Search

Consumer demand for automobiles is changing because of the vehicle's dependability and utility, and the superb design and high comfort make the vehicle a wealthy object class. The creation of object classes necessitates the creation of more sophisticated computer vision models. However, the critical issue is image quality, determined by lighting conditions, viewing angle, and physical vehicle construction. This work focuses on creating and implementing a deep learning-based traffic analysis system. Using a variety of video feeds and vehicle information, the developed model recognizes, categorizes, and counts vehicles in real-time traffic flow. The dynamic skipping method offered in the developed model speeds up the processing of a lengthy video stream while ensuring that the video picture is delivered accurately to the viewer. In real-time traffic, standard vehicle retrieval may assist in determining the make, model, and year of the vehicle. Previous MobileNet and VGG19 models achieved F-values of 0.81 and 0.91, respectively. However, the proposed solution raises MobileNet's frame rate from 71.2 to 89.17 and VGG19's frame rate from 48.2 to 59.14. The method may be applied to a wide range of applications that require a dedicated zone to monitor real-time data analysis and normal multimedia operations. [ABSTRACT FROM AUTHOR]

Training and inference of deep convolutional neural networks are usually slow due to the depth of the network and the number of parameters in the network. Although high-performance processors usually accelerate the training of these networks, their use on resource-constrained devices is still limited. Several compression-based acceleration methods have been presented to optimize the performance of neural networks. However, their use and adaptation are still limited due to their adverse effects on the network structure. Therefore, different filter pruning methods have been proposed to keep the network structure intact. To better solve the above limitations, we first propose a detailed classification of model acceleration method to explain the different ways of enhancing the inference performance of the convolutional neural network. Second, we present a broad classification of filter pruning methods including the comparison of these methods. Third, we present an empirical evaluation of four filter pruning methods to understand the effects of filter pruning on model accuracy and parameter reduction. Fourth, we perform several experiments with ResNet20, a pre-trained CNN, and with the proposed custom CNN to show the effect of filter pruning on them. ResNet20 is used to address the multiclass classification using CIFAR 10 dataset and custom CNN is used to address the binary classification using Leukaemia image classification dataset that includes low-information medical images. The experimental results show that among the four filter pruning methods, the soft filter pruning method best preserves the accuracy of the original model for both ResNet20 and the custom CNN. In addition, the sampling-based filter pruning method shows the highest reduction of 99.8% in parameters on custom CNN. The overall results show a reasonable pruning ratio within five training epochs for both the pre-trained CNN and custom CNN. In addition, our results show that pruning redundant filters significantly reduces the model size, and number of floating point operations. [ABSTRACT FROM AUTHOR]

Classification of surgical actions from a real anatomy video sequence is a challenging task due to limited visibility, poorer lighting conditions, lower contrast, and obscured frames of a video sequence. Several deep model-based spatiotemporal and spatial feature learning methods have been presented to classify surgical actions. However, some of the methods have been evaluated on synthetic data due to unavailability of sufficient labeled data. Conversely, some methods have been evaluated on a real anatomy dataset, but produce lower accuracy. Therefore, in this paper, first we analyze the effects of both feature learning methods on surgical action classificaiton from a real anatomy dataset. Thereafter, we propose new methods to enhance surgical action classification. Specific contributions in this paper are as follows. First, two novel deep model-based spatiotemporal feature learning methods are proposed to classify surgical actions. Second, a hypothesis is proposed, stating that the elimination of spatiotemporal features does not affect the performance of the method. Third, to test the proposed hypothesis, a spatial feature learning method comprised of a unique custom Convolutional Neural Network (CNN) is also proposed. Fourth, performance analysis of the proposed spatiotemporal and spatial feature learning methods is presented using a real anatomy Surgical Actions 160 dataset. The experimental results demonstrate that the MobileNetV2-based spatial feature learning method achieves the highest accuracy of 97% in classifying surgical actions. It outperforms the other pre-trained CNN models significantly. Additionally, MobileNetV2 exhibits lower training time. [ABSTRACT FROM AUTHOR]

Sensor data that is often collected in the Internet of Things (IoT) or any smart city environment should be protected against security and privacy concerns. Often, sensor data that is shared by devices in smart cities contains sensitive or private information that can often be shared over different networks and by different smart applications. In the last decade, the area of Privacy-Preserving Data Mining (PPDM) has received a lot of attention as the amount of data received and collected daily is huge. Unfortunately, PPDM mostly applies to binary data. To improve the usefulness of PPDM, we present a more usable version for smart cities called Privacy-Preserving Utility Mining (PPUM), in the form of a Maximal Sensitive Utility-Maximal Sensitive ConflIct (MSU-MSI) algorithm. MSU-MSI finds any conflicting items that may contain sensitive itemsets with high-utility and sanitizes them, stripping them of sensitive and private information while maintaining utility. Any transactions encountered that contain sensitive itemsets are first fed through sanitization processes. This is followed by calculating the total number of items that conflict, and then removing them so sanitization processes can operate more efficiently so as to not redo known sanitization processes. We conduct an in-depth experimental analysis, where our detailed methodology is compared directly with state-of-the-art frameworks such as MSU-MIU, MSU-MAU, HHUIF and MSCIF. Our proposed MSU-MSI shows a higher performance in missing cost, in particular when dealing with highly dense or highly sparse datasets. Moreover, our novel framework is shown to achieve an excellent performance with regards to similarity in database structure and database utility. [ABSTRACT FROM AUTHOR]

In the period of rapid development on the new information technologies, computer vision has become the most common application of artificial intelligence, which is represented by deep learning in the current society. As the most direct and effective application of computer vision, facial expression recognition (FER) has become a hot topic and used in many studies and domains. However, the existing FER methods focus on deep learning to generate increasingly complex attention structures, so they are unable to consider the connotative relationship between different parts of facial expressions. Moreover, the human expression recognition method based on complex deep learning network has serious interpretability issues. Therefore, in this paper, a novel Graph Neural Network (GNN) model is proposed to consider the systematic process of FER in human visual perception. Firstly, a region division mechanism is proposed, which divides the face region into six parts to unify the selection of key facial features. On this basis, in order to better consider the connotative relationship between different parts of facial expression, a human visual cognition strategy is proposed, which uses the divided six regions to learn facial expression features, and evenly selects the key features with high reliability as graph nodes. In combination with the human regional cooperative recognition process, the connotative relationship (such as relative position and similar structure) between graph nodes is extracted, so as to construct the GNN model. Finally, the effect of FER is obtained by the modeled GNN model. The experimental results compared with other related algorithms show that the model not only has stronger characterization and generalization ability, but also has better robustness compared with state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Chromosome images are commonly used in karyotype analysis to diagnose chromosomal diseases. However, there are often chromosome adhesion and overlaps in chromosome images, so effective chromosome segmentation is conducive to smooth karyotype analysis. To date, some progress has been made in automatic chromosome segmentation, and existing methods can be used to segment overlapping chromosomes in most cases. However, when two or more overlapping regions are too close to each other in the image of overlapping chromosomes, the existing segmentation methods adjust the non-overlapping regions that do not belong to the overlapping region, resulting in incomplete segmentation of chromatids. Therefore, we use a heuristic algorithm to solve this problem from the point of view of mathematics and geometry to improve the segmentation of overlapping chromosomes. Starting from chromosome images, the existing problems and solutions are explained and displayed in the way of visualized interpretable image features, which helps to better understand the algorithm. Our method achieves 92.86% splicing accuracy and 90.44% overall segmentation accuracy on open datasets. The experimental results show that our method can effectively improve the problem of incorrect chromosome segmentation when two or more overlapping parts of overlapping chromosomes are too close to each other. It can accelerate the development of artificial intelligence in computational pathology and provide patients with more accurate medical services. [ABSTRACT FROM AUTHOR]

There are some problems in abnormal node sensing in regional wireless networks, such as low sensing accuracy and poor judgment results of abnormal states of sensing nodes. Therefore, this paper develops a method for abnormal node sensing in regional wireless networks based on convolutional neural network. In addition, we will analyze the structure of regional wireless network nodes and determine the distribution mode of wireless network nodes. The regional wireless network node data are extracted and the pivot quantity and two-dimensional Gaussian distribution state are constructed using the median to build the regional wireless network node deployment model according to the confidence interval of the data characteristics; analyze the basic principle of convolution neural network, determine the operation mode of convolution kernel, classify the regional wireless network node data using Bayesian network, set a safety distance to determine the abnormal node of the regional wireless network, train the determined abnormal data as the input data of convolutional neural network and input it into the constructed perception model of the abnormal node of the regional wireless network, the loss function is set to continuously update the iterative results to realize the perception of abnormal node in the regional wireless network. The simulation results show that the sensing range of this method is relatively consistent with the range set by the sample, and the sensing accuracy reaches more than 95%, and the abnormal state error of abnormal nodes in the evaluation sample area is always less than 2%, which verifies that this method improves the sensing accuracy, reduces the error, and has higher application value. [ABSTRACT FROM AUTHOR]

In the process of engineering project construction, the balanced allocation of resources has an important impact on the purchase of actual materials, the progress of the site construction and the arrangement of temporary facilities. Although there have been many studies on the use of evolutionary computation (EC) to solve the fixed duration of the resource-leveling problem, the optimization effect of standard EC to solve the resource-balancing problem in large-scale network design planning is reduced due to the phenomenon of premature convergence of PSO and the complexity of large-scale network planning. Therefore, in this paper, an adaptive dissipative particle swarm optimization (ADPSO) algorithm is used to solve the resource balancing optimization problem for different network-plans scales, and the feasibility of the developed model is verified using four different scaled network plans (databases). At the same time, a solution to the dynamic resource-leveling problem with optimization for a fixed duration is proposed. The results show that the proposed models are suitable for solving the problem of resource-leveling with a fixed duration of optimization of different scale network plans, and their accuracy and stability are significantly improved compared to the state-of-the-art approaches. Moreover, the proposed models are accurate and suitable for combining the calculation method of network planning time parameters with the ADPSO algorithm to realize dynamic resource balancing optimization. [ABSTRACT FROM AUTHOR]

This paper introduces a novel and complete framework for solving different Internet of Things (IoT) applications, which explores eXplainable AI (XAI), deep learning, and evolutionary computation. The IoT data coming from different sensors is first converted into an image database using the Gamian angular field. The images are trained using VGG16, where XAI technology and hyper-parameter optimization are introduced. Thus, analyzing the impact of the different input values in the output and understanding the different weights of a deep learning model used in the learning process helps us to increase interpretation of the overall process of IoT systems. Extensive testing was conducted to demonstrate the performance of our developed model on two separate IoT datasets. Results show the efficiency of the proposed approach compared to the baseline approaches in terms of both runtime and accuracy. [ABSTRACT FROM AUTHOR]

The growth of media services, multimedia conferencing, interactive distance learning, and distributed interactive simulations is becoming more dependent on the security of multicast communication over 5 G and 6 G networks. In order to ensure the security and efficacy of multicast communication, our research came up with the idea of combining multicast public-key encryption with digital signatures known as "Multi-Receiver Signcryption (MRSC)." As compared to multicast encryption and signature primitives, MRSC significantly improves the effectiveness of secure information delivery through multicast communication over 5 G and 6 G networks. In this paper, we first provide the formal model of MRSC schemes used in Public Key Infrastructure, Identity-based Cryptography, and Certificateless Cryptography. Secondly, we present a typology of MRSC, as well as a summary of an in-depth investigation of the qualities of security, the cost of computing, and the overhead of communication over the networks. [ABSTRACT FROM AUTHOR]

This paper proposes a deep learning model that is robust and capable of handling highly uncertain inputs. The model is divided into three phases: creating a dataset, creating a neural network based on the dataset, and retraining the neural network to handle unpredictable inputs. The model utilizes entropy values and a non-dominant sorting algorithm to identify the candidate with the highest entropy value from the dataset. This is followed by merging the training set with adversarial samples, where a mini-batch of the merged dataset is used to update the dense network parameters. This method can improve the performance of machine learning models, categorization of radiographic images, risk of misdiagnosis in medical imaging, and accuracy of medical diagnoses. To evaluate the efficacy of the proposed model, two datasets, MNIST and COVID, were used with pixel values and without transfer learning. The results showed an increase of accuracy from 0.85 to 0.88 for MNIST and from 0.83 to 0.85 for COVID, which suggests that the model successfully classified images from both datasets without using transfer learning techniques. [ABSTRACT FROM AUTHOR]

Simultaneously, the application of resilient distributed datasets (RDD) in cloud computing provides a good environment for data analysis of big data. In addition, the combination of Machine Learning (ML) algorithms of the edge computing paradigm and the SFUP-SP algorithm may be able to also be used to improve local computing capabilities and speed up data analysis and user decision-making. [ABSTRACT FROM AUTHOR]

In today's society, investment wealth management has become a mainstream of the contemporary era. Investment wealth management refers to the use of funds by investors to arrange funds reasonably, for example, savings, bank financial products, bonds, stocks, commodity spots, real estate, gold, art, and many others. Wealth management tools manage and assign families, individuals, enterprises, and institutions to achieve the purpose of increasing and maintaining value to accelerate asset growth. Among them, in investment and financial management, people's favorite product of investment often stocks, because the stock market has great advantages and charm, especially compared with other investment methods. More and more scholars have developed methods of prediction from multiple angles for the stock market. According to the feature of financial time series and the task of price prediction, this article proposes a new framework structure to achieve a more accurate prediction of the stock price, which combines Convolution Neural Network (CNN) and Long–Short-Term Memory Neural Network (LSTM). This new method is aptly named stock sequence array convolutional LSTM (SACLSTM). It constructs a sequence array of historical data and its leading indicators (options and futures), and uses the array as the input image of the CNN framework, and extracts certain feature vectors through the convolutional layer and the layer of pooling, and as the input vector of LSTM, and takes ten stocks in U.S.A and Taiwan as the experimental data. Compared with previous methods, the prediction performance of the proposed algorithm in this article leads to better results when compared directly. [ABSTRACT FROM AUTHOR]

Federated learning is becoming a practical solution for machine learning (ML) in industry. This is due to the possibility of implementing artificial intelligence (AI) systems and training its models on private data sets. However, this is not an ideal solution as it is possible to manipulate or even intercept the model during its transmission between the server and workers. In this paper, we propose a solution to ensure the security of the model transmitted between units in FL. The given model is encrypted with AES, DES, RSA algorithms, then a checksum is determined. This checksum with a private key is stored as a transaction on a blockchain. In the case of sending the model and its modification, the recipient can easily verify whether it is correct. The proposed solution has been described, tested, and compared to indicate its advantages and disadvantages. Conducted experiments were based on analyzing the communication time between participants, the accuracy of machine learning models, and attack detection. In terms of attack detection on the blockchain, we reached 81% thanks to the checksum mechanism. [ABSTRACT FROM AUTHOR]

Cyber-physical transactions can be clustered and classified using an effective vector representation. Traditionally, methods for finding patterns have relied on heuristics and pruning. Despite the high dimensionality of transactional data, approaches that use frequent item sets as features can be affected by dimensionality, sparsity, and privacy issues. As a result, FIs are not uniformly distributed over the data. In this paper, we propose an embedded architecture for transaction classification based on distributed learning. The model transforms transaction data into frequent sets and implements attention mechanisms based on encoder-decoder structures. Consequently, the model can learn continuous vectors in low dimensions while maintaining context and colocation. We analyzed a high-dimensional transaction dataset to test attention-based methods and federated learning. To improve decision limits while maintaining privacy and security, the proposed model lowers the global loss function. In the experiment, four datasets are used for comparison. The data are randomly selected and distributed to different clients for each dataset. We run each experiment with five different random partitions of the dataset to evaluate the active learning. The training set is the same size in each round, and the test set is not evaluated in each round. Our methods are compared to the best performing baseline method using the F1 score and the percentage of the dataset used. Compared to the baseline model, the proposed model performed better in terms of percentage increases and output classes, namely retail (1.7%, 2), cancer (17.38%, 3), food (0.74%, 2), and snippet (3.47%, 8). [ABSTRACT FROM AUTHOR]

This paper introduces a novel framework for identifying a group of intrusions in the context of the Internet of Things (IoT). It combines both deep learning and decomposition. A set of data is first collected and a recurrent neural network is used to estimate the different individual intrusions. These individual intrusions are then used to identify a group of outliers based on a decomposition strategy. As case studies, the proposed solutions have been experimentally evaluated using two computer network intrusion datasets, namely (1) IDS 2018, and (2) LUFlow. The results show the benefits of the proposed framework and clear superiority in comparison to the state-of-the-art approaches. [ABSTRACT FROM AUTHOR]

High-utility itemset mining (HUIM) is an important task in data mining that can retrieve more meaningful and useful patterns for decision-making. One-phase HUIM algorithms based on the utility-list structure have been shown to be the most efficient as they can mine high-utility itemsets (HUIs) without generating candidates. However, storing itemset information for the utility-list is time-consuming and memory consuming. To address this problem, we propose an efficient simplified utility-list-based HUIM algorithm (HUIM-SU). In the proposed HUIM-SU algorithm, the simplified utility-list is proposed to obtain all HUIs effectively and reduce memory usage in the depth-first search process. Based on the the simplified utility-list, repeated pruning according to the transaction-weighted utilisation (TWU) reduces the number of items. In addition, a construction tree and compressed storage are introduced to further reduce the search space and the memory usage. The extension utility and itemset TWU are then proposed to be the upper bounds, which reduce the search space considerably. Extensive experimental results on dense and sparse datasets indicate that the proposed HUIM-SU algorithm is highly efficient in terms of the number of candidates, memory usage, and execution time. [ABSTRACT FROM AUTHOR]

The Industrial Internet of Things (IIoT) promises to deliver innovative business models across multiple domains by providing ubiquitous connectivity, intelligent data, predictive analytics, and decision-making systems for improved market performance. However, traditional IIoT architectures are highly susceptible to many security vulnerabilities and network intrusions, which bring challenges such as lack of privacy, integrity, trust, and centralization. This research aims to implement an Artificial Intelligence-based Lightweight Blockchain Security Model (AILBSM) to ensure privacy and security of IIoT systems. This novel model is meant to address issues that can occur with security and privacy when dealing with Cloud-based IIoT systems that handle data in the Cloud or on the Edge of Networks (on-device). The novel contribution of this paper is that it combines the advantages of both lightweight blockchain and Convivial Optimized Sprinter Neural Network (COSNN) based AI mechanisms with simplified and improved security operations. Here, the significant impact of attacks is reduced by transforming features into encoded data using an Authentic Intrinsic Analysis (AIA) model. Extensive experiments are conducted to validate this system using various attack datasets. In addition, the results of privacy protection and AI mechanisms are evaluated separately and compared using various indicators. By using the proposed AILBSM framework, the execution time is minimized to 0.6 seconds, the overall classification accuracy is improved to 99.8%, and detection performance is increased to 99.7%. Due to the inclusion of auto-encoder based transformation and blockchain authentication, the anomaly detection performance of the proposed model is highly improved, when compared to other techniques. [ABSTRACT FROM AUTHOR]

Massive urbanization and resource scarcity are global phenomena for which researchers around the world see urban transformation into a smart city as the most viable solution. Yet urban planners face many hurdles in transforming cities. These include economic factors such as the high cost of infrastructure development and maintenance, social paradigms, governance, and environmental issues that limit government ambitions. To design an optimal strategy for creating a smart city, the various barriers must be identified and prioritized. The various barriers to the successful implementation of the smart city mission were identified using an extensive literature review and expert opinions. To understand the complex interdependencies among these barriers, the interpretive structural modeling method (ISM) was used. MICMAC analyzed the ISM model to classify the barriers. Indian cities vary greatly in terms of infrastructure, size, population, and available facilities. Therefore, the impact of the barriers and their interdependencies also vary. The author has extended the study and used the fuzzy approach MICMAC to improve the model by incorporating the variations observed when the city category changes. This study will help urban developers and planners to identify the relationship between barriers and design strategic planning to make the city smarter. [ABSTRACT FROM AUTHOR]

It is always difficult and challenge to obtain suitable trading signals for the desired securities in financial markets. The popular way to deal with it is through the use of trading strategies (TSs) made up of technical or fundamental indicators. Due to the different properties of TSs, an algorithm was proposed to find trading signals by obtaining the group trading strategy portfolio (GTSP), which is composed of strategy groups that can be employed to generate various TS portfolios (TSP) instead of a single TS. The stop-loss and take-profit points (SLTP) are widely utilized by shareholders to avoid massive losses. However, the appropriate SLTP is hard to set by users. Therefore, in this paper, the algorithm, namely GTSP-SLTP algorithm, is proposed to not only obtain a reliable GTSP but also find appropriate SLTP using the grouping genetic algorithm. A chromosome is encoded by the generated SLTP and GTSP along with the weights for strategy groups that are the SLTP, grouping, weight, and strategy parts. To assess the goodness of a chromosome, the evaluation function that consists of the group balance, weight balance, risk factor, and profit factor, is employed. Genetic operators are then performed to produce new solutions for next population. The genetic process is performed iteratively until the stop conditions have achieved. Last but not the least, empirical experiments were conducted on three financial datasets with different trends and a case study is also given to reveal the effectiveness and robustness of the designed GTSP-SLTP algorithm. [ABSTRACT FROM AUTHOR]

To solve the problem of low recognition accuracy and long recognition time in conventional methods for digital moving image feature recognition, a method for digital image feature recognition on mobile platforms based on machine learning is proposed. The support vector machine method is used to cluster the dataset, and then the weight reconstruction method is used to reduce the dimension of digital moving image data. The reduced dimension of the digital moving image is embedded in the coordinates to obtain the reduced dimension of the image data information and complete the feature recognition of the digital moving image. The simulation results show that the recognition accuracy of this method in recognizing features of moving digital images is basically stable at more than 95% and the running time is 0.42s. The recognition effect of the proposed method is good, the recognition efficiency is high, and it has some practical application value. [ABSTRACT FROM AUTHOR]

Traditional motion amplitude normalization technology cannot get the whole process of motion amplitude in the time and space domain, and there is a big error in motion image contour extraction, which leads to low accuracy and long time. Therefore, a motion amplitude normalized feature localization technique based on intelligent vision is proposed. The contour of the moving image is extracted by the method of color information segmentation. The motion process in time and space is simulated by intelligent vision, and the feature data are captured by the parameters of the human motion sensor. According to the horizontal rotation record and geomagnetic flux of the sensor, the feature data is processed by normalization, and the moving image features are selected by normalization. The amplitude function is established to optimize the foreground region feature marker of the motion image, and the motion amplitude standard feature location is realized. Experimental results show that the proposed method is more accurate and less time-consuming than the traditional method. [ABSTRACT FROM AUTHOR]

In the Internet of Medical Things (IoMT), collaboration among institutes can help complex medical and clinical analysis of disease. Deep neural networks (DNN) require training models on large, diverse patients to achieve expert clinician-level performance. Clinical studies do not contain diverse patient populations for analysis due to limited availability and scale. DNN models trained on limited datasets are thereby constraining their clinical performance upon deployment at a new hospital. Therefore, there is significant value in increasing the availability of diverse training data. This research proposes institutional data collaboration alongside an adversarial evasion method to keep the data secure. The model uses a federated learning approach to share model weights and gradients. The local model first studies the unlabeled samples classifying them as adversarial or normal. The method then uses a centroid-based clustering technique to cluster the sample images. After that, the model predicts the output of the selected images, and active learning methods are implemented to choose the sub-sample of the human annotation task. The expert within the domain takes the input and confidence score and validates the samples for the model's training. The model re-trains on the new samples and sends the updated weights across the network for collaboration purposes. We use the InceptionV3 and VGG16 model under fabricated inputs for simulating Fast Gradient Signed Method (FGSM) attacks. The model was able to evade attacks and achieve a high accuracy rating of 95%. [ABSTRACT FROM AUTHOR]

In many recent applications, a graph is used to simulate many complex systems, such as social networks, traffic models or bioinformatics, and the underlying graphs for these systems are very large. Algorithms for mining all frequent subgraphs from a single large graph have attracted much attention and been studied in more detail lately. Mining frequent subgraphs is important, and defined as finding all subgraphs whose occurrences in a dataset are greater than or equal to a given frequency threshold. Among frequent subgraph algorithms, GraMi is considered as the state-of-the-art approach. However, GraMi has a huge search space, and therefore still needs a lot of time and memory to process a large graph. In this paper, we propose two effective strategies to balance and reduce the search space of GraMi, which can decrease the number of candidate subgraphs generated, with early pruning of a large portion of the domain for each candidate. Our experiments were performed on four real datasets and the results show that the performance of our balancing GraMi is better than those of the original algorithm GraMi and the optimized version SoGraMi. [ABSTRACT FROM AUTHOR]

The high average-utility itemset mining (HAUIM) was established to provide a fair measure instead of genetic high-utility itemset mining (HUIM) for revealing the satisfied and interesting patterns. In practical applications, the database is dynamically changed when insertion/deletion operations are performed on databases. Several works were designed to handle the insertion process but fewer studies focused on processing the deletion process for knowledge maintenance. In this paper, we then develop a PRE-HAUI-DEL algorithm that utilizes the pre-large concept on HAUIM for handling transaction deletion in the dynamic databases. The pre-large concept is served as the buffer on HAUIM that reduces the number of database scans while the database is updated particularly in transaction deletion. Two upper-bound values are also established here to reduce the unpromising candidates early which can speed up the computational cost. From the experimental results, the designed PRE-HAUI-DEL algorithm is well performed compared to the Apriori-like model in terms of runtime, memory, and scalability in dynamic databases. [ABSTRACT FROM AUTHOR]

This research work introduces a new intelligent framework for infectious disease detection by exploring various emerging and intelligent paradigms. We propose new deep learning architectures such as entity embedding networks, long-short term memory, and convolution neural networks, for accurately learning heterogeneous medical data in identifying disease infection. The multi-agent system is also consolidated for increasing the autonomy behaviours of the proposed framework, where each agent can easily share the derived learning outputs with the other agents in the system. Furthermore, evolutionary computation algorithms, such as memetic algorithms, and bee swarm optimization controlled the exploration of the hyper-optimization parameter space of the proposed framework. Intensive experimentation has been established on medical data. Strong results obtained confirm the superiority of our framework against the solutions that are state of the art, in both detection rate, and runtime performance, where the detection rate reaches 98% for handling real use cases. [ABSTRACT FROM AUTHOR]

Large annotated datasets are needed for successful Deep Learning methodologies to achieve human-level performance. These needs restrict the impact of Deep Learning and build the necessity to create smaller and richer representative datasets that can offer a potential solution to this problem. In this paper, we propose task-specific image corpus summarization using semantic information and self-supervision. Our methodology makes use of GAN for the generation of features and leverages rotational invariance for employing self-supervision. All these objectives are facilitated on features from Resnet34. A summary can be obtained efficiently by using k-means clustering on the semantic embedding space and then selecting examples nearest to centroids. In comparison to end-to-end trained models, the proposed model does not require retraining to obtain summaries of different lengths. We also test our model by extensive qualitative and quantitative experiments. [ABSTRACT FROM AUTHOR]

Background: Plant diseases significantly affect the crop, so their identification is very important. Correct identification of these diseases is crucial for establishing a good disease control strategy to avoid time and financial losses. In general, machines can greatly reduce the possibility of human error. In particular, computer vision techniques developed through deep learning have paved a way to detect and diagnose these plant diseases on the leaf. Methods: In this work, the model AFD-Net was developed to detect and identify various leaf diseases in apple trees. The dataset is from Kaggle 2020 and 2021 and was financially supported by the Cornell Initiative for Digital Agriculture. An AFD-Net was proposed for leaf disease classification in apple trees and the results of the efficiency of the model are compared with other state-of-the-art deep learning approaches. Results: The results of the experiments in the validation dataset show that the proposed AFD-Net model achieves the highest values of 98.7% accuracy for Plant Pathology 2020 and 92.6% for Plant Pathology 2021 compared to other deep learning models in the original and extended datasets. Discussion: The results also indicate the efficiency of the proposed model in identifying leaf diseases on apple trees for major and minor classes, i.e., for multiple classification. [ABSTRACT FROM AUTHOR]

HUIM has been an important issue in recent years, particularly in basket-market analysis, since it identifies useful information or goods for decision-making. Numerous research focused on extracting high-utility itemsets from datasets, revealing a tremendous amount of pattern information. This approach is incapable of providing correct choices in a short amount of time, e.g., real-time and online decision-making systems since it is difficult to extract relevant and important information in a short period of time from a huge body of found knowledge. Discovering closed patterns with high utilization (or closed pattern mining with high utility) is a market engineering method that discovers fewer but lucrative patterns. However, prior research has been unable to handle huge data, which is incompatible with today's Internet of Things (IoT) environments, where huge volumes of data are collected every second. We begin by introducing the multi-objective model for mining the closed high utility itemsets (called MCUI-Miner), which employs MapReduce frameworks of a Spark structure. To begin, the multi-objective k-means the model is used to categorize transactions based on their significant relationship to the frequency component. The MapReduce model and GA are used to examine potential and probable candidates for mining closed high-utility itemsets in a large-scale database. Experiments have shown that the proposed framework outperforms the conventional CLS-Miner in terms of runtime, memory use, and scalability. [ABSTRACT FROM AUTHOR]

In order to improve the quality of distance education and solve the problem of slow data processing of the teaching system, an intelligent distance education assistance system based on WEB is developed in this paper. After verification, students, teachers and administrators log into the distance intelligent teaching assistance system and transmit various information to the interface of the teaching and administration subsystems. The submitted information is merged using the Bayesian model for integrating educational resources in the digital cloud to create a distance education database that supports the system with data. At the same time, the evaluation of the business logic of the data is performed. After the data is converted to other formats in the subsequent convertible database, it returns to the user interface to provide browsing and consulting functions for users. The experimental results show that the designed system can realize the remote auxiliary function of intelligent education and effectively improve the quality of teaching. The real-time data acquisition rate of the system in this paper is always equal to the set value. The average acceleration of the system in this paper is 5.5 and the data processing efficiency is higher. The minimum safety factor of the system is between 7.8 and 8.5, and the system has high stability. The user satisfaction of the system is over 93%, and the accuracy of the collected data is relatively high. The auxiliary system designed in this paper can provide a stable and efficient application environment for distance education. [ABSTRACT FROM AUTHOR]

The Internet of Things is made up of a set of embedded devices such as light detection sensors, ranging (LiDAR) sensors, and millimetre-wave sensors. These sensors generate a massive amount of data in which limited communication capacity is available to share a massive amount of data to Fog-based Roadside Units (RSU) for data process and analysis service. Fog-based RSU has become an emerging paradigm in intelligent transportation but needs research attention to design intelligent decision-making methods for data communication and computation at Fog-based RSU. To address these issues, we design a two-level Quantum based D2D Computation, Communication ( Q D C 2 ) approach. First, design a bandwidth allocation strategy based on spatial importance score factors to resolve embedded devices' data transmission issues. Second, design an adaptive equilibrium service offloading strategy based on device-centric measurements to assess the computation capacity and performance rate for resolving Fog-node computation consistency issues. Additionally, Fog-based RSU is interconnected with LAN helps to optimize service latency. Simulation results show that our approach achieved a high service reliability rate (79.56%), low error rate (0.9%), and an execution delay of 22.5 s for 15 devices compared to state-of-the-art approaches. [ABSTRACT FROM AUTHOR]

Multi-source remote sensing images have the characteristics of large differences in texture and gray level. Mismatch and low recognition accuracy are easy to occur in the process of identifying targets. Thus, in this paper, the target recognition algorithm of multi-source remote sensing image based on IoT vision is investigated. The infrared sensor and SAR radars are set in the visual perception layer of the iVIOT. The visual perception layer transmits the collected remote sensing image information to the application layer through the wireless networks. The data processing module in the application layer uses the normalized central moment idea to extract the features of multi-source remote sensing image. Contourlet two-level decomposition is performed on the image after feature extraction to realize multi-scale and multi-directional feature fusion. A two-step method of primary fineness is used to match the fused features and the random sampling consensus algorithm is used to eliminate false matches for obtaining the correct match pairs. After the image feature matching is completed, the BVM target detection operator is used to complete the target recognition of multi-source remote sensing image. Experimental results show that the use of the IoT to visually recognizing the desired remote sensing image target has low communication overhead, and the recognition reaches 99% accuracy. [ABSTRACT FROM AUTHOR]

Recently, revealing more valuable information except for quantity value for a database is an essential research field. High utility itemset mining (HAUIM) was suggested to reveal useful patterns by average-utility measure for pattern analytics and evaluations. HAUIM provides a more fair assessment than generic high utility itemset mining and ignores the influence of the length of itemsets. There are several high-performance HAUIM algorithms proposed to gain knowledge from a disorganized database. However, most existing works do not concern the uncertainty factor, which is one of the characteristics of data gathered from IoT equipment. In this work, an efficient algorithm for HAUIM to handle the uncertainty databases in IoTs is presented. Two upper-bound values are estimated to early diminish the search space for discovering meaningful patterns that greatly solve the limitations of pattern mining in IoTs. Experimental results showed several evaluations of the proposed approach compared to the existing algorithms, and the results are acceptable to state that the designed approach efficiently reveals high average utility itemsets from an uncertain situation. [ABSTRACT FROM AUTHOR]

Discovering periodic patterns consists of identifying all sets of items (values) that periodically co-occur in a discrete sequence. Although traditional periodic pattern mining algorithms have multiple applications, they have two key limitations. First, they consider that a pattern is not periodic if the time difference between two of its successive occurrences is greater than a maxPer threshold. But this constraint is too strict, as a pattern may be discarded based on only two of its occurrences, although it may be usually periodic. Second, traditional algorithms use a constraint that the support (occurrence frequency) of a pattern must be no less than a minSup threshold. But setting that parameter is not intuitive. Hence, it is usually set by trial and error, which is time-consuming. This paper addresses the first limitation by introducing a concept of stability to find periodic patterns that have a stable periodic behavior. Then, the second limitation is addressed by proposing an algorithm named TSPIN (Top-k Stable Periodic pattern mINer) to find the top-k stable periodic patterns, where the user can directly specify the number of patterns k to be found rather than using the minSup threshold. Several experiments have been performed to assess TSPIN's performance, and it was found that it is efficient and can discover patterns that reveal interesting insights in real data. [ABSTRACT FROM AUTHOR]

The article titled "Retraction Note: Robust adversarial uncertainty quantification for deep learning fine-tuning" has been retracted by the Editor-in-Chief at the author's request. An investigation conducted by the Western Norway University of Applied Sciences revealed that parts of the article bear significant resemblance to previously published articles by different authors. All authors of the article have agreed to the retraction. Springer Nature, the publisher, maintains a neutral stance on jurisdictional claims and institutional affiliations. The full text of the retracted article can be accessed online. [Extracted from the article]

Portfolio management involves position sizing and resource allocation. Traditional and generic portfolio strategies require forecasting of future stock prices as model inputs, which is not a trivial task since those values are difficult to obtain in the real-world applications. To overcome the above limitations and provide a better solution for portfolio management, we developed a Portfolio Management System (PMS) using reinforcement learning with two neural networks (CNN and RNN). A novel reward function involving Sharpe ratios is also proposed to evaluate the performance of the developed systems. Experimental results indicate that the PMS with the Sharpe ratio reward function exhibits outstanding performance, increasing return by 39.0% and decreasing drawdown by 13.7% on average compared to the reward function of trading return. In addition, the proposed PMS_CNN model is more suitable for the construction of a reinforcement learning portfolio, but has 1.98 times more drawdown risk than the PMS_RNN. Among the conducted datasets, the PMS outperforms the benchmark strategies in TW50 and traditional stocks, but is inferior to a benchmark strategy in the financial dataset. The PMS is profitable, effective, and offers lower investment risk among almost all datasets. The novel reward function involving the Sharpe ratio enhances performance, and well supports resource-allocation for empirical stock trading. [ABSTRACT FROM AUTHOR]

High utility itemset mining is a popular pattern mining task, which aims at revealing all sets of items that yield a high profit in a transaction database. Although this task is useful to understand customer behavior, an important limitation is that high utility itemsets do not provide information about the purchase quantities of items. Recently, some algorithms were designed to address this issue by finding quantitative high utility itemsets but they can have very long execution times due to the larger search space. This paper addresses this issue by proposing a novel efficient algorithm for high utility quantitative itemset mining, called FHUQI-Miner (Fast High Utility Quantitative Itemset Miner). It performs a depth-first search and adopts two novel search space reduction strategies, named Exact Q-items Co-occurrence Pruning Strategy (EQCPS) and Range Q-items Co-occurrence Pruning Strategy (RQCPS). Experimental results show that the proposed algorithm is much faster than the state-of-art HUQI-Miner algorithm on sparse datasets. [ABSTRACT FROM AUTHOR]

Effective vector representation has been proven useful for transaction classification and clustering tasks in Cyber-Physical Systems. Traditional methods use heuristic-based approaches and different pruning strategies to discover the required patterns efficiently. With the extensive and high dimensional availability of transactional data in cyber-physical systems, traditional methods that used frequent itemsets (FIs) as features suffer from dimensionality, sparsity, and privacy issues. In this paper, we first propose a federated learning-based embedding model for the transaction classification task. The model takes transaction data as a set of frequent item-sets. Afterward, the model can learn low dimensional continuous vectors by preserving the frequent item-sets contextual relationship. We perform an in-depth experimental analysis on the number of high dimensional transactional data to verify the developed models with attention-based mechanism and federated learning. From the results, it can be seen that the designed model can help and improve the decision boundary by reducing the global loss function while maintaining both security and privacy. [ABSTRACT FROM AUTHOR]

The multi-dimensional characteristics of public opinion in online education lead to the difficulty of data cross-dimensional mining. To solve this problem, this paper designs a cross-dimensional data mining model of public opinion in online education based on fuzzy association rules. Based on the public opinion subject, object, and ontology to analyze the characteristics of public opinion in online education, Yaahp software is used to calculate the influence factor weight of public opinion in online education. According to the weight analysis results, the relationship between the dimensions of various public opinion data is clarified by using data semantic association. This paper introduces the fuzzy set theory into the database and uses crawlers to obtain public opinion data and stores them in the database, to complete the data preprocessing through distributed text preprocessing, feature selection distributed computing, and text vectorization distributed computing. Taking the cloud computing platform as the core, the cross-dimension mining model of public opinion in online education data is constructed according to the dimension correlation analysis and preprocessing results. The simulation results show that the model has the advantages of wide range, fast speed, and high accuracy, and can provide data support for education reform. [ABSTRACT FROM AUTHOR]

Traditional association-rule mining (ARM) considers only the frequency of items in a binary database, which provides insufficient knowledge for making efficient decisions and strategies. The mining of useful information from quantitative databases is not a trivial task compared to conventional algorithms in ARM. Fuzzy-set theory was invented to represent a more valuable form of knowledge for human reasoning, which can also be applied and utilized for quantitative databases. Many approaches have adopted fuzzy-set theory to transform the quantitative value into linguistic terms with its corresponding degree based on defined membership functions for the discovery of FFIs, also known as fuzzy frequent itemsets. Only linguistic terms with maximal scalar cardinality are considered in traditional fuzzy frequent itemset mining, but the uncertainty factor is not involved in past approaches. In this paper, an efficient fuzzy mining (EFM) algorithm is presented to quickly discover multiple FFIs from quantitative databases under type-2 fuzzy-set theory. A compressed fuzzy-list (CFL)-structure is developed to maintain complete information for rule generation. Two pruning techniques are developed for reducing the search space and speeding up the mining process. Several experiments are carried out to verify the efficiency and effectiveness of the designed approach in terms of runtime, the number of examined nodes, memory usage, and scalability under different minimum support thresholds and different linguistic terms used in the membership functions. [ABSTRACT FROM AUTHOR]

This paper introduces a secure and scalable intelligent transportation and human behavior system to accurately discover knowledge from urban traffic data. The data are secured using blockchain learning technology, where the scalability is ensured by a threaded GPU. In addition, different optimizations are provided to efficiently process data on the GPU. A reinforcement deep learning algorithm is also established to merge local knowledge discovered on each site into global knowledge. To demonstrate the applicability of the proposed framework, intensive experiments have been carried out on well-known intelligent transportation and human behavior data. Our results show that our proposed framework outperforms the baseline solutions for the outlier detection use case. [ABSTRACT FROM AUTHOR]

This paper addresses the problem of responding to user queries by fetching the most relevant object from a clustered set of objects. It addresses the common drawbacks of cluster-based approaches and targets fast, high-quality information retrieval. For this purpose, a novel cluster-based information retrieval approach is proposed, named Cluster-based Retrieval using Pattern Mining (CRPM). This approach integrates various clustering and pattern mining algorithms. First, it generates clusters of objects that contain similar objects. Three clustering algorithms based on k-means, DBSCAN (Density-based spatial clustering of applications with noise), and Spectral are suggested to minimize the number of shared terms among the clusters of objects. Second, frequent and high-utility pattern mining algorithms are performed on each cluster to extract the pattern bases. Third, the clusters of objects are ranked for every query. In this context, two ranking strategies are proposed: i) Score Pattern Computing (SPC), which calculates a score representing the similarity between a user query and a cluster; and ii) Weighted Terms in Clusters (WTC), which calculates a weight for every term and uses the relevant terms to compute the score between a user query and each cluster. Irrelevant information derived from the pattern bases is also used to deal with unexpected user queries. To evaluate the proposed approach, extensive experiments were carried out on two use cases: the documents and tweets corpus. The results showed that the designed approach outperformed traditional and cluster-based information retrieval approaches in terms of the quality of the returned objects while being very competitive in terms of runtime. [ABSTRACT FROM AUTHOR]

When focusing on the general area of data mining, high-utility itemset mining (HUIM) can be defined as an offset of frequent itemset mining (FIM). It is known to emphasize more factors critically, which gives HUIM its intrinsic edge. Due to the flourishing development of the IoT technique, the uncertainty patterns mining is also attractive. Potential high-utility itemset mining (PHUIM) is introduced to reveal valuable patterns in an uncertainty database. Unfortunately, even though the previous methods are all very effective and powerful to mine, the potential high-utility itemsets quickly. These algorithms are not specifically designed for a database with an enormous number of records. In the previous methods, uncertainty transaction datasets would be load in the memory ultimately. Usually, several pre-defined operators would be applied to modify the original dataset to reduce the seeking time for scanning the data. However, it is impracticable to apply the same way in a big-data dataset. In this work, a dataset is assumed to be too big to be loaded directly into memory and be duplicated or modified; then, a MapReduce framework is proposed that can be used to handle these types of situations. One of our main objectives is to attempt to reduce the frequency of dataset scans while still maximizing the parallelization of all processes. Through in-depth experimental results, the proposed Hadoop algorithm is shown to perform strongly to mine all of the potential high-utility itemsets in a big-data dataset and shows excellent performance in a Hadoop computing cluster. [ABSTRACT FROM AUTHOR]

Analysis of medical CT images directly affects the accuracy of clinical case diagnosis. Therefore, feature extraction problem of medical CT images is extremely important. A feature extraction algorithm for medical CT images of sports tear injury is proposed. First, CT images are decomposed into a low frequency component and a series of high frequency components in different directions by wavelet fast decomposition method. The high- and low-frequency information of CT images is enhanced by wavelet layered multi-directional image enhancement algorithm, and the multi-scale enhancement for medical CT images of sports tear injury is completed. Then, edge of the enhanced CT images is extracted using an image edge extraction algorithm based on extended mathematical morphology. Finally, based on the extracted edge information of CT images, feature extraction for medical CT images of sports tear injury is completed by the NSCT-GLCM based CT image feature extraction algorithm. Research results show that the proposed algorithm effectively extracts CT image features of sports tear injury and provides auxiliary information for doctor diagnosis. [ABSTRACT FROM AUTHOR]

Nowadays, embedded systems are comprised of heterogeneous multi-core architectures, i.e., CPUs and GPUs. If the application is mapped to an appropriate processing core, then these architectures provide many performance benefits to applications. Typically, programmers map sequential applications to CPU and parallel applications to GPU. The task mapping becomes challenging because of the usage of evolving and complex CPU- and GPU-based architectures. This paper presents an approach to map the OpenCL application to heterogeneous multi-core architecture by determining the application suitability and processing capability. The classification is achieved by developing a machine learning-based device suitability classifier that predicts which processor has the highest computational compatibility to run OpenCL applications. In this paper, 20 distinct features are proposed that are extracted by using the developed LLVM-based static analyzer. In order to select the best subset of features, feature selection is performed by using both correlation analysis and the feature importance method. For the class imbalance problem, we use and compare synthetic minority over-sampling method with and without feature selection. Instead of hand-tuning the machine learning classifier, we use the tree-based pipeline optimization method to select the best classifier and its hyper-parameter. We then compare the optimized selected method with traditional algorithms, i.e., random forest, decision tree, Naïve Bayes and KNN. We apply our novel approach on extensively used OpenCL benchmarks, i.e., AMD and Polybench. The dataset contains 653 training and 277 testing applications. We test the classification results using four performance metrics, i.e., F-measure, precision, recall and R 2 . The optimized and reduced feature subset model achieved a high F-measure of 0.91 and R 2 of 0.76. The proposed framework automatically distributes the workload based on the application requirement and processor compatibility. [ABSTRACT FROM AUTHOR]

High-utility itemset mining (HUIM) is considered as an emerging approach to detect the high-utility patterns from databases. Most existing algorithms of HUIM only consider the itemset utility regardless of the length. This limitation raises the utility as a result of a growing itemset size. High average-utility itemset mining (HAUIM) considers the size of the itemset, thus providing a more balanced scale to measure the average-utility for decision-making. Several algorithms were presented to efficiently mine the set of high average-utility itemsets (HAUIs) but most of them focus on handling static databases. In the past, a fast-updated (FUP)-based algorithm was developed to efficiently handle the incremental problem but it still has to re-scan the database when the itemset in the original database is small but there is a high average-utility upper-bound itemset (HAUUBI) in the newly inserted transactions. In this paper, an efficient framework called PRE-HAUIMI for transaction insertion in dynamic databases is developed, which relies on the average-utility-list (AUL) structures. Moreover, we apply the pre-large concept on HAUIM. A pre-large concept is used to speed up the mining performance, which can ensure that if the total utility in the newly inserted transaction is within the safety bound, the small itemsets in the original database could not be the large ones after the database is updated. This, in turn, reduces the recurring database scans and obtains the correct HAUIs. Experiments demonstrate that the PRE-HAUIMI outperforms the state-of-the-art batch mode HAUI-Miner, and the state-of-the-art incremental IHAUPM and FUP-based algorithms in terms of runtime, memory, number of assessed patterns and scalability. [ABSTRACT FROM AUTHOR]

This paper investigates the use of recurrent neural network to predict urban long-term traffic flows. A representation of the long-term flows with related weather and contextual information is first introduced. A recurrent neural network approach, named RNN-LF, is then proposed to predict the long-term of flows from multiple data sources. Moreover, a parallel implementation on GPU of the proposed solution is developed (GRNN-LF), which allows to boost the performance of RNN-LF. Several experiments have been carried out on real traffic flow including a small city (Odense, Denmark) and a very big city (Beijing). The results reveal that the sequential version (RNN-LF) is capable of dealing effectively with traffic of small cities. They also confirm the scalability of GRNN-LF compared to the most competitive GPU-based software tools when dealing with big traffic flow such as Beijing urban data. [ABSTRACT FROM AUTHOR]

This paper explores five pattern mining problems and proposes a new distributed framework called DT-DPM: Decomposition Transaction for Distributed Pattern Mining. DT-DPM addresses the limitations of the existing pattern mining problems by reducing the enumeration search space. Thus, it derives the relevant patterns by studying the different correlation among the transactions. It first decomposes the set of transactions into several clusters of different sizes, and then explores heterogeneous architectures, including MapReduce, single CPU, and multi CPU, based on the densities of each subset of transactions. To evaluate the DT-DPM framework, extensive experiments were carried out by solving five pattern mining problems (FIM: Frequent Itemset Mining, WIM: Weighted Itemset Mining, UIM: Uncertain Itemset Mining, HUIM: High Utility Itemset Mining, and SPM: Sequential Pattern Mining). Experimental results reveal that by using DT-DPM, the scalability of the pattern mining algorithms was improved on large databases. Results also reveal that DT-DPM outperforms the baseline parallel pattern mining algorithms on big databases. [ABSTRACT FROM AUTHOR]

Ontology matching aims at identifying the correspondences between instances and data properties of different ontologies. The use of data mining approach in matching ontology problem is reviewed in this article. We propose DMOM (Data Mining for Ontology Matching based instances) framework to select data properties of instances efficiently. The framework exploits data mining techniques to select the most appropriate features to match ontologies. Moreover, three strategies have been investigated to select the relevant features for the matching process. The first one called exhaustive, explores the enumerate search tree randomly by generating at each iteration a subset of feature attributes, where each node is evaluated by running the matching process on its selected attributes. The second approach called statistical, it uses some statistical values to select the most relevant properties. The third one called FIM (Frequent Itemsets Mining), it explores the correlation between different properties and selects the most frequent properties describing the overall instances of the given ontology. To demonstrate the usefulness of DMOM framework, several experiments have been carried out on OAEI (Ontology Alignment Evaluation Initiative) and DBpedia ontology databases. The results show that the third strategy, FIM, outperforms the two other strategies (Exhaustive, and Statistical). The results also reveal that DMOM outperforms the state-of-the-art ontology matching approaches in terms of execution time and the quality of the matching process. [ABSTRACT FROM AUTHOR]