Search

Wireless body area networks (WBAN) have improved healthcare industries to a large extent by providing contactless measurements and remote data analysis. However, the challenges encountered are mostly in the form of energy depletion scenarios, which results in the reduction of network lifetime to a large extent. This work presents an effective model to provide energy-efficient routing and enhanced energy harvesting mechanisms to improve network lifetime. The ant colony optimization (ACO) method has been extended to include a fitness function that takes into account several factors, and this is the basis for the routing model. These processes ensure effective routing, which conserves energy and, in turn, results in enhanced network lifetime. Performance of the proposed model has been compared with the existing state-of-the-art models in the domain. Comparison with the metaheuristic-based model, cooperative energy efficient and priority based reliable routing protocol with network coding (CEPRAN), indicates the efficiency of the energy harvesting mechanism used in the proposed work. When compared with models using energy harvesting mechanisms, results exhibit higher network lifetime, depicting the efficiency of the proposed routing mechanism.

Public transport network design deals with finding efficient network solution(s) from a set of alternatives that best satisfies the often-conflicting objectives of stakeholders like passengers and operators. This work presents a simulation-based optimization (SBO) model for designing public transport networks. The work’s novelty is in developing such a network design model that fully accounts for the stochastic behavior of commuters on the transit network. The SBO discipline solves decision-based problems like the transit network design problem (TNDP) by combining simulation and optimization models. The proposed model integrates a disaggregated activity-based travel demand simulation with a multi-objective network optimization algorithm. Trip-based travel demand models are commonly used to represent traveler behavior in the literature. The approach limits its ability to accommodate the stochastic realities of traveler behavior in a transit network design solution. Using activity-based simulation instead makes it possible to account for a more realistic traveler behavior, especially real-time decisions made in response to changing network dynamics which ultimately affect the distribution of demand over time on the network. The proposed model is applied to the improved design of the integrated public transport network in the City of Cape Town, South Africa. The results show SBO can design efficient network solutions that reflect the objectives of network stakeholders.

A metaheuristic approach based on the nature-inspired and well-known Grey Wolf Optimization algorithm (GWO) was employed in this study to design an approach for retrieving strong designs of 8×8 substitution boxes (S-boxes). The GWO was developed as a novel metaheuristic based on inspiration from grey wolves and how they hunt. The ability of the GWO to quickly explore the search space for the near/optimal feature subsets that maximize any given fitness function (in consideration of its distinctive hierarchical structure) aids in the construction of strong S-boxes that can satisfy the required criteria. However, when tackling optimization problems, GWO may experience the problem of premature convergence. Therefore, a variant of GWO called Crossover Grey Wolf Optimizer (XGWO) has been proposed in this study. The performance of the proposed novel approach was evaluated using numerous cryptographic performance metrics, including bijective property, bit independence, strict avalanche, linear probability, and I/O XOR distribution and the result was contrasted with a couple of existing S-box creation techniques. Overall, the results of the experiment showed that the suggested S-box design had adequate cryptographic features. Author

RESUMEN: El presente estudio muestra la potencialidad de uso de las técnicas metaheurísticas, para abordar problemas de toma de decisiones en contextos de gran complejidad y con aplicabilidad a una gran variedad de campos. Se presentan casos prácticos en el campo de la acuicultura y la gestión de granjas de cultivo de peces; así como, en la ingeniería inversa. Estos casos abordan problemas de optimización de gran complejidad por el alto número de variables y restricciones. Para resolver este tipo de problemas se presentan algunos métodos metaheurísticos implementados en Python, como herramientas útiles para conseguir soluciones satisfactorias. Las soluciones obtenidas han sido óptimas o cuasióptimas conseguidas en tiempos aceptables. ABSTRACT: This study shows the potential use of metaheuristic techniques to address decision-making problems in contexts of great complexity and with applicability to a wide variety of fields. Practical cases are presented in the field of aquaculture and the management of fish farms; as well as, in reverse engineering. These cases address highly complex optimization problems due to the high number of variables and restrictions. To solve this type of problems, some metaheuristic methods implemented in Python are presented as useful tools to achieve satisfactory solutions. The solutions obtained have been optimal or quasi-optimal achieved in acceptable times.

Requirements for information security dictate the necessity of developing new methods of cryptanalysis. Modern cryptanalysis depend on mathematics, in particular on theory and optimization methods. Taking into account the generally recognized requirements for attack resistance of ciphers, the decryption problem should be considered as a combinatorial optimization problem The paper proves the necessary of the development of new methods of cryptanalysis using metaheuristics, contains a retrospective review of publications in the last period in this area. The number of publications indicates the relevance of the research direction. Specialities of the application of the Ant Colony Optimization algorithm to cryptanalysis problems, in particular, factorization problem, are considered. The structure and general principles of the ACO algorithm are described, as well as the adaptation of this algorithm to the solution of a specific problem of combinatorial optimization. Various variants of the fitness function, features of their application, methods of narrowing the search space, rules for choosing the direction of movement on the graph, modification of local search are discussed. The addition of genetic operators of crossover, mutation, and selection is considered as one of the modification options. The conditions for stopping the operation of the algorithm are described. The various facts of using metaheuristics for solving combinatorial optimization problems arising in numerous subject areas, in particular, in cryptanalysis, are described. It is emphasized that since theoretical studies of combinatorial optimization algorithms rarely allow obtaining results that can be applied in practice. The main tool for analyzing their effectiveness is a computational experiment. Вимоги до інформаційної безпеки диктують неохідність розвитку нових методів криптоаналізу. Сучасний криптоаналіз спирається на математику, зокрема на теорію та методи оптимізації. Враховучи загальновизнані вимоги до зламостійкості шифрів, задача розшифрування мусить розглядатися, як задача комбінаторної оптимізації. В роботі обґрунтовується необхідність розвитку нових методів криптоаналізу із застосуванням метаевристик, міститься ретрспективний огляд публікацій за останній період в даній області. Кількість публікацій свідчить про актуальність напрямку досліджень. Розглядаються особливості застосування алгоритму АСО (Ant Colony Optimization) до задач криптоаналізу, зокрема, задачі факторизації. Описується структура і загальні принципи роботи алгоритму АСО, адаптація даного алгоритму до розв’язання конкретної задачі комбінаторної оптимізації. Розглянуто різні варіанти фітнес-функції, особливості їх застосування, способи звуження простору пошуку, правила вибору напрямку руху на графі, модифікація локального пошуку. Як один із варіантів модифікації розглядається додавання генетичних операторів кросоверу, мутації, селекції. Описано умови припинення роботи алгоритму. Обґрунтовано доцільність застосування метаевристик для розв’зання задач комбінаторної оптимізації що виникають у різних предметних областях, зокрема, у криптоаналізі. Підкреслюється, що так як теоретичні дослідження алгоритмів комбінаторної оптимізації рідко дозволяють отримувати результати, які можуть бути застосовані на практиці, то основним інструментом аналізу їх ефективності є обчислювальний експеримент.

Usmjeravanje vozila često je prisutan element mnogih industrija pa tako i morskih luka. Ovaj problem predstavlja ogroman logistički izazov zbog velike količine robe koja se mora prevesti. Kvalitetno rješenje navedenog problema može rezultirati velikom uštedom sredstava. Postoje različiti pristupi rješavanju ovog problema, no i dalje je ovo česta tema istraživanja zbog rastuće složenosti uzrokovane povećanjem industrija. Evolucijski algoritmi igraju veliku ulogu u ovom području jer nude obećavajuće rezultate. Međutim, i dalje ima puno prostora za poboljšanja i istraživanje novih, efektivnijih metoda. Ovaj rad opisuje problem usmjeravanja vozila u morskim lukama i predlaže prikladan evolucijski algoritam koji nastoji kreirati optimalan skup ruta za vozila. Opisana je korištena metoda te analizirani rezultati u usporedbi s relevantnim suvremenim pristupima. Vehicle routing is an often present element of many industries, including seaports. This problem presents a huge logistical challenge due to the large amount of goods that have to be transported. A high-quality solution to the above-mentioned problem can result in a large saving of funds. There are different approaches to solving this problem, but it is still a frequent topic of research due to the growing complexity caused by the increasing demands of industries. Evolutionary algorithms play a major role in this field as they offer promising results. However, there is still a lot of room for improvement and research into new, more effective methods. This paper describes the vehicle routing problem in seaports and proposes a suitable evolutionary algorithm that seeks to create an optimal set of vehicle routes. The method used is described and the results analyzed in comparison with relevant modern approaches.

After a disaster, often roads are damaged and blocked, hindering accessibility for relief efforts. It is essential to dispatch work teams to restore the blocked roads by clearance or repair operations. With the goal of enabling access between critical locations in the disaster area in shortest time, we propose algorithms that determine the schedule and routes of multiple work teams. We minimize the total latency of reaching the critical locations, where the latency of a location is defined as the time it takes from the start of the operation until its first visit by one of the work teams. Coordination among the teams is needed since some blocked edges might be opened by a certain team and utilized by other teams later on. First, we develop an exact mathematical model that handles the coordination requirement. After observing the intractability of this formulation, we introduce two heuristic methods and a lower bounding procedure. In the first method, we develop a mathematical model based on a novel multi-level network representation that yields solutions with disjoint paths. Given that it does not coordinate the teams, we present a matheuristic based on a cluster-first-route-second approach embedded into a local search algorithm together with an additional coordination step to obtain alternative solutions with higher quality and in a shorter time. We test our heuristics on data sets coming from a real network from the literature (180 instances) and randomly generated ones (640 instances) and observe the superiority of the solutions obtained by incorporation of coordination., NA

Metaheuristics are found to be efficient in different applications where the use of exact algorithms becomes short-handed. In the last decade, many of these algorithms have been introduced and used in a wide range of applications. Nevertheless, most of those approaches share similar components leading to a concern related to their novelty or contribution. Thus, in this paper, a pool template is proposed and used to categorize algorithm components permitting to analyze them in a structured way. We exemplify its use by means of continuous optimization metaheuristics, and provide some measures and methodology to identify their similarities and novelties. Finally, a discussion at a component level is provided in order to point out possible design differences and commonalities.

Feature selection (FS) is a pre-processing step that aims to eliminate the redundant and less-informative features to enhance the performance of data mining techniques. It is also considered as one of the key success factors for classification problems in high-dimensional datasets. This paper proposes an efficient wrapper feature selection method based on Grey Wolf Optimizer (GWO). GWO is a recent metaheuristic algorithm that has been widely employed to solve diverse optimization problems. However, GWO mainly follows the search directions toward the leading wolves, making it prone to fall into local optima, especially when dealing with high-dimensional problems, which is the case when dealing with many biological datasets. An enhanced variation of GWO called EGWO, which adapts two enhancements, is introduced to overcome this specific shortcoming. In the first place, the transition parameter concept is incorporated to move GWO from the exploration phase to the exploitation phase. Several adaptive non-linear decreasing formulas are introduced to control the transition parameters. In the second place, a random-based search strategy is exploited to empower diversity during the search process. Two binarization schemes using S-shaped and V-shaped transfer functions are incorporated to map the continuous search space into a binary one for FS. The efficiency of the proposed EGWO is validated on ten high-dimensional low-samples biological data. Our experiments show the promising performance of EGWO compared to the original GWO approach and other state-of-the-art techniques in terms of dimensionality reduction and the enhancement of classification performance.

Differential evolution (DE) is a popular evolutionary algorithm inspired by Darwin’s theory of evolution and has been studied extensively to solve different areas of optimisation and engineering applications since its introduction by Storn in 1997. This study aims to review the massive progress of DE in the research community by analysing the 192 articles published on this subject from 1997 to 2021, particularly studies in the past five years. The methodology used to search for relevant DE papers and an overview of the original DE are firstly explained. Recent advances in the modifications proposed to enhance the effectiveness and efficiency of the original DE are reviewed by analysing the strengths and weaknesses of each published work, followed by the potential applications of these DE variants in solving different real-world engineering problems. In contrast to most existing DE review papers, additional analyses are performed in this survey by investigating the impacts of various parameter settings on given DE variants to identify their optimal values required for solving certain problem classes. The qualities of modifications incorporated into selected DE variants are also evaluated by measuring the performance gains achieved in terms of search accuracy and/or efficiency against the original DE. The additional surveys conducted in this study are anticipated to provide more insightful perspectives for both beginners and experts of DE research, enabling their better understanding about current research trends and new motivations to outline appropriate strategic planning for future development works.

In this research, a hybrid algorithm called VMC_2OptBI_r was developed, which improved the solution search criteria applied by the 2Opt Exchange Heuristic based on a node exchange policy called Best Improvement (BI) and introducing an additional factor "1+r" to perform the 2Opt exchanges. The "1+r" factor allows slightly modifying the selection of nodes to be exchanged, thus exploring new solutions. To measure the level of efficacy of the new algorithm, instances of the Symmetric Traveling Salesman Problem from TSPLIB were selected and compared with their basic versions: Nearest Neighbor (VMC) and VMC_2OptBI. Then, they were compared with the solutions of the SCA_2Opt and SCA_2Opt_r algorithms, where they were compared with the solutions of four metaheuristics published in recent articles (2019-2022). The results showed that the solutions obtained by the new VMC_2OptBI_r algorithm achieved an efficacy level of 100% compared to VMC, VMC_2OptBI, SCA_2Opt, a value higher than 81% compared to SCA_2Opt_r, and a range between 11% to 88% compared to the four reviewed metaheuristics for the compared instances., En esta investigación se desarrolló un algoritmo híbrido denominado VMC_2OptBI_r, que a partir de la solución inicial construida bajo el pensamiento del Vecino más Cercano se buscó mejorar el criterio de búsqueda de soluciones aplicado por la Heurística de Intercambio 2Opt basada en una política de intercambio de nodos denominada Best Improvement (BI) e introduciendo un factor adicional “1+r” al criterio para realizar los intercambios 2Opt. El factor “1+r” permite modificar ligeramente la selección de nodos a intercambiar provocando así la exploración de nuevas soluciones. Para medir el nivel de eficacia del nuevo algoritmo se seleccionaron instancias del problema del agente viajero simétrico de TSPLIB, las cuales en primer lugar se compararon con sus versiones básicas: El Vecino más Cercano (VMC) y el VMC_2OptBI; luego se compararon con las soluciones de los algoritmos SCA_2Opt, SCA_2Opt_r, donde se comparó con las soluciones de cuatro metaheurísticas publicadas en artículos recientes (2019-2022). Los resultados mostraron que las soluciones obtenidas por el nuevo algoritmo VMC_2OptBI_r alcanzaron un nivel de eficacia de 100% con respecto al VMC, VMC_2OptBI, SCA_2Opt, un valor mayor al 81% con respecto al SCA_2Opt_r y un rango entre el 11% al 88% con respecto a las cuatro metaheurísticas revisadas para las instancias comparadas.

Neural networks have revolutionised the way we approach problem solving across multiple domains; however, their effective design and efficient use of computational resources is still a challenging task. One of the most important factors influencing this process is model hyperparameters which vary significantly with models and datasets. Recently, there has been an increased focus on automatically tuning these hyperparameters to reduce complexity and to optimise resource utilisation. From traditional human-intuitive tuning methods to random search, grid search, Bayesian optimisation, and evolutionary algorithms, significant advancements have been made in this direction that promise improved performance while using fewer resources. In this article, we propose HyperGE, a two-stage model for automatically tuning hyperparameters driven by grammatical evolution (GE), a bioinspired population-based machine learning algorithm. GE provides an advantage in that it allows users to define their own grammar for generating solutions, making it ideal for defining search spaces across datasets and models. We test HyperGE to fine-tune VGG-19 and ResNet-50 pre-trained networks using three benchmark datasets. We demonstrate that the search space is significantly reduced by a factor of ~90% in Stage 2 with fewer number of trials. HyperGE could become an invaluable tool within the deep learning community, allowing practitioners greater freedom when exploring complex problem domains for hyperparameter fine-tuning.

International audience; Reconfigurable manufacturing systems (RMS) intend to bridge the gap between dedicated and flexible manufacturing systems. If the literature is mainly focused on the design step and tactical planning of such systems, few research projects have addressed scheduling at the operational level. While setup times may occur in flexible manufacturing systems, reconfiguration times considered in RMS may affect several resources at once, and hence require specific modelling and solving approaches to be considered. This paper first formalises the problem at hand through integer linear programming. An iterative search method is then provided to obtain solutions to larger-scale instances. Results obtained on generated instances show that managing even few possible configurations can yield significant improvements in solutions’ quality. Meanwhile, the extended search space implied by the increase in available configurations hinders the convergence to a good solution in a reasonable computation time, which suggests further investigations.

In this paper, we propose a new approach to dynamic wireless charging that allows electric vehicles to charge wirelessly while in motion in both lanes on highways. The challenge is to locate the charging infrastructure on a highway between origin O and destination S (round trip) with heterogeneous battery vehicles, where each type of vehicle requires its allocation of charging segments on the road. We aim to ensure that each type of vehicle can complete a round trip without running out of battery charge while minimizing the number of charging segments and inverters on the road by studying both lanes simultaneously. We model the problem mathematically and validate it using a CPLEX optimizer for limited instances. Finally, we solve the problem using a hybrid approach that combines genetic algorithms and local search techniques to balance diversification and intensification. We have significantly improved the results found in the literature by reducing the number of inverters, which are expensive components in the charging infrastructure. Our approach takes advantage of utilizing a single inverter for both lanes of the highway, leading to cost savings and improved efficiency.

In recent years, metaheuristic methods have shown remarkable efficacy in resolving complex combinatorial challenges across a broad spectrum of fields. Nevertheless, the escalating complexity of these problems necessitates the continuous development of innovative techniques to enhance the performance and reliability of these methods. This paper aims to contribute to this endeavor by examining the impact of solution initialization methods on the performance of a hybrid algorithm applied to the set union knapsack problem (SUKP). Three distinct solution initialization methods, random, greedy, and weighted, have been proposed and evaluated. These have been integrated within a sine cosine algorithm employing k-means as a binarization procedure. Through testing on medium- and large-sized SUKP instances, the study reveals that the solution initialization strategy influences the algorithm’s performance, with the weighted method consistently outperforming the other two. Additionally, the obtained results were benchmarked against various metaheuristics that have previously solved SUKP, showing favorable performance in this comparison.

This paper proposes an enhanced Particle Swarm Optimization (PSO) algorithm in order to deal with the issue that the time and cost of the PSO algorithm is quite high when scheduling workflow tasks in a cloud computing environment. To reduce particle dimensions and ensure initial particle quality, intensive tasks are combined when scheduling workflow tasks. Next, the particle initialization is optimized to ensure better initial particle quality and reduced search space. Then, a suitable self-adaptive function is integrated to determine the best direction of the particles. The experiments show that the proposed enhanced PSO algorithm has better convergence speed and better performance in the execution of workflow tasks.

This paper presents a new hybrid algorithm that combines genetic algorithms (GAs) and the optimizing spotted hyena algorithm (SHOA) to solve the production shop scheduling problem. The proposed GA-SHOA algorithm incorporates genetic operators, such as uniform crossover and mutation, into the SHOA algorithm to improve its performance. We evaluated the algorithm on a set of OR library instances and compared it to other state-of-the-art optimization algorithms, including SSO, SCE-OBL, CLS-BFO and ACGA. The experimental results show that the GA-SHOA algorithm consistently finds optimal or near-optimal solutions for all tested instances, outperforming the other algorithms. Our paper contributes to the field in several ways. First, we propose a hybrid algorithm that effectively combines the exploration and exploitation capabilities of SHO and GA, resulting in a balanced and efficient search process for finding near-optimal solutions for the FSSP. Second, we tailor the SHO and GA methods to the specific requirements of the FSSP, including encoding schemes, objective function evaluation and constraint handling, which ensures that the hybrid algorithm is well suited to address the challenges posed by the FSSP. Third, we perform a comprehensive performance evaluation of the proposed hybrid algorithm, demonstrating its effectiveness in terms of solution quality and computational efficiency. Finally, we provide an in-depth analysis of the behavior of the hybrid algorithm, discussing the roles of the SHO and GA components and their interactions during the search process, which can help understand the factors contributing to the success of the algorithm and provide insight into potential improvements or adaptations to other combinatorial optimization problems.

Stochastic, as well as fuzzy uncertainty, can be found in most real-world systems. Considering both types of uncertainties simultaneously makes optimization problems incredibly challenging. In this paper we propose a fuzzy simheuristic to solve the Time Capacitated Arc Routing Problem (TCARP) when the nature of the travel time can either be deterministic, stochastic or fuzzy. The main goal is to find a solution (vehicle routes) that minimizes the total time spent in servicing the required arcs. However, due to uncertainty, other characteristics of the solution are also considered. In particular, we illustrate how reliability concepts can enrich the probabilistic information given to decision-makers. In order to solve the aforementioned optimization problem, we extend the concept of simheuristic framework so it can also include fuzzy elements. Hence, both stochastic and fuzzy uncertainty are simultaneously incorporated into the CARP. In order to test our approach, classical CARP instances have been adapted and extended so that customers' demands become either stochastic or fuzzy. The experimental results show the effectiveness of the proposed approach when compared with more traditional ones. In particular, our fuzzy simheuristic is capable of generating new best-known solutions for the stochastic versions of some instances belonging to the tegl, tcarp, val, and rural benchmarks. This work has been partially supported by the Spanish Ministry of Science (PID2019-111100RB-C21/AEI/10.13039/01100011033), as well as by the Barcelona Council and the “laCaixa” Foundation under the framework of the Barcelona Science Plan 2020-2023 (grant21S09355-01) and Generalitat Valenciana (PROMETEO/2021/065).

A Cyber-Physical System (CPS) is a network of cyber and physical elements that interact with each other. In recent years, there has been a drastic increase in the utilization of CPSs, which makes their security a challenging problem to address. Intrusion Detection Systems (IDSs) have been used for the detection of intrusions in networks. Recent advancements in the fields of Deep Learning (DL) and Artificial Intelligence (AI) have allowed the development of robust IDS models for the CPS environment. On the other hand, metaheuristic algorithms are used as feature selection models to mitigate the curse of dimensionality. In this background, the current study presents a Sine-Cosine-Adopted African Vultures Optimization with Ensemble Autoencoder-based Intrusion Detection (SCAVO-EAEID) technique to provide cybersecurity in CPS environments. The proposed SCAVO-EAEID algorithm focuses mainly on the identification of intrusions in the CPS platform via Feature Selection (FS) and DL modeling. At the primary level, the SCAVO-EAEID technique employs Z-score normalization as a preprocessing step. In addition, the SCAVO-based Feature Selection (SCAVO-FS) method is derived to elect the optimal feature subsets. An ensemble Deep-Learning-based Long Short-Term Memory–Auto Encoder (LSTM-AE) model is employed for the IDS. Finally, the Root Means Square Propagation (RMSProp) optimizer is used for hyperparameter tuning of the LSTM-AE technique. To demonstrate the remarkable performance of the proposed SCAVO-EAEID technique, the authors used benchmark datasets. The experimental outcomes confirmed the significant performance of the proposed SCAVO-EAEID technique over other approaches with a maximum accuracy of 99.20%.

Intrusion detection systems (IDS) play a crucial role in securing networks and identifying malicious activity. This is a critical problem in cyber security. In recent years, metaheuristic optimization algorithms and deep learning techniques have been applied to IDS to improve their accuracy and efficiency. Generally, optimization algorithms can be used to boost the performance of IDS models. Deep learning methods, such as convolutional neural networks, have also been used to improve the ability of IDS to detect and classify intrusions. In this paper, we propose a new IDS model based on the combination of deep learning and optimization methods. First, a feature extraction method based on CNNs is developed. Then, a new feature selection method is used based on a modified version of Growth Optimizer (GO), called MGO. We use the Whale Optimization Algorithm (WOA) to boost the search process of the GO. Extensive evaluation and comparisons have been conducted to assess the quality of the suggested method using public datasets of cloud and Internet of Things (IoT) environments. The applied techniques have shown promising results in identifying previously unknown attacks with high accuracy rates. The MGO performed better than several previous methods in all experimental comparisons.

In this research, the effectiveness of a novel optimizer dubbed as decomposition-based multi-objective symbiotic organism search (MOSOS/D) for multi-objective problems was explored. The proposed optimizer was based on the symbiotic organisms’ search (SOS), which is a star-rising metaheuristic inspired by the natural phenomenon of symbioses among living organisms. A decomposition framework was incorporated in SOS for stagnation prevention and its deep performance analysis in real-world applications. The investigation included both qualitative and quantitative analyses of the MOSOS/D metaheuristic. For quantitative analysis, the MOSOS/D was statistically examined by using it to solve the unconstrained DTLZ test suite for real-parameter continuous optimizations. Next, two constrained structural benchmarks for real-world optimization scenario were also tackled. The qualitative analysis was performed based on the characteristics of the Pareto fronts, boxplots, and dimension curves. To check the robustness of the proposed optimizer, comparative analysis was carried out with four state-of-the-art optimizers, viz., MOEA/D, NSGA-II, MOMPA and MOEO, grounded on six widely accepted performance measures. The feasibility test and Friedman’s rank test demonstrates the dominance of MOSOS/D over other compared techniques and exhibited its effectiveness in solving large complex multi-objective problems.

Planning problems are particularly important for the production processes of precast reinforced concrete elements. Currently used modeling of these processes is based on the flow shop problem. Flow shop models are usually used in Enterprise Resource Planning systems, which, however, may not take into account the specifics of the production of such elements. The article presents a new model for scheduling the production of reinforced concrete prefabricated elements, which is distinguished by the possibility of carrying out activities by more than one working group. An additional new constraint is the possibility of parallel performance of some works, which may occur during their production. Also, there will be an individual order of elements assumed for each of the activities. New objective functions will be considered – the sum of idle times of working groups and the total type changes of precast components. The presented scheduling model contains an NP-hard discrete optimization problem. For this reason, metaheuristics were used in the article to solve optimization problems: the simulated annealing algorithm and the tabu search algorithm. Verification of the results obtained with the use of these algorithms confirmed their high efficiency. The application of the presented scheduling model illustrates a practical case study showing the effectiveness of the used algorithms.

The location routing problem (LRP) integrates operational decisions on vehicle routing operations with strategic decisions on the location of the facilities or depots from which the distribution will take place. In other words, it combines the well-known vehicle routing problem (VRP) with the facility location problem (FLP). Hence, the LRP is an NP-hard combinatorial optimization problem, which justifies the use of metaheuristic approaches whenever large-scale instances need to be solved. In this paper, we explore a realistic version of the LRP in which facilities of different capacities are considered, i.e., the manager has to consider not only the location but also the size of the facilities to open. In order to tackle this optimization problem, three mixed-integer linear formulations are proposed and compared. As expected, they have been proved to be cost- and time- inefficient. Hence, a biased-randomized iterated local search algorithm is proposed. Classical instances for the LRP with homogeneous facilities are naturally extended to test the performance of our approach.

The Multidimensional Multiple-choice Knapsack Problem (MMKP) is a complex combinatorial optimization problem for which finding high quality feasible solutions is very challenging. Recently, several heuristic approaches and a few exact algorithms have been proposed for its solution. These methods have been able to provide new best-known values for benchmark instances although many of them still remain unclosed to optimality. In this paper, we provide a new variant of the heuristic framework Kernel Search and apply it to the MMKP. The proposed variant keeps the method’s main idea of solving a sequence of restricted mixed-integer subproblems but innovates by partitioning the solution process into two different phases with complementary goals. The first phase strives for feasibility and collects important information to dynamically adapt subproblems’ dimension and solution time in the second phase that is focused on getting high quality solutions. This makes the global approach more scalable and efficient. Computational results on different sets of benchmark instances demonstrate that our method is extremely effective outperforming all state-of-the-art heuristics for the MMKP. The method compares extremely well also with respect to exact approaches running for five hours. The proposed algorithm is able to improve the best-known value of 185 out of 276 open benchmark instances and gets 8 out of 17 optimal solutions for the closed ones. The average deviation from optimality is always negligible.

Big data in healthcare defines a massive quantity of healthcare data accumulated from massive sources like electronic health records (EHR), medical imaging, genomic sequence, pharmacological research, wearable, medical gadgets, etc. One of the data mining approaches commonly employed to classify big data is the MapReduce model. Data clustering, a significant data mining technique has been extensively investigated in recent years in handling the diversity in data and various sets of application necessities. In this view, this paper develops an enhanced metaheuristic algorithm based clustering and classification model with MapReduce (EMACC-MR) framework for big data environment. The presented EMACC-MR model involves an oppositional cuckoo search optimization algorithm (OCSOA) with Density-Based Spatial Clustering of Applications with Noise (DBSCAN) and wavelet kernel extreme learning machine (WKELM) model based on the classification model. The inclusion of oppositional based learning (OBL) concept helps to improve the convergence rate of the CSOA. For handling big data, the Hadoop MapReduce environment is employed. The proposed OCSOA model improves the clustering quality and MapReduce architecture to cope with the large-scale dataset. For validating the experimental analysis of the proposed model, two benchmark datasets namely Activity recognition and diabetes datasets are used. The simulation outcomes confirmed that the presented model outperforms the compared methods in terms of several evaluation parameters.

This theoretical research study proposes a novel Chaotic Quasi-Oppositional Arithmetic Optimization Algorithm (COAOA) for thermo-economic optimization of a shell and tube condenser working with refrigerant mixtures. Arithmetic Optimization Algorithm (AOA) is a recently emerged metaheuristic algorithm considering different mathematical operators to optimize the candidate solutions over a wide range of search domains. The effectiveness the COAOA is assessed by applying it to a set of benchmark optimization problems and comparing the obtained solutions with that of the original AOA and its quasi-oppositional variant. The COAOA has been employed to acquire the minimum value of the total annual cost of the shell and tube condenser by iteratively varying nine decision variables of mass flow rate, shell diameter, the tube inside diameter, tube length, number of tube passes, tube layout, tube pitch ratio, the total number of baffles, and diameter ratio. Three different case studies are solved using different refrigerant pairs used for in-tube flow to show the proposed metaheuristic optimizer's efficiency and effectivity on real-world mixed-integer optimization problem. Optimal results retrieved for different mixture pairs with varying mass fractions are compared with each other, and parametric configuration yielding the minimum total cost is decided. Finally, a comprehensive sensitivity analysis is performed to investigate the influences of the design variables over the considered problem objective. Overall analysis results indicate that COAOA can be an excellent optimizer to obtain a shell and tube condenser's optimal configuration within a reasonable computation time.

In recent years, the power demonstrated by Machine Learning (ML) has increasingly attracted the interest of the optimization community that is starting to leverage ML for enhancing and automating the design of algorithms. One combinatorial optimization problem recently tackled with ML is the Job Shop scheduling Problem (JSP). Most of the works on the JSP using ML focus on Deep Reinforcement Learning (DRL), and only a few of them leverage supervised learning techniques. The recurrent reasons for avoiding supervised learning seem to be the difficulty in casting the right learning task, i.e., what is meaningful to predict, and how to obtain labels. Therefore, we first propose a novel supervised learning task that aims at predicting the quality of machine permutations. Then, we design an original methodology to estimate this quality, and we use these estimations to create an accurate sequential deep learning model (binary accuracy above 95%). Finally, we empirically demonstrate the value of predicting the quality of machine permutations by enhancing the performance of a simple Tabu Search algorithm inspired by the works in the literature.

Currently, the industry is experiencing an exponential increase in dealing with binary-based combinatorial problems. In this sense, metaheuristics have been a common trend in the field in order to design approaches to solve them successfully. Thus, a well-known strategy consists in the use of algorithms based on discrete swarms transformed to perform in binary environments. Following the No Free Lunch theorem, we are interested in testing the performance of the Fruit Fly Algorithm, this is a bio-inspired metaheuristic for deducing global optimization in continuous spaces, based on the foraging behavior of the fruit fly, which usually has much better sensory perception of smell and vision than any other species. On the other hand, the Set Coverage Problem is a well-known NP-hard problem with many practical applications, including production line balancing, utility installation, and crew scheduling in railroad and mass transit companies. In this paper, we propose different binarization methods for the Fruit Fly Algorithm, using S-shaped and V-shaped transfer functions and various discretization methods to make the algorithm work in a binary search space. We are motivated with this approach, because in this way we can deliver to future researchers interested in this area, a way to be able to work with continuous metaheuristics in binary domains. This new approach was tested on benchmark instances of the Set Coverage Problem and the computational results show that the proposed algorithm is robust enough to produce good results with low computational cost.

Inspired by an automated teller machine (ATM) cash replenishment problem involving population coverage requirements (PCRs) in the Netherlands, we propose the vehicle tour problem with minimum coverage requirements. In this problem, a set of minimum-cost routes is constructed subject to constraints on the duration of each route and the population coverage of the replenished ATMs. A compact formulation incorporating a family of valid inequalities and an efficient tour-splitting metaheuristic are proposed and tested on 77 instances derived from real-life data involving up to 98 ATMs and 237,604 citizens and on 144 newly generated synthetic instances. Our results for the real-life instances indicate significant cost differences in replenishing ATMs for seven major Dutch cities when the PCRs vary. Additionally, we illustrate the impact of different PCRs on the ATM replenishment costs for seven major cities in the Netherlands by presenting an aggregated cost evaluation of 11 PCRs involving 1,003,519 citizens, 338 ATMs, and 19 cash distribution vehicles.

This paper examines a permutation flow-shop scheduling problem, which is a complex combinatorial problem encountered in many practical applications. The objective of the research is to reduce the maximum completion time, i.e., the makespan of all jobs. In order to increase productivity and to meet the demand, manufacturers are continuously under pressure to attain the shortest possible completion time. Estimation of accurate cycle time can tremendously help production planning and scheduling in manufacturing industries. Since production planning is characterised by NP-hardness and a wide range, traditional optimization methods and heuristic rules are unable to find satisfactory solutions. Q-learning, a type of reinforcement learning algorithm, is used in this paper to find a solution that is close to being optimal. Q-learning is a branch of machine learning referring to the way an intelligent agent should act in order to maximize the concept of cumulative reward in a given environment. To validate the performance of the algorithm, Taillard’s benchmark problems were solved and compared with the upper bound value. The results showed that the performance of the algorithm is better and has low computational time. Based on the performance of the proposed algorithm, two case studies were done and the solutions are compared with the performance of a metaheuristic algorithm. The result shows that the proposed algorithm can effectively and efficiently solve the problem stated above and that it is an interesting solution to resolving complex scheduling problems.

The Quadratic Assignment Problem (QAP) is one of the classical combinatorial optimization problems and is known for its diverse applications. The QAP is an NP-hard optimization problem which attracts the use of heuristic or metaheuristic algorithms that can find quality solutions in an acceptable computation time. On the other hand, there is quite a broad spectrum of mathematical programming techniques that were developed for finding the lower bounds for the QAP. This paper presents a fusion of the two approaches whereby the solutions from the computations of the lower bounds are used as the starting points for a metaheuristic, called HC12, which is implemented on a GPU CUDA platform. We perform extensive computational experiments that demonstrate that the use of these lower bounding techniques for the construction of the starting points has a significant impact on the quality of the resulting solutions.

With the advent of Service-Oriented Architecture (SOA), services can be registered, invoked, and combined by their identical Quality of Services (QoS) attributes to create a new value-added application that fulfils user requirements. Efficient QoS-aware service composition has been a challenging task in cloud computing. This challenge becomes more formidable in emerging resource-constrained computing paradigms, such as the Internet of Things and Fog. Service composition has regarded as a multi-objective combinatorial optimization problem that falls in the category of NP-hard. Historically, the proliferation of services added to problem complexity and navigated solutions from exact (none-heuristics) approaches to near-optimal heuristics and metaheuristics. Although metaheuristics have fulfilled some expectations, the quest for finding a high-quality, near-optimal solution has led researchers to devise hybrid methods. As a result, research on service composition shifts towards the hybridization of metaheuristics. Hybrid metaheuristics have been promising efforts to transcend the boundaries of metaheuristics by leveraging the strength of complementary methods to overcome base algorithm shortcomings. Despite the significance and frontier position of hybrid metaheuristics, to the best of our knowledge, there is no systematic research and survey in this field with a particular focus on strategies to hybridize traditional metaheuristics. This study’s core contribution is to infer a framework for hybridization strategies by conducting a mapping study that analyses 71 papers between 2008 and 2020. Moreover, it provides a panoramic view of hybrid methods and their experiment setting in respect to the problem domain as the main outcome of this mapping study. Finally, research trends, directions and challenges are discussed to benefit future endeavours.

An improved variant of the Jaya optimization algorithm, called Jaya2, is proposed to enhance the performance of the original Jaya sacrificing its algorithmic design. The proposed approach arranges the solutions in a ring topology to reduce the likelihood of premature convergence. In addition, the population size reduction is used to automatically adjust the population size during the optimization process. Moreover, the translation dependency problem of the original Jaya is discussed, and an alternative solution update operation is proposed. To test Jaya2, we compare it with nine different optimization methods on the CEC 2020 benchmark functions and the CEC 2011 real-world optimization problems. The results show that Jaya2 is highly competitive on the tested problems where it generally outperforms most approaches. Having an easy-to-implement approach with little parameter tuning is highly desirable since researchers from different disciplines with basic programming skills can use it to solve their optimization problems.

In this study, we present a hybrid approach of Ant Colony Optimization algorithm (ACO) with fuzzy logic and clustering methods to solve multiobjective path planning problems in the case of swarm Unmanned Surface Vehicles (USVs). This study aims to further explore the performance of the ACO algorithm by integrating fuzzy logic in order to cope with the multiple contradicting objectives and generate quality solutions by in-parallel identifying the mission areas of each USV to reach the desired targets. The design of the operational areas for each USV in the swarm is performed by a comparative evaluation of three popular clustering algorithms: Mini Batch K-Means, Ward Clustering and Birch. Following the identification of the operational areas, the design of each USV path to perform the operation is performed based on the minimization of traveled distance and energy consumption, as well as the maximization of path smoothness. To solve this multiobjective path planning problem, a comparative evaluation is conducted among ACO and fuzzy inference systems, Mamdani (ACO-Mamdani) and Takagi–Sugeno–Kang (ACO-TSK). The results show that depending on the needs of the application, each methodology can contribute, respectively. ACO-Mamdani generates better paths, but ACO-TSK presents higher computation efficiency.

Optimal power flow (OPF) is a complex, highly nonlinear, NP-hard optimization problem, in which the goal is to determine the optimal operational parameters of a power-related system (in many cases a type of smart or micro grid) which guarantee an economic and effective power dispatch. In recent years, a number of approaches based on metaheuristics algorithms have been proposed to solve OPF problems. In this paper, we propose the use of the Cross-Entropy (CE) method as a first step depth search operator to assist population-based evolutionary methods in the framework of an OPF problem. Specifically, a new variant of the Coral Reefs Optimization with Substrate Layers algorithm boosted with CE method (CE+CRO-SL) is presented in this work. We have adopted the IEEE 57-Bus System as a test scenario which, by default, has seven thermal generators for power production for the grid. We have modified this system by replacing three thermal generators with renewable source generators, in order to consider a smart grid approach with renewable energy production. The performance of CE+CRO-SL in this particular case study scenario has been compared with that of well-known techniques such as population?s methods CMA-ES and EPSO (both boosted with CE). The results obtained indicate that CE+CRO-SL showed a superior performance than the alternative techniques in terms of efficiency and accuracy. This is justified by its greater exploration capacity, since it has internally operations coming from different heuristics, thus surpassing the performance of classic methods. Moreover, in a projection analysis, the CE+CRO-SL provides a profit of millions of dollars per month in all cases tested considering the modified version of the IEEE 57-Bus smart grid system., European Commission

Post weld heat treatment, or PWHT, is often used to improve the mechanical properties of materials that have been welded. Several publications have investigated the effects of the PWHT process using experimental designs. However, the modeling and optimization using the integration of machine learning (ML) and metaheuristics have yet to be reported, which are fundamental steps toward intelligent manufacturing applications. This research proposes a novel approach using ML techniques and metaheuristics to optimize PWHT process parameters. The goal is to determine the optimal PWHT parameters for both single and multiple objective perspectives. In this research, support vector regression (SVR), K-nearest neighbors (KNN), decision tree (DT), and random forest (RF) were ML techniques employed to obtain a relationship model between PWHT parameters and mechanical properties: ultimate tensile strength (UTS) and elongation percentage (EL). The results show that the SVR demonstrated superior performance among ML techniques for both UTS and EL models. Then, SVR is used with metaheuristics such as differential evolution (DE), particle swarm optimization (PSO), and genetic algorithms (GA). SVR-PSO shows the fastest convergence among other combinations. The final solutions of single-objective and Pareto solutions were also suggested in this research.

In this paper, we propose an algorithmic framework to automatically generate efficient deep neural networks and optimize their associated hyperparameters. The framework is based on evolving directed acyclic graphs (DAGs), defining a more flexible search space than the existing ones in the literature. It allows mixtures of different classical operations: convolutions, recurrences and dense layers, but also more newfangled operations such as self-attention. Based on this search space we propose neighbourhood and evolution search operators to optimize both the architecture and hyper-parameters of our networks. These search operators can be used with any metaheuristic capable of handling mixed search spaces. We tested our algorithmic framework with an evolutionary algorithm on a time series prediction benchmark. The results demonstrate that our framework was able to find models outperforming the established baseline on numerous datasets.

This work proposes a smart protection system for microgrids, which relies on an adaptive metaheuristic for the automatic calculation of optimal settings for directional overcurrent relays (DOCRs). The adaptive fuzzy directional bat algorithm (AFDBA) associated with a fuzzy inference system (FIS) is used for this purpose. A prominent advantage of this solution is that there is no need for an initial tuning of the parameters associated with the algorithm, unlike many traditional approaches reported in the literature. Such a metaheuristic is used in the conception of an adaptive protection system (APS) in the context of a microgrid while taking into account the connection status of distributed generation (DG) units under distinct scenarios. A performance comparison with a protection system with fixed optimal settings (PSFOS) is also presented. The results demonstrate that the proposed APS outperforms the PSFOS while providing faster response, higher reliability and less susceptibility to miscoordination. In other words, it presents a shorter trip time when compared with the PSFOS, with a reduction of 6.83% and 26.58% when considering the DG penetration and the islanded microgrid, respectively.

In this paper, we propose an algorithmic framework to automatically generate efficient deep neural networks and optimize their associated hyperparameters. The framework is based on evolving directed acyclic graphs (DAGs), defining a more flexible search space than the existing ones in the literature. It allows mixtures of different classical operations: convolutions, recurrences and dense layers, but also more newfangled operations such as self-attention. Based on this search space we propose neighbourhood and evolution search operators to optimize both the architecture and hyper-parameters of our networks. These search operators can be used with any metaheuristic capable of handling mixed search spaces. We tested our algorithmic framework with an evolutionary algorithm on a time series prediction benchmark. The results demonstrate that our framework was able to find models outperforming the established baseline on numerous datasets.

Email is a commonly available communication technology used to share information among people via the Internet. But the drastic upsurge in email misuses/abuses has led to a rising quantity of spam emails in recent times. Spam email classification by the use of data mining and machine learning (ML) models has gained significant attention among researchers owing to the positive effect on saving Internet users. Different ML and feature selection (FS) techniques can be employed to design effective email spam detection and classification approaches. In this aspect, this paper devises a novel feature subset selection with deep learning-based email spam detection and classification (FSSDL-ESDC) technique. The FSSDL-ESDC technique encompasses two major processes namely tokenization and stops word removal. In addition, a feature selection approach based on fruitfly optimization (FFO) is used to find an optimum subset of characteristics. Furthermore, for the categorization of email spam, the bidirectional long short-term memory (BiLSTM) approach is applied. In order to boost the email spam detection performance of the BiLSTM model, grasshopper optimization algorithm (GOA) is applied to finely tune the hyper parameters of the BiLSTM model. The improved performance of the FSSDL-ESDC approach is shown by a rigorous simulation study. The experimental results demonstrated that the FSSDL-ESDC approach outperformed the other state-of-the-art procedures.

This thesis addresses two distinct problems: Professional Timetabling and Recommender Systems. Both issues take place at Mandarine Academy, a french educational technology company that specializes in innovative corporate training techniques such as personalized e-learning platforms, training logistics, web conferences, etc. We start with the timetabling problem related to handling training logistics, the company proposes a tool named “Dileap Logistic” that automatically handles the assignment of resources (rooms, teachers, and equipment) to time-slots (days/hours) at specific locations. Previously, this was done manually, which took a long time and resulted in inaccurate results (wrong trainers, days, or equipment). Despite being an NP-Complete problem, a mathematical formulation of the problem has been developed after reviewing both literature works and company requirements. It includes 18 constraints (hard/soft) and 5 objectives, two of which are competing. We evaluate five state of the art Multi-Objective Evolutionary Algorithms (MOEA's) starting with the non-dominated sorting genetic algorithm (NSGA II and NSGA III), the Multi-objective Evolutionary Algorithm Based on Decomposition (MOEA/D), the indicator-based Evolutionary Algorithm (IBEA) and finally the Strength Pareto Evolutionary Algorithm (SPEA 2). Two custom genetic operators (for Mutation and Crossover) were proposed and compared against classical operators (PMX and Swap mutation). A tuning phase involving all the above-mentioned algorithms is performed to obtain elite configurations. Experiments are divided by each problem size (small, medium, and large instances) with 3 to 5 objectives tested. We discuss results such as comparing the performance of each algorithm using various metrics, as well as convergence graphs and population evolution. Aside from proposing solutions to training logistics, the company creates online pedagogical content (videos, quizzes, documents, etc.) on daily basis to support the digitization of work environments and to keep up with current trends. With more than 550K users spread across 100 active e-learning platforms. However, with an overgrowing catalog, problems like information overload can increase the lack of motivation among users. Mandarine Academy provided us with access to one of its public e-learning platform to conduct our research. After literature review and extended data analysis of user interactions. We propose enhancements to the overall user experience and interface. Similar steps taken in the previous timetabling work are also applied on the recommendation engine. This includes the mathematical formulation of a multi-objective optimization problem having 5 objectives (Similarity, Diversity, Novelty, Root Mean Square Error (RMSE), and Normalized Discounted Cumulative Gain (NDCG)) and the implementation of different recommendation techniques to generate initial solutions. All of the above-mentioned evolutionary algorithms (NSGA II, NSGAIII, IBEA, SPEA2, and MOEA/D) are assessed in our experiments. Real-world implicit interactions in the form of view portions are selected to conduct predictions. Tests were run on two different groups based on user behavior. Both groups are used in the parameter tuning and experimentation phase. Initial findings under various objectives show promising results considering production mode scenarios and the choice of genetic operators. We further discuss performance graphs and conclude with future work.; Cette thèse aborde deux problèmes distincts : les emplois du temps des formations professionnelles et les systèmes de recommandation. Les deux problèmes ont lieu à Mandarine Academy, une société française de technologie éducative qui se spécialise dans les techniques de formation d'entreprise innovantes telles que les plateformes d'apprentissage en ligne personnalisées, la logistique de formation, conférences Web, etc.Nous commençons par le problème des emplois du temps lié à la gestion de la logistique de formation, l'entreprise propose un outil nommé « Dileap Logistic» qui gère automatiquement l'affectation des ressources (salles, enseignants et équipement) aux créneaux horaires (jours et heures) à des endroits précis. Auparavant, cela se faisait manuellement, ce qui prenait beaucoup de temps et donnait des résultats inexacts (mauvais formateurs, mauvais jours ou mauvais équipement).Bien qu'il s'agisse d'un problème NP-Complete, une formulation mathématique du problème a été élaborée après avoir examiné les ouvrages de littérature et les exigences de l'entreprise. Il comprend 18 contraintes (hard/soft) et 5 objectifs, dont deux sont concurrents. Nous évaluons 5 algorithmes évolutifs multi-objectifs (AEMO) en commençant par l'algorithme génétique de tri non dominant (NSGA II et NSGA III), l'algorithme évolutif multi-objectif basé sur la décomposition (MOEA/D), l'algorithme évolutif basé sur des indicateurs (IBEA) et enfin l'algorithme évolutif de Pareto Strength (SPEA 2). Deux opérateurs génétiques personnalisés (pour Mutation et Crossover) ont été proposés et comparés à des opérateurs classiques (PMX et Swap mutation).Une phase de réglage impliquant tous les algorithmes mentionnés ci-dessus est effectuée pour obtenir des configurations d'élite. Les expérimentations sont divisées par taille de problème (petites, moyennes et grandes instances) avec 3 à 5 objectifs testés. Nous discutons de résultats tels que la comparaison de la performance de chaque algorithme ainsi que des graphes de convergence.En plus de proposer des solutions à la logistique de la formation, l'entreprise crée quotidiennement du contenu pédagogique en ligne (vidéos, quiz, documents, etc.) pour soutenir la numérisation des environnements de travail et suivre les tendances actuelles. Avec plus de 550K utilisateurs répartis sur 100 plateformes d'apprentissage. Cependant, avec un catalogue en croissance excessive, des problèmes comme la surcharge d'information peut augmenter le manque de motivation chez les utilisateurs. Mandarine Academy nous a donné accès à l'une de ses plateformes publiques d'apprentissage pour mener nos recherches. Après une analyse de la littérature et une analyse approfondie des données d'utilisateurs. Nous proposons des améliorations à l'expérience utilisateur globale et à l'interface graphique.Des mesures similaires ont été prises dans le cadre des travaux antérieurs de planification du temps sur le moteur de recommandation. Ceci inclut la formulation mathématique d'un problème d'optimisation multi-objectif ayant 5 objectifs (Similarité, Diversité, Nouveauté, Erreur quadratique moyenne racine (RMSE) et Gain cumulé actualisé normalisé (NDCG) et la mise en œuvre de différentes techniques de recommandation pour générer des solutions initiales. Tous les algorithmes évolutifs mentionnés ci-dessus (NSGA II, NSGAIII, IBEA, SPEA2 et MOEA/D) sont évalués dans nos expériences.Les interactions implicites du monde réel sous forme de temps de visionnage sont choisies pour effectuer des prédictions. Les tests ont été exécutés sur deux groupes différents en fonction du comportement de l'utilisateur. Les deux groupes sont utilisés dans la phase de réglage des paramètres et les expérimentations finales. Les premiers résultats de divers objectifs sont prometteurs, compte tenu des scénarios de mode de production et du choix des opérateurs génétiques. Nous discutons davantage des graphes de performance et concluons avec les travaux futurs.

Cette thèse aborde deux problèmes distincts : les emplois du temps des formations professionnelles et les systèmes de recommandation. Les deux problèmes ont lieu à Mandarine Academy, une société française de technologie éducative qui se spécialise dans les techniques de formation d'entreprise innovantes telles que les plateformes d'apprentissage en ligne personnalisées, la logistique de formation, conférences Web, etc.Nous commençons par le problème des emplois du temps lié à la gestion de la logistique de formation, l'entreprise propose un outil nommé « Dileap Logistic» qui gère automatiquement l'affectation des ressources (salles, enseignants et équipement) aux créneaux horaires (jours et heures) à des endroits précis. Auparavant, cela se faisait manuellement, ce qui prenait beaucoup de temps et donnait des résultats inexacts (mauvais formateurs, mauvais jours ou mauvais équipement).Bien qu'il s'agisse d'un problème NP-Complete, une formulation mathématique du problème a été élaborée après avoir examiné les ouvrages de littérature et les exigences de l'entreprise. Il comprend 18 contraintes (hard/soft) et 5 objectifs, dont deux sont concurrents. Nous évaluons 5 algorithmes évolutifs multi-objectifs (AEMO) en commençant par l'algorithme génétique de tri non dominant (NSGA II et NSGA III), l'algorithme évolutif multi-objectif basé sur la décomposition (MOEA/D), l'algorithme évolutif basé sur des indicateurs (IBEA) et enfin l'algorithme évolutif de Pareto Strength (SPEA 2). Deux opérateurs génétiques personnalisés (pour Mutation et Crossover) ont été proposés et comparés à des opérateurs classiques (PMX et Swap mutation).Une phase de réglage impliquant tous les algorithmes mentionnés ci-dessus est effectuée pour obtenir des configurations d'élite. Les expérimentations sont divisées par taille de problème (petites, moyennes et grandes instances) avec 3 à 5 objectifs testés. Nous discutons de résultats tels que la comparaison de la performance de chaque algorithme ainsi que des graphes de convergence.En plus de proposer des solutions à la logistique de la formation, l'entreprise crée quotidiennement du contenu pédagogique en ligne (vidéos, quiz, documents, etc.) pour soutenir la numérisation des environnements de travail et suivre les tendances actuelles. Avec plus de 550K utilisateurs répartis sur 100 plateformes d'apprentissage. Cependant, avec un catalogue en croissance excessive, des problèmes comme la surcharge d'information peut augmenter le manque de motivation chez les utilisateurs. Mandarine Academy nous a donné accès à l'une de ses plateformes publiques d'apprentissage pour mener nos recherches. Après une analyse de la littérature et une analyse approfondie des données d'utilisateurs. Nous proposons des améliorations à l'expérience utilisateur globale et à l'interface graphique.Des mesures similaires ont été prises dans le cadre des travaux antérieurs de planification du temps sur le moteur de recommandation. Ceci inclut la formulation mathématique d'un problème d'optimisation multi-objectif ayant 5 objectifs (Similarité, Diversité, Nouveauté, Erreur quadratique moyenne racine (RMSE) et Gain cumulé actualisé normalisé (NDCG) et la mise en œuvre de différentes techniques de recommandation pour générer des solutions initiales. Tous les algorithmes évolutifs mentionnés ci-dessus (NSGA II, NSGAIII, IBEA, SPEA2 et MOEA/D) sont évalués dans nos expériences.Les interactions implicites du monde réel sous forme de temps de visionnage sont choisies pour effectuer des prédictions. Les tests ont été exécutés sur deux groupes différents en fonction du comportement de l'utilisateur. Les deux groupes sont utilisés dans la phase de réglage des paramètres et les expérimentations finales. Les premiers résultats de divers objectifs sont prometteurs, compte tenu des scénarios de mode de production et du choix des opérateurs génétiques. Nous discutons davantage des graphes de performance et concluons avec les travaux futurs. This thesis addresses two distinct problems: Professional Timetabling and Recommender Systems. Both issues take place at Mandarine Academy, a french educational technology company that specializes in innovative corporate training techniques such as personalized e-learning platforms, training logistics, web conferences, etc. We start with the timetabling problem related to handling training logistics, the company proposes a tool named “Dileap Logistic” that automatically handles the assignment of resources (rooms, teachers, and equipment) to time-slots (days/hours) at specific locations. Previously, this was done manually, which took a long time and resulted in inaccurate results (wrong trainers, days, or equipment). Despite being an NP-Complete problem, a mathematical formulation of the problem has been developed after reviewing both literature works and company requirements. It includes 18 constraints (hard/soft) and 5 objectives, two of which are competing. We evaluate five state of the art Multi-Objective Evolutionary Algorithms (MOEA's) starting with the non-dominated sorting genetic algorithm (NSGA II and NSGA III), the Multi-objective Evolutionary Algorithm Based on Decomposition (MOEA/D), the indicator-based Evolutionary Algorithm (IBEA) and finally the Strength Pareto Evolutionary Algorithm (SPEA 2). Two custom genetic operators (for Mutation and Crossover) were proposed and compared against classical operators (PMX and Swap mutation). A tuning phase involving all the above-mentioned algorithms is performed to obtain elite configurations. Experiments are divided by each problem size (small, medium, and large instances) with 3 to 5 objectives tested. We discuss results such as comparing the performance of each algorithm using various metrics, as well as convergence graphs and population evolution. Aside from proposing solutions to training logistics, the company creates online pedagogical content (videos, quizzes, documents, etc.) on daily basis to support the digitization of work environments and to keep up with current trends. With more than 550K users spread across 100 active e-learning platforms. However, with an overgrowing catalog, problems like information overload can increase the lack of motivation among users. Mandarine Academy provided us with access to one of its public e-learning platform to conduct our research. After literature review and extended data analysis of user interactions. We propose enhancements to the overall user experience and interface. Similar steps taken in the previous timetabling work are also applied on the recommendation engine. This includes the mathematical formulation of a multi-objective optimization problem having 5 objectives (Similarity, Diversity, Novelty, Root Mean Square Error (RMSE), and Normalized Discounted Cumulative Gain (NDCG)) and the implementation of different recommendation techniques to generate initial solutions. All of the above-mentioned evolutionary algorithms (NSGA II, NSGAIII, IBEA, SPEA2, and MOEA/D) are assessed in our experiments. Real-world implicit interactions in the form of view portions are selected to conduct predictions. Tests were run on two different groups based on user behavior. Both groups are used in the parameter tuning and experimentation phase. Initial findings under various objectives show promising results considering production mode scenarios and the choice of genetic operators. We further discuss performance graphs and conclude with future work.

Das Directed Feedback Vertex Set (DFVS) Problem ist ein NP-vollständiges, kombinatorisches Optimierungsproblem, das für ungewichtete, gerichtete Graphen spezifiziert ist, und dessen Ziel es ist, ein DFVS minimaler Größe zu finden. Eine Knotenmenge wird als Directed Feedback Vertex Set (gerichtete kreiskritische Knotenmenge) bezeichnet, wenn sie zumindest einen Knoten eines jeden gerichteten Kreises im Graphen enthält, sodass das Entfernen der Knotenmenge aus dem Graphen zu einem gerichteten azyklischen Graphen (engl. directed acyclic graph, DAG) führt. Das Feedback Vertex Set Problem findet in vielen verschiedenen Bereichen Anwendung, wie zum Beispiel bei der Verklemmungserkennung und -beseitigung in Betriebssystemen, bei computerunterstützten Konstruktionsanwendungen und in der chemischen Verfahrenstechnik. Das Ziel dieser Diplomarbeit ist die Entwicklung hybrider Metaheuristiken, die innerhalb eines Zeitlimits hochwertige Lösungen für das DFVS Problem erzeugen. Am Beginn unserer Lösungsansätze steht jeweils eine Prozedur zur Reduktion der Graphengröße, in der aus der Literatur bekannte Reduktionsregeln zur Anwendung kommen, und die den Suchraum für mögliche Lösungen verkleinert. Danach wird eine Greedy-Heuristik verwendet, um eine erste Lösung zu erzeugen, welche im Weiteren mit einer lokalen Suche verbessert wird. Unsere Metaheuristik besteht aus einer Large Neighborhood Search (LNS), deren Nachbarschaften jeweils von einer zerstörenden und einer reparierenden Operation bestimmt werden. Wir präsentieren zwei Nachbarschaften, in denen entweder Knoten von der momentanen Lösung zum entsprechenden DAG hinzugefügt werden (enlarge-DAG) oder umgekehrt (enlarge-(partial-)DFVS). Eine gültige Lösung wird dann wiederhergestellt, indem das resultierende kleinere Teilproblem in ein gemischt-ganzzahliges lineares Optimierungsmodell (engl. mixed integer linear program, MILP) überführt und mit entsprechender Software gelöst wird. Diese Kombination aus LNS und MILP führt zu einer Hybridisierung der Metaheuristik. Wir stellen zwei MILP Modelle für das DFVS Problem vor, welche sowohl als Bestandteil der hybriden LNS als auch in eigenständigen Lösungsverfahren verwendet werden können. Die CEC genannte Formulierung basiert auf Constraints zur Entfernung von Kreisen (engl. cycle elimination constraints). Die Formulierung namens MTZ wurde inspiriert von den Kurzzyklusungleichungen, auch Subtour-Eliminationsbedingungen genannt, nach Miller, Tucker und Zemlin und ist eine Abwandlung eines Modells aus der Literatur. Weiters präsentieren wir verschiedene Strategien zur Wahl des Zerstörungsgrades in der LNS und auch eine Vorgehensweise, wie die jeweils passende MILP Formulierung für jede Instanz gewählt werden kann. Das Hinzufügen dieser Erweiterungen ergibt mehrere Varianten unserer hybriden Metaheuristik, welche auf typischen Instanzen getestet und bezüglich ihrer durchschnittlich erreichten Lösungsqualität miteinander verglichen werden. Unsere praktische Untersuchung zeigt, dass alle Varianten der hybriden LNS höhere Durchschnittswerte für die Lösungsqualität erzielen als die eigenständigen Lösungsverfahren basierend auf den MILP Formulierungen. Außerdem übertrifft die Leistung der CEC Formulierung zusammen mit der enlarge-DAG Nachbarschaft die der anderen Ansätze, welche MTZ oder auch die Kombination von MTZ und der enlarge-partial-DFVS Nachbarschaft verwenden, wenn eine zeitliche Beschränkung festgelegt wird. Das dynamische Auswahlverfahren für den Zerstörungsgrad, das auf der Anzahl an Kreisen mit Länge zwei im Graphen beruht, erzielt die besten Ergebnisse aller untersuchten Strategien. Die durchschnittliche Lösungsqualität wird weiter verbessert, wenn dieser Ansatz mit der dynamischen Auswahl der MILP Formulierung gepaart wird. Des Weiteren ist es uns gelungen, mit den Hauptvarianten unserer hybriden Metaheuristik die Leistungen zweier state-of-the-art Lösungsverfahren aus der Literatur, welche auf Simulated Annealing basieren, zu übertreffen., The directed feedback vertex set (DFVS) problem is an NP-complete combinatorial optimization problem defined on unweighted, directed graphs with the goal of finding a DFVS of minimum cardinality. A set of vertices is called a directed feedback vertex set if it contains at least one vertex of every directed cycle in the graph, such that the removal of the set from the graph leads to a directed acyclic graph (DAG). The general feedback vertex set problem is relevant for many different areas and real world scenarios including deadlock detection and recovery in operating systems, computer-aided design applications, and chemical engineering. The aim of this work is to develop hybrid metaheuristics that provide solutions of high quality for challenging DFVS problem instances within a given time limit. Our solving approaches start with a graph reduction procedure using reduction rules from the literature to decrease the size of the graph and reduce the search space for possible solutions. An initial solution is generated with a greedy construction heuristic and improved with a classical local search procedure. For the metaheuristic framework, we employ large neighborhood search (LNS) with neighborhoods that are each defined by a destroy and a repair method. We introduce two neighborhood structures which are based on either moving vertices from the current solution to the corresponding DAG (enlarge-DAG) or the other way around (enlarge-(partial-)DFVS). A valid solution is restored by encoding the resulting smaller subproblem with a mixed integer linear programming (MILP) formulation and solving it with an existing solver. This combination of LNS and MILP leads to the hybridization of the metaheuristic. We propose two MILP formulations for the DFVS problem, which can be used within the hybrid LNS as well as in standalone solving procedures. One formulation called CEC is based on cycle elimination constraints and the other formulation MTZ is inspired by the subtour elimination constraints from Miller, Tucker, and Zemlin and derived from the literature. We also investigate various strategies for selecting the degree of destruction for the LNS and a technique for dynamically choosing the MILP formulation for each instance. These enhancements lead to multiple variants of our hybrid metaheuristic, which are tested on benchmark instances and compared in terms of the achieved average solution quality. The results of our computational study show that all hybrid LNS variants achieve higher average solution qualities than the standalone MILP solving procedures. Furthermore, the CEC-based formulation together with the enlarge-DAG neighborhood outperforms other approaches using the MTZ-based formulation as well as the combination of MTZ and enlarge-partial-DFVS neighborhood when having a time limit. The dynamic selection strategy for the degree of destruction that is based on the number of cycles of length two in the respective graph performs the best out of all tested strategies. Enhancing this approach with the dynamic selection of the MILP formulation leads to further improvements of the solution quality. Moreover, the main variants of our hybrid metaheuristic manage to outperform two state-of-the-art solving procedures from the literature, which are based on simulated annealing.

Community detection is a prominent research topic in Complex Network Analysis, and it constitutes an important research field on all those areas where complex networks represent a powerful interpretation tool for describing and understanding systems involved in neuroscience, biology, social science, economy, and many others. A challenging approach to uncover the community structure in complex network, and then revealing the internal organization of nodes, is Modularity optimization. In this research paper, we present an immune optimization algorithm (opt-IA) developed to detect community structures, with the main aim to maximize the modularity produced by the discovered communities. In order to assess the performance of opt-IA, we compared it with an overall of 20 heuristics and metaheuristics, among which one Hyper-Heuristic method, using social and biological complex networks as data set. Unlike these algorithms, opt-IA is entirely based on a fully random search process, which in turn is combined with purely stochastic operators. According to the obtained outcomes, opt-IA shows strictly better performances than almost all heuristics and metaheuristics to which it was compared; whilst it turns out to be comparable with the Hyper-Heuristic method. Overall, it can be claimed that opt-IA, even if driven by a purely random process, proves to be reliable and with efficient performance. Furthermore, to prove the latter claim, a sensitivity analysis of the functionality was conducted, using the classic metrics NMI, ARI and NVI.

Engineering and Applied Science Research, 50, 4, 291-297

SAT solvers determine whether a given propositional formula is satisfiable. Today’s highly engineered SAT solvers can determine the satisfiability of huge formulas with millions of constraints and thousands of variables. Further, stat- ing other combinatorial problems in terms of propositional satisfiability allows the use of this great SAT-solving performance for a large range of combinatorial problems. Despite the great performance of SAT solvers, scalability becomes an issue whenever the instances become too large or too complex. For complex instances, the SAT solver may not be able to determine satisfiability in a reasonable amount of time and for large instances, the corresponding formula becomes too large, and solving becomes practically impossible. In this thesis, we propose different methods for overcoming these scalability issues. In particular, we discuss examples of efficient encoding design, scalability using a lazy approach, and embedding SAT-based methods into a heuristic approach using SAT-based local improvement (SLIM). These methods are discussed in the context of applications: we propose encodings for computing hypertree width and twin-width; a lazy approach to the directed feedback vertex problem; and SLIM approaches to graph coloring and decision tree induction.

Increased demand on the package delivery industry is prompting companies to make use of crowdsourced delivery drivers. In this industry, when crowdsourced drivers are used the drivers need to be quickly provided with an optimal delivery route. This thesis addresses a new, real-world, crowdsourcing variant of the Vehicle Routing Problem, with deviation constraints and both heterogeneous and homogeneous capacity constraints using ad-hoc drivers. Previously, the problem has only been solved with exact techniques. Although exact techniques guarantee exact optimal solutions, they can have very long running times making them impractical in a fast paced, real-world situation. For this reason, this thesis explores a metaheuristic approach to solving the crowdsourced VRP, where solutions are not guaranteed to be exact, but can be close to exact and the running times are much shorter, thus practical in a real-world setting. In this thesis, we develop two Memetic Algorithms (HA1 and HA2) by combining a Genetic Algorithm with a Randomized Hill Climbing Local Search. The algorithms aim is the problem’s objective of minimizing total deviations of all the drivers while satisfying the problem's constraints. Problem-specific crossover and mutation operations are developed, and two variants of the crossover operation are explored. The results of the algorithms are empirically evaluated. The algorithms best fitness values and running times are compared to results of a Constraint Programming solver and the algorithms are compared against each other.One Memetic Algorithm in particular, HA1, provided positive results when evaluated. This algorithm is able to provide feasible solutions to instances with up to 135 packages with average running times of only a few minutes. On small instances the Memetic and Genetic Algorithms found feasible, but not exact, solutions, compared to the exact solutions of the Constraint Programming solver. Both Memetic Algorithms found more feasible and better results than the Genetic Algorithm, with HA1 performing the best in this area. The results found with a Memetic approach are practical for real-world problems since results that are both feasible and practical are provided within very short running times.

Gerçek hayat uygulamaları arasında yer alan esnek atölye çizelgeleme problemi (FJSSP), atölye çizelgeleme problemlerinin (JSSP) genişletilmiş versiyonudur. Problem esas olarak iki zorluktan oluşmaktadır. Birincisi, işlemi gerçekleştirebilecek bir makine setinden her işlem için bir makinenin atanması, ikincisi ise işlemlerin fizibilite koşulları sağlanırken sıralanmasıdır. Çizelgeleme problemlerini çözmek, iyi bilinen NP-zor problem alanında oldukları için oldukça zordur. Literatür araştırmaları bu iki alt problemi bütünleşik bir yaklaşımla birlikte veya sıralı bir şekilde tek tek ele alarak çözmeye çalışmaktadır.Pazardaki yoğun rekabet nedeniyle, etkin çizelgeleme artık endüstrilerin büyümesi ve hayatta kalması için önemli bir konu haline gelmiştir. Mevcut rekabet ortamında ayakta kalabilmek için, endüstrilerin, üretim süresi, akış süresi ve gecikme gibi performans ölçütlerinin eşzamanlı optimizasyonuna dayalı programı iyileştirmeleri esastır. Tüm çizelgeleme kriterleri iş operasyonu açısından önemli olduğundan, tek bir hedef yerine tüm hedefleri aynı anda optimize etmek hayati önem taşımaktadır. Ayrıca, makine arızaları, arızalar, teslim süreleri veya son teslim tarihleri gibi ek kısıtlamalar getirilirse sorun daha da zorlaşabilir. Bu amaçla, dinamik ortamlarda çalıştıkları için gerçek dünya uygulamaları için sağlam ve esnek çizelgeler bulma sorunu son derece önemlidir. Son zamanlarda, çok amaçlı FJSSP'leri çözmek için birçok çalışma yapılmıştır. Bu çalışmalar, meta-sezgisel, matematiksel programlama, simülasyon tabanlı yöntemler, yapay zeka tabanlı yöntemler veya bunların bir kombinasyonu gibi çeşitli teknikleri kullanan algoritmalar geliştirerek en uygun çözümü belirlemeyi amaçlamaktadır.Bu tez çalışmasında, FJSSP'yi çözmek için GA'yı paralel yürütülen yerel arama prosedürleriyle bütünleştiren verimli bir algoritma önerilmiştir. Başlangıç popülasyonu, arama uzayında daha iyi alanlar ve erken yakınsamayı önlemek için yeterli çeşitlilik sağlamak için gönderme kuralları kullanılarak oluşturulur. Ayrıca, çözümün sağlamlığını, yakınsama ve yürütme süresini iyileştirmek için kritik bir yol tabanlı çoklu arama prosedürü paralel olarak uygulanır ve yürütülür. Hedef fonksiyonlar, Pareto Optimal yaklaşımı ile değerlendirilir ve çözümler, Pareto sıralamasına ve hiperhacim katkı ölçüsüne (Δs) dayalı iki seviyeli bir seçim stratejisi ile seçilir. Önerilen algoritmanın etkinliği ve performansı, literatürde sunulan iyi bilinen veri setleri kullanılarak diğer algoritmalarla kıyaslanır ve karşılaştırılır. The Flexible Job-Shop Scheduling Problem (FJSSP) is a real-world problem that builds upon the Job Shop Scheduling Problem (JSSP). It involves two main challenges: assigning the appropriate machine for each operation from a set of machines that can perform the operation and arranging the operations in a sequence while meeting feasibility conditions. These scheduling problems are known to be challenging as they fall under the NP-hard category. Researchers have attempted to solve these sub-problems by either addressing them as a combined approach or tackling them individually in a sequential manner.With the market becoming increasingly competitive, effective scheduling has become a crucial aspect for the success and survival of industries. In order to remain competitive, companies must improve their schedules by simultaneously optimizing various performance measures such as makespan, flow time, and tardiness. All of these scheduling criteria are important from a business perspective, therefore it is essential to optimize all objectives at once, rather than just one. Additionally, the problem can become even more complex when additional constraints are introduced, such as machine breakdowns, failures, release times, or deadlines. This is why it is important to find robust and flexible schedules that can adapt to changing conditions in real-world applications, as they often operate in dynamic environments. In recent times, there has been a significant amount of research focused on resolving multi-objective Flexible Job-Shop Scheduling Problems (FJSSPs). These studies aim to identify the optimal solution through the development of algorithms that employ various techniques such as meta-heuristics, mathematical programming, simulation-based methods, artificial intelligence-based methods or a combination of them.This dissertation presents a new algorithm that utilizes a combination of Genetic Algorithm (GA) and parallelized local search procedures to effectively solve the Flexible Job-Shop Scheduling Problem (FJSSP). The initial population is generated using dispatching rules to identify promising regions in the search space and to ensure adequate diversity to prevent early convergence. Additionally, a critical path-based multi-search procedure is applied and executed in parallel to enhance the robustness, convergence, and computational efficiency of the solution. The algorithm evaluates the objective functions using a Pareto Optimal approach and employs a two-level selection strategy that combines Pareto ranking and the hypervolume contribution measure (Δs) to choose solutions. The proposed algorithm has been evaluated and compared to other existing methods using well-known datasets from the literature to demonstrate its effectiveness and performance.