BACKGROUND: DNA sequences are fundamental carriers of genetic information, and their accurate classification is essential for understanding gene regulation, disease mechanisms, and translational genomics. Existing encoding methods often fail to capture both local and long-range dependencies simultaneously. RESULTS: We introduce EDEN (Expected Density of Nucleotide Encoding), a unified multiscale encoding framework based on kernel density estimation. EDEN captures position-specific and context-dependent nucleotide patterns and integrates them into a hybrid deep learning architecture. Across sixteen benchmark datasets covering promoter detection, core promoter detection, and transcription factor binding prediction, EDEN achieves the best average performance while using orders of magnitude fewer parameters compared with state-of-the-art models. All source code, pretrained models, and datasets are publicly available at: https://github.com/zabihis/EDEN . CONCLUSIONS: EDEN provides an efficient, biologically informed, and interpretable multiscale representation for genomic sequence classification. Its favorable parameter-performance ratio and robust consistency across tasks underscore its practicality for large-scale genomic applications.

Summary The increasing popularity of Internet of Things (IoT) devices makes them an attractive target for malware authors. In this paper, we use sequential pattern mining technique to detect most frequent opcode sequences of malicious IoT applications. Detected maximal frequent patterns (MFP) of opcode sequences can be used to differentiate malicious from benign IoT applications. We then evaluate the suitability of MFPs as a classification feature for K nearest neighbors (KNN), support vector machines (SVM), multilayer perceptron (MLP), AdaBoost, decision tree, and random forest classifier. Specifically, we achieve an accuracy rate of 99% in the detection of unseen IoT malware. We also demonstrate the utility of our approach in detecting polymorphed IoT malware samples.

In bug fixing process, estimating the ‘Time to Fix Bug’ is one of the factors that helps the triager to allocate jobs in a better way. Due to the limitation of resources for bug fixing, the bugs with long fixing time must be identified, as soon as possible, after receiving the report. This helps the prioritisation and fixing process of the bug reports. In the process of bug fixing, a temporal sequence of activities is done. Each activity is represented by a term. Useful semantic information and long‐term dependency are available between terms in the sequence, but it is usually underutilised by existing bug fixing time predictor approaches. This work presents a novel deep learning‐based model (called DeepLSTMPred) that (i) converts constituent terms to a vector of real numbers by considering their semantic meaning, (ii) finds the long‐term dependencies between terms by deep long short term memory (LSTM) and (iii) classifies sequences to short fixing time or long fixing time. DeepLSTMPred is evaluated on bug reports extracted from the Mozilla project. The results show that the proposed method has better performance in comparison with a state‐of‐the‐art approach (that is the hidden Markov‐based model). The experimental results show that DeepLSTMPred achieves 15–20% improvement in terms of accuracy, precision, f ‐score, and recall.

Ransomware, a malware designed to encrypt data for ransom payments, is a potential threat to fog layer nodes as such nodes typically contain considerably amount of sensitive data. The capability to efficiently hunt abnormalities relating to ransomware activities is crucial in the timely detection of ransomware. In this paper, we present our Deep Ransomware Threat Hunting and Intelligence System (DRTHIS) to distinguish ransomware from goodware and identify their families. Specifically, DRTHIS utilizes Long Short-Term Memory (LSTM) and Convolutional Neural Network (CNN), two deep learning techniques, for classification using the softmax algorithm. We then use 220 Locky, 220 Cerber and 220 TeslaCrypt ransomware samples, and 219 goodware samples, to train DRTHIS. In our evaluations, DRTHIS achieves an F-measure of 99.6% with a true positive rate of 97.2% in the classification of ransomware instances. Additionally, we demonstrate that DRTHIS is capable of detecting previously unseen ransomware samples from new ransomware families in a timely and accurate manner using ransomware from the CryptoWall, TorrentLocker and Sage families. The findings show that 99% of CryptoWall samples, 75% of TorrentLocker samples and 92% of Sage samples are correctly classified.

One of the serious challenges in machine learning and pattern recognition is to transfer knowledge from related but different domains to a new unlabeled domain. Feature selection with maximum mean discrepancy (f-MMD) is a novel and effective approach to transfer knowledge from source domain (training set) into target domain (test set) where training and test sets are drawn from different distributions. However, f-MMD has serious challenges in facing datasets with large number of samples and features. Moreover, f-MMD ignores the feature-label relation in finding the reduced representation of dataset. In this paper, we exploit jointly transfer learning and class discrimination to cope with domain shift problem on which the distribution difference is considerably large. We therefore put forward a novel transfer learning and class discrimination approach, referred to as RandOm k-samplesets feature Weighting Approach (ROWA). Specifically, ROWA reduces the distribution difference across domains in an unsupervised manner where no label is available in the test set. Moreover, ROWA exploits feature-label relation to separate various classes alongside the domain transfer, and augments the relation of selected features and source domain labels. In this work, we employ disjoint/overlapping small-sized samplesets to iteratively converge to final solution. Employment of local sets along with a novel optimization problem constructs a robust and effective reduced representation for adaptation across domains. Extensive experiments on real and synthetic datasets verify that ROWA can significantly outperform state-of-the-art transfer learning approaches.

Class imbalance problem is quite pervasive in our nowadays human practice. This problem basically refers to the skewness in the data underlying distribution which, in turn, imposes many difficulties on typical machine learning algorithms. To deal with the emerging issues arising from multi-class skewed distributions, existing efforts are mainly divided into two categories: model-oriented solutions and data-oriented techniques. Focusing on the latter, this paper presents a new over-sampling technique which is inspired by Mahalanobis distance. The presented over-sampling technique, called MDO (Mahalanobis Distance-based Over-sampling technique), generates synthetic samples which have the same Mahalanobis distance from the considered class mean as other minority class examples. By preserving the covariance structure of the minority class instances and intelligently generating synthetic samples along the probability contours, new minority class instances are modelled better for learning algorithms. Moreover, MDO can reduce the risk of overlapping between different class regions which are considered as a serious challenge in multi-class problems. Our theoretical analyses and empirical observations across wide spectrum multi-class imbalanced benchmarks indicate that MDO is the method of choice by offering statistical superior MAUC and precision compared to the popular over-sampling techniques.

Recently, the hybrid Particle Swarm Optimisation/Ant Colony Optimisation (PSO/ACO) has been proposed for discovery of classification rules. An improved version of this hybrid scheme, PSO/ACO2 algorithm, can directly cope with nominal attributes without converting them into numerical ones. Although PSO/ACO2 can handle nominal values, it suffers from high computational complexity for large datasets. Beside variety of classification methods which exist to provide more compact set of rules, this study propose an approach which reduces the computational complexity of PSO/ACO2 in order to make it suitable for classification of large datasets. This work is developed the K-mode as a method of sampling from the datasets. In this regard a modification is employed to this algorithm in order to decrease the computational time as well as maintaining the accuracy of the algorithm. Further contribution of this paper is utilizing a new fitness measure for the algorithm. This measure has a robust theoretical background. The combination of the proposed modified K-mode method and the introduced fitness measure led to speed the obtained results up. The experimental result shows the efficiency of the proposed algorithm in comparison with its competitors.

Recently, Indicator-based Evolutionary Algorithms are considered the main issue for researchers in the evolutionary multi-objective frameworks. Due to the capability of the Indicator-based approaches in obtaining a finest non-dominated solutions and the potential of these approaches on achieving th e well-distributed solutions, these approaches become popular among modern Multi-Objective Evolutionary Algorithms (MOEAs). Most modern MOEAs are intended to converge to the Pareto optimal front through preserving the population diversity in the objective space. In this regard, the intention of this work is presenting a novel MOEA to enhance the population diversity among the non-dominated vectors in the solution space. The idea of this method is inspired by the Hierarchical clustering. In this attitude, an adept approach is planned to present a new indicator as a selection method during the optimization cycle. The gain of this technique is a desirable set with more diverse solutions in the solution space during the environmental selection operator. In the last part, this work also improved the rate of the convergence by introducing a parent selection mechanism. The selection method is simple and effective, which is worked base on the selection of proper members of parents' population instead of a random mechanism. This bright parent selection is adopted to accelerate the convergence of the proposed method. This work is applied to a wide range of well-established test problems. The obtained results validate the motivation on the basis of diversity and performance measures in comparison with the state of the art algorithms.

Discovering communities in popular social networks like Facebook has been receiving significant attentions recently. In this paper, inspired from real life, we have addressed the community detection problem by a framework based on Information Diffusion Model and Game Theory. In this approach, we consider each node of the social network as a selfish agent which has interactions with its neighbors and tries to maximize its total utility (i.e. received information). Finally community structure of the graph reveals after reaching to the local Nash equilibrium of the game. Experimental results on the benchmark social media datasets, synthetic and real world graphs demonstrate that our method is superior compared with the other state-of-the-art methods.

Generalization is the most challenging issue in XCS research area. One of the main components of XCS managing to remedy this issue is knowledge representation. In this paper, a knowledge representation based on fuzzy membership function offering certain and vague regions is described. We extend the Michigan learning classifier system using this approach to be improved in terms of both performance and interpretability. The contribution of this paper is three-folds: 1) updating main parameters of classifiers based on their certainty factor in matching of incoming data, 2) enhancing essential components of XCS to be compatible with such fuzzy representation schema and 3) proposing a novel rule set reduction method named Reduction based on Least Reward Prediction (RLRP) to improve the interpretability of the evolved model. Furthermore, an inference methodology which is compatible with RLRP is suggested to maintain the similar performance. The obtained results are promising due to the effectiveness of proposed method in dealing with real world problems. Furthermore, the proposed reduction method can upgrade the interpretability of final rule set by boiling its size down by 94% on average while slightly degrading the prediction accuracy.

Distance learning is an important notion and has played a critical role in success of various machine learning algorithms. Any learning algorithm that requires dissimilarity/similarity measures has to assume some forms of distance functions, either explicitly or implicitly. Hence, in recent years a considerable amount of research has been devoted to distance learning. Despite great achievements in this field, a number of important issues need to be further explored for real world datasets mainly containing categorical attributes. Based on these considerations, the current research presents a Context-Based Distance Learning approach (CBDL) to advance the state of the art existing researches on distance metric learning for categorical datasets. CBDL is designed and developed based on the idea that distance between two values of a given categorical attribute can be estimated by using information inherently exists within subset of attributes called context. CBDL composes of two main components: context extraction component and distance learning component. Context extraction component is responsible for extracting the relevant subset of feature set for a given attribute, while distance learning component tries to learn distance between each pair of values based on the extracted context. To have a comprehensive analysis, we conduct wide range of experiments in both supervised and unsupervised environments in the presence of noise. Our experimental results reveal that CBDL is the method of choice distance learning approach by offering a comparable or better performance compared to the state of the art existing distance learning schemes according to studied evaluation measures. © 2011 Wiley Periodicals, Inc.

Previously we introduced Parallel Specialized XCS (PSXCS), a distributed-architecture classifier system that detects aliased environmental states and assigns their handling to created subordinate XCS classifier systems. PSXCS uses a history-window approach, but with novel efficiency since the subordinateXCSs, which employ the windows, are only spawned for parts of the state space that are actually aliased. However, because the window lengths are finite and set manually, PSXCS may fail to be optimal in difficult test mazes. This paper introduces Recursive PSXCS (RPSXCS) that automatically spawns windows wherever more history is required. Experimental results show that RPSXCS is both more powerful and learns faster than PSXCS. The present research suggests new potential for history approaches to partially observable environments.

Decision trees are one of the most effective and widely used induction methods that have received a great deal of attention over the past twenty years. When decision tree induction algorithms were used with uncertain rather than deterministic data, the result is a complete tree, which can classify most of the unseen samples correctly. This tree would be pruned in order to reduce its classification error and over-fitting. Recently, multi agent researchers concentrated on learning from large databases. In this paper we present a novel multi agent learning method that is able to induce a decision tree from distributed training sets. Our method is based on combination of separate decision trees each provided by one agent. Hence an agent is provided to aggregate results of the other agents and induces the final tree. Our empirical results suggest that the proposed method can provide significant benefits to distributed data classification.

One versus all (OVA) decision trees learn k individual binary classifiers, each one to distinguish the instances of a single class from the instances of all other classes. Thus OVA is different from existing data stream classification schemes whose majority use multiclass classifiers, each one to discriminate among all the classes. This paper advocates some outstanding advantages of OVA for data stream classification. First, there is low error correlation and hence high diversity among OVA's component classifiers, which leads to high classification accuracy. Second, OVA is adept at accommodating new class labels that often appear in data streams. However, there also remain many challenges to deploy traditional OVA for classifying data streams. First, as every instance is fed to all component classifiers, OVA is known as an inefficient model. Second, OVA's classification accuracy is adversely affected by the imbalanced class distribution in data streams. This paper addresses those key challenges and consequently proposes a new OVA scheme that is adapted for data stream classification. Theoretical analysis and empirical evidence reveal that the adapted OVA can offer faster training, faster updating and higher classification accuracy than many existing popular data stream classification algorithms.

Abstract: The purpose of the intrusion detection system (IDS) database is to detect transactions that access data without permission. This paper proposes a novel approach to identifying malicious transactions. The approach concentrates on two aspects of database transactions: (1) dependencies among data items and (2) variations of each individual data item which can be considered as time‐series data. The advantages are threefold. First, dependency rules among data items are extended to detect transactions that read or write data without permission. Second, a novel behaviour similarity criterion is introduced to reduce the false positive rate of the detection. Third, time‐series anomaly analysis is conducted to pinpoint intrusion transactions that update data items with unexpected pattern. As a result, the proposed approach is able to track normal transactions and detect malicious ones more effectively than existing approaches.