Search

In this paper, a hybrid sunflower-teaching learning-based optimization (SFO-TLBO) algorithm has been proposed, integrating two well-known algorithms, namely sunflower optimization, and teaching–learning-based optimization. The proposed hybrid SFO-TLBO algorithm has been incorporated based on the ratio of the fitness value of the search agent and the average fitness value of the population. The SFO has more tendency of global exploration but has a problem of slow convergence speed. On the other hand, the convergence of the TLBO algorithm is fast, but it may get stuck in local optima. In this work, we have hybridized SFO with TLBO to take advantage of both the algorithms and also to balance the trade-off between exploitation and exploration. The proposed hybrid SFO-TLBO has been first applied to nineteen test benchmark functions. The outcomes have been compared with twelve state-of-the-art algorithms. In seventeen benchmark test functions, better performance has been achieved by the proposed hybrid SFO-TLBO. Also, the proposed hybrid SFO-TLBO algorithm has been used to solve an Internet of Vehicles (IoV) optimization problem and performance has been compared with fifteen state-of-the-art algorithms. In solving the IoV problem, hybrid SFO-TLBO has been achieved better performance in comparison with all fifteen considered algorithms which showcases the applicability of hybrid SFO-TLBO algorithm in resolving large-scale, real-world problems. In addition, for feature selection problems, a binary version of hybrid SFO-TLBO has been proposed, in which a wrapper approach is used to decide an optimal subset of features with a k-nearest neighbor classifier. The proposed binary method also determines an optimal value of k for the k-nearest neighbor classifier. The binary hybrid SFO-TLBO algorithm has been applied to the QSAR biodegradation dataset for classification, and for evidencing its efficiency, its performance has been compared with the other state-of-the-art algorithms. The results of the proposed approach are quite encouraging, and an average accuracy of 89.10% has been achieved. [ABSTRACT FROM AUTHOR]

Recently, some statistical topic modeling approaches based on LDA have been applied in the field of supervised document classification, where the model generation procedure incorporates prior knowledge to improve the classification performance. However, these customizations of topic modeling are limited by the cumbersome derivation of a specific inference algorithm for each modification. In this paper, we propose a new supervised topic modeling approach for document classification problems, Neural Labeled LDA (NL-LDA), which builds on the VAE framework, and designs a special generative network to incorporate prior information. The proposed model can support semi-supervised learning based on the manifold assumption and low-density assumption. Meanwhile, NL-LDA has a consistent and concise inference method while semi-supervised learning and predicting. Quantitative experimental results demonstrate our model has outstanding performance on supervised document classification relative to the compared approaches, including traditional statistical and neural topic models. Specially, the proposed model can support both single-label and multi-label document classification. The proposed NL-LDA performs significantly well on semi-supervised classification, especially under a small amount of labeled data. Further comparisons with related works also indicate our model is competitive with state-of-the-art topic modeling approaches on semi-supervised classification. [ABSTRACT FROM AUTHOR]

Class imbalance problems have attracted the research community, but a few works have focused on feature selection with imbalanced datasets. To handle class imbalance problems, we developed a novel fitness function for feature selection using the chaotic salp swarm optimization algorithm, an efficient meta-heuristic optimization algorithm that has been successfully used in a wide range of optimization problems. This paper proposes an AdaBoost algorithm with chaotic salp swarm optimization. The most discriminating features are selected using salp swarm optimization, and AdaBoost classifiers are thereafter trained on the features selected. Experiments show the ability of the proposed technique to find the optimal features with performance maximization of AdaBoost. [ABSTRACT FROM AUTHOR]

An intrusion detection system is a security system that aims to detect sabotage and intrusions on networks to inform experts of the attack and abuse of the network. Different classification methods have been used in the intrusion detection systems such as fuzzy, genetic algorithms, decision trees, artificial neural networks, and support vector machines. Moreover, ensemble classifiers have shown more robust and effective performance for various tasks in the field. In this paper, we adopt ensemble models in order to improve the performance of intrusion detection and, at the same time, decrease the false alarm rate. We use kNN for multi-class classification, as well as SVM to approach the classification problem in normal-based detection. In order to combine multiple outputs, we use the Dempster–Shafer method in which there is the possibility of explicit retrieval of uncertainty. Moreover, we utilize deep learning for extracting features to train the samples, selected by the sample selection algorithm based on ensemble margin. We compare our results with state-of-the-art methods on benchmarking datasets such as UNSW-NB15, CICIDS2017, and NSL-KDD. Our proposed method indicates the superiority in terms of prominent metrics Accuracy, Precision, Recall, and F-measure. [ABSTRACT FROM AUTHOR]

The diagnosis of heart disease is found to be a serious concern, so the diagnosis has to be done remotely and regularly to take the prior action. In the present years, the diagnosis of heart disease has become a key research area for researchers and many models have been proposed in recent years. The diagnosis of heart disease can be done using optimization algorithms, and it provides results with good efficiency. The main objective of this paper is to propose a hybrid fuzzy-based decision tree algorithm for the process of prediction of heart disease at an early stage through the continuous and remote patient monitoring system. The results obtained from the proposed algorithm are compared with the various number of classifier algorithms like decision tree J48, naïve Bayes, GA with FCM, KNN with NB, ANN, SVM with fuzzy in which the proposed HFDT algorithm provides better accuracy of 98.30%. From the above-obtained results, the proposed hybrid fuzzy-based decision tree algorithm efficiently predicts heart disease compared to the other classifier algorithms in the literature. The proposed work is implemented in the MATLAB environment using the heart disease dataset. [ABSTRACT FROM AUTHOR]

In this paper, improved salp swarm algorithm is proposed. The algorithm integrates (1) random opposition-based learning (2) multiple leadership and (3) simulated annealing in swarm intelligence-based metaheuristic salp swarm algorithm. This integration increases the exploration and exploitation of the original salp swarm algorithm. Hence, the effectiveness of the proposed algorithm is better for complex multi-modal problems. The algorithm is tested on several standard numerical benchmark functions and CEC-2015 benchmarks. Results are compared with some well-known metaheuristics. The results represent the merit of the proposed algorithm with respect to other algorithms. The improved salp swarm algorithm is applied for feed-forward neural network training. Performance is compared with other metaheuristic-based feedforward neural network trainers for different data sets. The results show the efficiency and effectiveness of proposed algorithm in solving complex multi-modal problems. [ABSTRACT FROM AUTHOR]

Classification is one of the foremost machine learning tasks in this modern era. Neural Network (NN) is one of the powerful classification techniques. NN can achieve high classification accuracy on highly imbalanced and complex datasets, but lacks in explanation of its reasoning process which limits its applicability in various domains which require transparent decision along with good accuracy. There are some techniques which extract rules from NN and make it transparent; however, attribute pruning, rule pruning and class overlap algorithms are not sufficiently effective. Therefore, this paper proposes a rule extraction algorithm, called Transparent Rule Extraction using Neural Network (TRENN) to convert NN into white box with greater emphasis on attribute pruning and rule pruning. The proposed TRENN is a pedagogical approach and an extension of one of the existing algorithms named Rule Extraction from Neural Network using Classified and Misclassified data (RxNCM). The proposed TRENN extends the RxNCM with sequential floating backward search for feature and rule selection to improve the comprehensibility of the generated rules. Besides, the proposed TRENN uses probabilistic approach for the treatment of class overlapping problem in the rule updating phase instead of reclassification used in RxNCM where the overlap may persist. Experiments are conducted with eight real datasets collected from the UCI repository. Performance of the TRENN is measured with Precision, Recall, FP-Rate, F-measure, and local and global comprehensibility. It is observed from the experimental results that TRENN performs better than Re-RX, RxNCM and RxREN. [ABSTRACT FROM AUTHOR]

In the bioinformatics field, the classification of unknown biological sequences is a key task that is fundamental for simplifying the consistency, aggregation, and survey of organisms and their evolution. We can view biological sequences as data components of higher non-fixed dimensions, corresponding to the length of the sequences. Numerical encoding performs an important function in DNA sequence evaluation via computational procedures such as one-hot encoding (OHE). However, the OHE method has drawbacks: 1) it does not add any details that may produce the additional predictive variable, and 2) if the variable has many classes, then OHE increases the feature space significantly. To overcome these drawbacks, this paper presents a computationally effective framework for classifying DNA sequences of living organisms in the image domain. The proposed strategy relies upon multilayer perceptron trained by a pseudoinverse learning autoencoder (PILAE) algorithm. The PILAE training process does not have to set the learning control parameters or indicate the number of hidden layers. Therefore, the PILAE classifier can accomplish better performance contrasting with other deep neural network (DNNs) strategies such as VGG-16 and Xception models. Experimental results have demonstrated that this proposed strategy achieves high prediction accuracy as well as to a significant degree high computational efficiency over different datasets. [ABSTRACT FROM AUTHOR]

Feature selection (FS) is the process of finding the least possible number of features that are able to describe a dataset in the same way as the original features. Feature selection is a crucial preprocessing step for data mining techniques as it improves the performance of the prediction process in terms of speed and accuracy and also provides a better understanding of stored data. The success of the FS process depends on achieving a balance between two important factors, namely selecting the minimal number of features and maintaining the maximum accuracy in the results. In this paper, two methods are proposed to improve the FS process. Firstly, the mine blast algorithm (MBA) is introduced to optimize the FS process in the exploration phase. Secondly, the MBA is hybridized with simulated annealing as a local search in the exploitation phase to enhance the solutions located by the MBA. The proposed approaches (MBA and MBA–SA) are tested on 18 benchmark datasets from the UCI repository, and the comprehensive experimental results indicate that MBA–SA achieved good performance when compared with five approaches in the literature. [ABSTRACT FROM AUTHOR]

The qualities of new data used in the sequential learning phase of the online sequential extreme learning machine algorithm (OS-ELM) have a significant impact on the performance of OS-ELM. This paper proposes a novel data filter mechanism for OS-ELM from the perspective of fuzziness and a fuzziness-based online sequential extreme learning machine algorithm (FOS-ELM). In FOS-ELM, when new data arrive, a fuzzy classifier first picks out the meaningful data according to the fuzziness of each sample. Specifically, the new samples with high-output fuzziness are selected and then used in sequential learning. The experimental results on eight binary classification problems and three multiclass classification problems have shown that FOS-ELM updated by the new samples with high-output fuzziness has better generalization performance than OS-ELM. Since the unimportant data are discarded before sequential learning, FOS-ELM can save more memory and have higher computational efficiency. In addition, FOS-ELM can handle data one-by-one or chunk-by-chunk with fixed or varying sizes. The relationship between the fuzziness of new samples and the model performance is also studied in this paper, which is expected to provide some useful guidelines for improving the generalization ability of online sequential learning algorithms. [ABSTRACT FROM AUTHOR]

Feature selection is one of the hottest machine learning topics in recent years. The main purposes of it are to simplify the original model, improve the readability of the model, and prevent over-fitting by searching for a suitable subset of features. There are many methods for this problem, including evolutionary algorithms and particle swarm optimization. Among them, the competitive swarm optimizer is a new optimization algorithm proposed in recent years, which is based on particle swarm optimization algorithm, and has achieved good results in high-dimensional feature selection problems, but it also has the problems of high computation time cost and easily being premature. Aiming at these problems, this paper proposes to add the crossover operator and mutation operator in the genetic algorithm to the competitive swarm optimization, so as to improve the generation speed of new individuals in the algorithm and prevent premature population. After testing on UC Irvine Machine Learning Repository, the new algorithm not only improves the computational efficiency, but also avoids the problem that the competitive swarm optimization algorithm is easy to fall into the local optimum, which greatly improves the calculation effect. [ABSTRACT FROM AUTHOR]

Document representation is a key problem in document analysis and processing tasks, such as document classification, clustering and information retrieval. Especially for unstructured text data, the use of a suitable document representation method would affect the performance of the subsequent algorithms for applications and research. In this paper, we propose a novel document representation method called the conditional co-occurrence degree matrix document representation method (CCODM), which is based on word co-occurrence. CCODM not only considers the co-occurrence of terms but also considers the conditional dependencies of terms in a specific context, which leads to more available and useful structural and semantic information being retained from the original documents. Extensive experimental classification results with different supervised and unsupervised feature selection methods show that the proposed method, CCODM, achieves better performance than the vector space model, latent Dirichlet allocation, the general co-occurrence matrix representation method and the document embedding method. [ABSTRACT FROM AUTHOR]

The nearest neighbor rule is one of the most considered algorithms for supervised learning because of its simplicity and fair performance in most cases. However, this technique has a number of disadvantages, being the low computational efficiency the most prominent one. This paper presents a strategy to overcome this obstacle in multi-class classification tasks. This strategy proposes the use of Prototype Reduction algorithms that are capable of generating a new training set from the original one to try to gather the same information with fewer samples. Over this reduced set, it is estimated which classes are the closest ones to the input sample. These classes are referred to as promising classes. Eventually, classification is performed using the original training set using the nearest neighbor rule but restricted to the promising classes. Our experiments with several datasets and significance tests show that a similar classification accuracy can be obtained compared to using the original training set, with a significantly higher efficiency. [ABSTRACT FROM AUTHOR]

In many real-world applications, instances (data) arrive sequentially in the form of streams. Processing such data poses challenges to machine learning. While adhering to on-line learning strategies, in this paper we extend the Flexible Fuzzy Decision Tree (FlexDT) algorithm with multiple partitioning that makes it possible to carry out automatic on-line fuzzy data classification. The proposed method is aimed to balance accuracy and tree size in data stream mining. The objective of the classification problem is to predict the true class of each incoming instances in real time. In terms of evaluation of the method, accuracy, tree depth, and the learning time are significant factors influencing the performance. A series of experiments demonstrate that the proposed method produces optimal trees for both numeric and nominal features (variables). [ABSTRACT FROM AUTHOR]

The changes of face images with poses and polarized illuminations increase data uncertainty in face recognition. In fact, synthesized mirror samples can be recognized as representations of the left-right deflection of poses or illuminations of the face. Symmetrical face images generated from the original face images also provide more observations of the face which is useful for improving the accuracy of face recognition. In this paper, to the best of our knowledge, it is the first time that the well-known minimum squared error classification (MSEC) algorithm is used to perform face recognition on an extended face database using synthesized mirror training samples, which is titled as extended minimum squared error classification (EMSEC). By modifying the MSE classification rule, we append the mirror samples to the training set for gaining better classification performance. First, we merge original training samples and mirror samples synthesized from original training samples per subject as mixed training samples. Second, EMSEC algorithm exploits mixed training samples to obtain the projection matrix that can best transform the mixed training samples into predefined class labels. Third, the projection matrix is exploited to simultaneously obtain transform results of the test sample and its nearest neighbor from the mixed training sample set. Finally, we ultimately classify the test sample by combining the transform results of the test sample and the nearest neighbor. As an extension of MSEC, EMSEC reduces the uncertainty of the face observation by auxiliary mirror samples, so that it has better robustness classification performance than traditional MSEC. Experimental results on the ORL, GT, and FERET databases show that EMSEC has better generalization ability than traditional MSEC. [ABSTRACT FROM AUTHOR]

Supervised discretization is one of basic data preprocessing techniques used in data mining. CAIM (class-attribute interdependence maximization) is a discretization algorithm of data for which the classes are known. However, new arising challenges such as the presence of unbalanced data sets, call for new algorithms capable of handling them, in addition to balanced data. This paper presents a new discretization algorithm named ur-CAIM, which improves on the CAIM algorithm in three important ways. First, it generates more flexible discretization schemes while producing a small number of intervals. Second, the quality of the intervals is improved based on the data classes distribution, which leads to better classification performance on balanced and, especially, unbalanced data. Third, the runtime of the algorithm is lower than CAIM's. The algorithm has been designed free-parameter and it self-adapts to the problem complexity and the data class distribution. The ur-CAIM was compared with 9 well-known discretization methods on 28 balanced, and 70 unbalanced data sets. The results obtained were contrasted through non-parametric statistical tests, which show that our proposal outperforms CAIM and many of the other methods on both types of data but especially on unbalanced data, which is its significant advantage. [ABSTRACT FROM AUTHOR]

In this paper, we present a method for the filteriNg of ovEr-sampled dAta using non-cooperaTive gamE theoRy (NEATER) to address the imbalanced data problem. Specifically, the problem is formulated as a non-cooperative game where all the data are players and the goal is to uniformly and consistently label all of the synthetic data created by any over-sampling technique. The proposed algorithm does not require any prior assumptions and selects representative synthetic instances while generating a very small number of noisy data. We present extensive experimental results over a large collection of datasets using three different classifiers to demonstrate the advantages of our method. [ABSTRACT FROM AUTHOR]

This paper introduces an electromagnetism-like (EM) approach for solving the problem of parameter tuning in the support vector machine (SVM). The proposed method is used to tune binary SVM classifiers in single and multiple kernel mode. The internal kernel structure is based on linear and radial basis functions (RBF). An appropriate encoding scheme of EM enables easy transformation of real-valued EM points directly to real-valued parameter combinations. Estimations of the generalization error based on the cross-validation and validation set error are used as objective functions. The efficient local search procedure uses variable size interval movement in order to improve the convergence of the method. The quality of the proposed method is tested on four collections of testing benchmarks through five separate experiments. The first three collections consist of small-size to medium-size classification data sets with up to 60 features and 1,300 training vectors, while the fourth collection is formed of large heterogeneous data sets with up to 1,554 features and 2,186 training vectors. The obtained results indicate that EM outperforms the comparison algorithms in 10 out of 13 instances from the first collection, 5 out of 5 instances from the second, and 13 out of 15 instances from the third collection. The last two experiments, conducted on the fourth collection, show that the proposed method outperforms 14 successful methods in 3 out of 5 data sets where RBF multiple kernel learning is used, and behaves competitively in cases when linear kernels are used. [ABSTRACT FROM AUTHOR]

K-nearest neighbours algorithms are among the most popular existing classification methods, due to their simplicity and good performances. Over the years, several extensions of the initial method have been proposed. In this paper, we propose a K-nearest neighbours approach that uses the theory of imprecise probabilities, and more specifically lower previsions. We show that the proposed approach has several assets: it can handle uncertain data in a very generic way, and decision rules developed within this theory allow us to deal with conflicting information between neighbours or with the absence of close neighbour to the instance to classify. We show that results of the basic k-NN and weighted k-NN methods can be retrieved by the proposed approach. We end with some experiments on the classical data sets. [ABSTRACT FROM AUTHOR]

The efficiency of evolutionary algorithms has become a studied problem since it is one of the major weaknesses in these algorithms. Specifically, when these algorithms are employed for the classification task, the computational time required by them grows excessively as the problem complexity increases. This paper proposes an efficient scalable and massively parallel evaluation model using the NVIDIA CUDA GPU programming model to speed up the fitness calculation phase and greatly reduce the computational time. Experimental results show that our model significantly reduces the computational time compared to the sequential approach, reaching a speedup of up to 820×. Moreover, the model is able to scale to multiple GPU devices and can be easily extended to any evolutionary algorithm. [ABSTRACT FROM AUTHOR]

In the classification framework there are problems in which the number of examples per class is not equitably distributed, formerly known as imbalanced data sets. This situation is a handicap when trying to identify the minority classes, as the learning algorithms are not usually adapted to such characteristics. An usual approach to deal with the problem of imbalanced data sets is the use of a preprocessing step. In this paper we analyze the usefulness of the data complexity measures in order to evaluate the behavior of undersampling and oversampling methods. Two classical learning methods, C4.5 and PART, are considered over a wide range of imbalanced data sets built from real data. Specifically, oversampling techniques and an evolutionary undersampling one have been selected for the study. We extract behavior patterns from the results in the data complexity space defined by the measures, coding them as intervals. Then, we derive rules from the intervals that describe both good or bad behaviors of C4.5 and PART for the different preprocessing approaches, thus obtaining a complete characterization of the data sets and the differences between the oversampling and undersampling results. [ABSTRACT FROM AUTHOR]

To avoid the problems of slow and premature convergence of the differential evolution (DE) algorithm, this paper presents a new DE variant named p-ADE. It improves the convergence performance by implementing a new mutation strategy 'DE/rand-to-best/pbest', together with a classification mechanism, and controlling the parameters in a dynamic adaptive manner, where the 'DE/rand-to-best/pbest' utilizes the current best solution together with the best previous solution of each individual to guide the search direction. The classification mechanism helps to balance the exploration and exploitation of individuals with different fitness characteristics, thus improving the convergence rate. Dynamic self-adaptation is beneficial for controlling the extent of variation for each individual. Also, it avoids the requirement for prior knowledge about parameter settings. Experimental results confirm the superiority of p-ADE over several existing DE variants as well as other significant evolutionary optimizers. [ABSTRACT FROM AUTHOR]

This paper presents a new evolutionary cooperative–competitive algorithm for the design of Radial Basis Function Networks (RBFNs) for classification problems. The algorithm, CO2RBFN, promotes a cooperative–competitive environment where each individual represents a radial basis function (RBF) and the entire population is responsible for the final solution. The proposal considers, in order to measure the credit assignment of an individual, three factors: contribution to the output of the complete RBFN, local error and overlapping. In addition, to decide the operators’ application probability over an RBF, the algorithm uses a Fuzzy Rule Based System. It must be highlighted that the evolutionary algorithm considers a distance measure which deals, without loss of information, with differences between nominal features which are very usual in classification problems. The precision and complexity of the network obtained by the algorithm are compared with those obtained by different soft computing methods through statistical tests. This study shows that CO2RBFN obtains RBFNs with an appropriate balance between accuracy and simplicity, outperforming the other methods considered. [ABSTRACT FROM AUTHOR]

Abstract  Aiming at the common support vector machine’s biased disadvantage and computational complexity, an unbiased least squares support vector machine (LSSVM) model is proposed in this paper. The model eliminates the bias item of LSSVM by improving the form of structure risk, then the unbiased least squares support vector classifier and the unbiased least squares support vector regression are deduced. Based on this model, we design a new learning algorithm using Cholesky factorization according to the characteristic of kernel function matrix, in this way the calculation of Lagrangian multipliers is greatly simplified. Several experiments on diffenert datasets are carried out, including the common datasets classification, synthetic aperture radar image automatic target recognition and chaotic time series prediction. The experimental results of correct recognition rate and the fitting precision testify that the unbiased LSSVM model has good universal ability and fitting accuracy, better generalization capability and stability, and have a great improvement in learning speed. [ABSTRACT FROM AUTHOR]