Search

Feature selection, both for supervised as well as for unsupervised classification is a relevant problem pursued by researchers for decades. There are multiple benchmark algorithms based on filter, wrapper and hybrid methods. These algorithms adopt different techniques which vary from traditional search-based techniques to more advanced nature inspired algorithm based techniques. In this paper, a hybrid feature selection algorithm using graph-based technique has been proposed. The proposed algorithm has used the concept of Feature Association Map (FAM) as an underlying foundation. It has used graph-theoretic principles of minimal vertex cover and maximal independent set to derive feature subset. This algorithm applies to both supervised and unsupervised classification. The performance of the proposed algorithm has been compared with several benchmark supervised and unsupervised feature selection algorithms and found to be better than them. Also, the proposed algorithm is less computationally expensive and hence has taken less execution time for the publicly available datasets used in the experiments, which include high-dimensional datasets. [ABSTRACT FROM AUTHOR]

This paper presents an improved version of the well-established k nearest neighbor ( k -NN) and fuzzy NN (FNN), termed the multi-objective genetic-algorithm-modified FNN (MOGA-MFNN). The MFNN design problem is converted into a multi-modal objective maximization problem constrained by four objective functions. Thereafter, the associated parameter set of the MFNN and the feature attributes can be determined optimally and automatically via the non-dominated sorting genetic algorithm II. We introduce two new objective functions termed the Margin-I and Margin-II, which are used to improve the generalization capability of the MFNN for the unknown data, along with two existing performance functions: the geometric mean and the area under the receiver-operated characteristic curve for the training accuracy. Moreover, we proposed a novel data-dependent weight-assignment technique for local class membership functions of the MFNN. The technique enables the MFNN to determine its local neighbors adaptively through the MOGA algorithm. To expedite the classification, the MOGA-MFNN is implemented on a graphical processing unit (GPU), which significantly increases the computation speed. Furthermore, the local class-membership function of the MFNN can be computed in advance, rather than delaying it to the classification stage. This again can improve the classification speed. The MOGA-MFNN is evaluated on 20 datasets obtained from the repository of the University of California, Irvine (UCI). The experiments with rigorous statistical significance tests demonstrate that the proposed method performs competitively with the existing methods. [ABSTRACT FROM AUTHOR]

Learning from imbalanced datasets is challenging for standard algorithms, as they are designed to work with balanced class distributions. Although there are different strategies to tackle this problem, methods that address the problem through the generation of artificial data constitute a more general approach compared to algorithmic modifications. Specifically, they generate artificial data that can be used by any algorithm, not constraining the options of the user. In this paper, we present a new oversampling method, Self-Organizing Map-based Oversampling (SOMO), which through the application of a Self-Organizing Map produces a two dimensional representation of the input space, allowing for an effective generation of artificial data points. SOMO comprises three major stages: Initially a Self-Organizing Map produces a two-dimensional representation of the original, usually high-dimensional, space. Next it generates within-cluster synthetic samples and finally it generates between cluster synthetic samples. Additionally we present empirical results that show the improvement in the performance of algorithms, when artificial data generated by SOMO are used, and also show that our method outperforms various oversampling methods. [ABSTRACT FROM AUTHOR]

With the advent of Big Data, data is being collected at an unprecedented fast pace, and it needs to be processed in a short time. To deal with data streams that flow continuously, classical batch learning algorithms cannot be applied and it is necessary to employ online approaches. Online learning consists of continuously revising and refining a model by incorporating new data as they arrive, and it allows important problems such as concept drift or management of extremely high-dimensional datasets to be solved. In this paper, we present a unified pipeline for online learning which covers online discretization, feature selection and classification. Three classical methods—the k -means discretizer, the χ 2 filter and a one-layer artificial neural network—have been reimplemented to be able to tackle online data, showing promising results on both synthetic and real datasets. [ABSTRACT FROM AUTHOR]

• This paper studies the knapsack problem with setups. • This paper presents an algorithm based on a Lagrangean relaxation of the problem. • We used very large data sets to test the algorithm. • The algorithm can solve near optimally the problem in reasonable amount of time. We consider the knapsack problem with setups which is a generalization of the classical knapsack problem where the items belong to families and an item can be placed in the knapsack only if its family is selected. The problem has received increasing attention by researchers because of its theoretical significance and practical applications related to resource allocation. This paper presents an algorithm based on a Lagrangean relaxation of the problem that produces solutions whose quality can be assessed automatically with the algorithm itself without ever knowing the optimal solutions. We report results of an extensive computational study which show that the method can solve near optimally very large instances of the problem with up to 500 families and 2,000,000 items in reasonable amount of time. This study shows the merit of using the Lagrangean relaxation method to solve the current problem when the constraints to relax are chosen properly. [ABSTRACT FROM AUTHOR]

• Proposing a fast and accurate version of DBSCAN algorithm for big data. • Calculating density based on a small subset of data called operational set. • Creating and updating the operational set at a very low computational cost. • Comprehensive evaluation of the proposed method in compared to other recent works. Today, data is being generated with a high speed. Managing large volume of data has become a challenge in the current age. Clustering is a method to analyze data that is generated in the Internet. Various approaches have been presented for data clustering until now. Among them, DBSCAN is a most well-known density-based clustering algorithm. This algorithm can detect clusters of different shapes and does not require prior knowledge about the number of clusters. A major part of the DBSCAN run-time is spent to calculate the distance of data from each other to find the neighbors of each sample in the dataset. The time complexity of this algorithm is O(n2); Therefore, it is not suitable for processing big datasets. In this paper, DBSCAN is improved so that it can be applied to big datasets. The proposed method calculates accurately each sample density based on a reduced set of data. This reduced set is called the operational set. This collection is updated periodically. The use of local samples to calculate the density has greatly reduced the computational cost of clustering. The empirical results on various datasets of different sizes and dimensions show that the proposed algorithm increases the clustering speed compared to recent related works while having similar accuracy as the original DBSCAN algorithm. [ABSTRACT FROM AUTHOR]

The recent developments in cellular networks, along with the increase in services, users and the demand of high quality have raised the Operational Expenditure (OPEX). Self-Organizing Networks (SON) are the solution to reduce these costs. Within SON, self-healing is the functionality that aims to automatically solve problems in the radio access network, at the same time reducing the downtime and the impact on the user experience. Self-healing comprises four main functions: fault detection, root cause analysis, fault compensation and recovery. To perform the root cause analysis (also known as diagnosis), Knowledge-Based Systems (KBS) are commonly used, such as fuzzy logic. In this paper, a novel method for extracting the Knowledge Base for a KBS from solved troubleshooting cases is proposed. This method is based on data mining techniques as opposed to the manual techniques currently used. The data mining problem of extracting knowledge out of LTE troubleshooting information can be considered a Big Data problem. Therefore, the proposed method has been designed so it can be easily scaled up to process a large volume of data with relatively low resources, as opposed to other existing algorithms. Tests show the feasibility and good results obtained by the diagnosis system created by the proposed methodology in LTE networks. [ABSTRACT FROM AUTHOR]

• A cellular-based evolutionary fuzzy system for the extraction of emerging patterns in high-speed, massive data streams is proposed. • Smart triggering of the learning method which updates the model only when required. • A reinitialisation and filtering strategy for reducing the extraction of redundant patterns is also defined. • The quality of knowledge extracted outperforms state-of-the-art methods. • The proposed method is able to process batches of data up to 750,000 instances. Today, the number of existing devices generates immense amounts of data on a continuous basis that must be processed by new distributed data stream mining approaches. In this paper we present a new approach for extracting descriptive emerging patterns in massive data streams from different sources through Apache Kafka and Apache Spark Streaming whose objective is to monitor the state of the system with respect to a variable of interest. For this purpose, the proposed algorithm is a cellular-based multi-objective evolutionary fuzzy system that uses an informed strategy for efficient data processing and a re-initialisation and filtering mechanism to eliminate redundant and low-reliable patterns. The experimental study carried out demonstrates an interpretability improvement of 25% in the extraction of high-interest knowledge by the proposed algorithm, which would make it easier for experts to analyse the problem. Finally, the proposed algorithm is up to five times faster than another proposal on the processing of the same amount of data. In this experimental study, up to 750,000 instances have been processed in approximately four seconds. [ABSTRACT FROM AUTHOR]

• The challenge of algorithm runtime has been addressed. • The graph modification step of the algorithm is focused. • The proposed algorithm has two steps of adding and deleting edges. • It uses an efficient algorithm to select the most appropriate edges. • It improves the algorithm runtime and the utility of the graph. Privacy on social networks is one of the most important and well-known issues. Various algorithms have been proposed to preserve the privacy of social network, all of which try to change the graph structure such that the utility of the graph is maintained. Although these algorithms have been successful in protecting the privacy and the utility of social networks, they are not suitable for anonymizing big data, because of the high cost of processing. Some of these algorithms have a high runtime. In addition, they should be improved from the aspect of preserving graph utility. In this paper, an effective algorithm has been introduced to increase the anonymization speed, as well as improving the graph utility. This algorithm uses number factorization to remove the best edges from the graph in the graph modification step of the algorithm. Since the appropriate edges are selected just through one scan of the edges, the runtime is reduced. In order to add edges to the graph, all the appropriate edges are selected simultaneously and added to the graph using NaFa algorithm. Some limitations are applied to the selection of edges to choose the correct ones. Taking into account the appropriate criteria to select edges in both the removal and addition steps improves the graph utility. The evaluation results of the proposed algorithm on real data sets show the efficiency of the algorithm to sharply reduce the runtime and increase the graph utility, simultaneously. [ABSTRACT FROM AUTHOR]

• Integrates multi-objective in an evolutionary-based technique for subspace clustering. • A first attempt in simultaneously optimizing the subspace feature set and compactness. • A new validity index, PSM-index designed to optimizing the subspace feature sets. • A modified and a new mutation operator is developed to explore the search space. • Application of the proposed method is shown in bi-clustering the gene expression data. Subspace clustering technique divides the data set into different groups or clusters where each cluster comprises of objects that share some similar properties. Again, the feature sets or the subspace features that are used to represent clusters are different for different clusters. Moreover, in subspace clustering, the grouping of similar objects and the subspace feature set representing that group are identified simultaneously. In evolutionary-based machine learning problems, two critical measures to determine the quality of the generated clusters are compactness within and separation between the clusters. However, the distance-based separation between two clusters may not be useful in the context of subspace clustering, as the clusters may belong to two different subspaces. Again, in the case of subspace clustering, the selection of relevant subspace features plays a primary role in generating good quality subspace clusters. Therefore, the proposed approach optimizes the subspace features by considering two new objective functions, feature non-redundancy (FNR) and feature per cluster (FPC) represented in the form of PSM-index. Another objective function, intra-cluster compactness (ICC-index), is modified and used to optimize the compactness among objects within the cluster. Finally, an evolutionary-based multi-objective subspace clustering technique is developed in this paper optimizing these validity indices. A new mutation operator, namely duplication and deletion along with the modified version of the exogenous genetic material uptake , are developed to explore the search space effectively. The developed algorithm is tested on sixteen synthetic data sets and seven standard real-life data sets for identifying different subspace clusters. Again, to show the effectiveness of using multiple objectives, the algorithm is also tested on three big data sets and a MNIST data set. Also, an application of the proposed method is shown in bi-clustering the gene expression data. The results obtained by the proposed algorithm are compared against some state-of-the-art methods. Experimentation reveals that the proposed algorithm can take advantage of its evolvable genomic structure and the newly defined objective functions on the multi-objective based framework. [ABSTRACT FROM AUTHOR]

• We propose a Semantic-kNN (Sk-NN) algorithm for ML. • Sk-NN deploys a process by leveraging on the semantic itemization. • Sk-NN used to filter the training dataset. • The proposed method is more efficient, intelligent and better accuracy. The k-NN algorithm is one of the most renowned ML algorithms widely used in the area of data classification research. With the emergence of big data, the performance and the efficiency of the traditional k-NN algorithm is fast becoming a critical issue. The traditional k-NN algorithm is inefficient to solve the high volume multi-categorical training datasets Traditional k-NN algorithm has a constraint in filtering the training dataset to yield training data that are most relevant to the intended or the targeted test dataset/file. It has to scan through all the training datasets categories to classify the intended/targeted data. As such, traditional k-NN is considered not intelligent and consequently is suffering poor accuracy performance with high computational complexity. A Semantic-kNN (Sk-NN) algorithm for ML is thus proposed in this paper to address the limitations in the traditional k-NN. The proposed Sk-NN deploys a process by leveraging on the semantic itemization and bigram model to filter the training dataset in accordance with the relevant information engaged in the test dataset. It is aimed for general security applications such as finding (the confidentiality level of the data when the algorithm is trained with multiple training categories during the data classification phase. Ultimately, Sk-NN is to elevate the ML performance in pattern extraction and labeling in the big data context. [ABSTRACT FROM AUTHOR]