Detecting fraud accurately in credit cards is critical as this financial sector incurs significant losses for cardholders. Nonetheless, most studies adopted standard machine learning and few incremental learning, which are inadequate for addressing credit card challenges, such as rapid data arrival, unlimited data, data sensitivity, and performance decline over time. For this purpose, we propose a chunk-based incremental feature learning approach that optimises the fraud model topology for each new chunk and keeps track of one chunk each time. The model consists of several connected sub-models, where a new sub-model is optimally created for each new chunk. To avoid the network growing indefinitely, we limit the number of sub-models. To this end, we retain the most relevant sub-models to the current chunk’s data distribution and re-combine them to create the optimal model. We evaluate our approach using two credit card datasets: the first of medium scale contains 2-day payments in 2013, and the second of considerable scale possesses 6-month payments in 2019. We split these datasets into multiple chunks to learn and test incrementally. We compare our approach with static learning methods trained with different scenarios. Moreover, we vary the number of historical sub-models to check their impact on the predictive performance.

Data clustering has many applications in machine learning, data mining and image processing. K-means is the most popular clustering algorithm due to its efficiency and simplicity of implementation. However, K-means has limitations, such as large feature spaces, which may affect its effectiveness. To improve K-means accuracy, we adopt the Biogeography-Based Optimization (BBO) evolutionary technique to select the most relevant features of datasets. We conducted several experiments to compare our approach with other methods, such as PCA and Particle Swarm Optimization (PSO). The results demonstrate the effectiveness of BBO for feature selection.

Network traces represent a critical piece of data for network security. Due to lack of expertise, companies are forced to outsource their network traces to third parties to perform analytics on the traces and provide security feedback and recommendations. However, these companies are reluctant to share their network traces, as they comprise sensitive information (e.g., IP addresses). Therefore, the network traces are anonymized to ensure the privacy of the data and preserve its utility. The latter guarantees that the essence of the data remains valid after anonymization, otherwise the analytics are useless. Existing solutions, such as CryptoPAN, preserves the data utility (by preserving the IP prefixes), but are vulnerable to semantic attacks.In this paper, we propose an anonymization solution, which is based on the Feistel, which is widely used in encryption systems, such as DES and Twofish. Our solution preserves both data privacy and its utility at the same time. We validate our solution using the Kddcup99 dataset and measure the data leakage (dual of privacy) provided by our solution. We evaluate the security of our solution using the avalanche property, which is widely used to measure the security of encryption systems. Moreover, the efficacy of our solution is evaluated against Injection attacks. Overall, the obtained results, avalanche property and resistance to Injection attacks, are appealing.

This work presents a deep averaged reinforcement-learning approach to learn improvement heuristics for route planning. The proposed method is tested on the Traveling Salesman Problem (TSP). While learning improvement heuristics using machine learning models are prosperous, these methods suffer from low generalization and forgetfulness of the agents during the training process. We have applied the stochastic weight averaging method during the training phase to solve these issues, which smothers the training convergence and prevents the forgetting of optimized learned policies, and consequently provides better results. The agent can learn the optimized policy while holding a moving average of the previously learned policies during the training epochs. In order to assess the performance of our proposed approach, we conducted comparative experiments considering other known methods from the literature. The results demonstrate our proposed method’s superiority in training trends and optimization.

SAT problems are fundamental in representing and solving combinatorial applications. Over the past years, many sophisticated SAT solvers have been proposed. Due to the topic’s relevance, a SAT competition is scheduled yearly to promote solving hard SAT instances. There is no unique solver to tackle all SAT problems efficiently. Indeed, some solvers work best for some SAT instances but perform poorly for others. This limitation has been addressed, in the literature, by identifying a pool of solvers that complement each other for efficiently tackling a given set of SAT instances. This pool of solvers is called a portfolio. Several studies have been conducted to find the optimal portfolio maximizing the number of solved SAT instances, minimizing the overall running time, or a trade-off between both. In this context, we present a new approach that first finds the suitable portfolio meeting each of these objectives. Then, the approach predicts the best solver for any new SAT instance. Our approach is based on Greedy search techniques, clustering, and deep learning. More precisely, we investigate two different scenarios. In the first one, our goal is to find the best portfolio capable of solving the largest number of instances within a given time limit. Both the Greedy-based method and clustering are used in this case. The second scenario aims to find the optimal portfolio to minimize the penalized average running time. The latter objective captures a good trade-off between the objective in the first scenario and the overall average running time. In addition to Greedy search and clustering, we consider a variant of the Beam-search technique to address this scenario. To assess the performance of our approach regarding the two scenarios, we conduct multiple experiments on the SAT2021 competition datasets that include SAT instances together with participants’ solvers’ results for each instance. The outcomes from the conducted experiments are encouraging and promising.

Mobile devices are used by numerous applications that continuously need computing power to grow. Due to limited resources for complex computing, offloading, a service offered for mobile devices, is commonly used in cloud computing. In Mobile Cloud Computing (MCC), offloading decides where to execute the tasks to efficiently maximize the benefits. Hence, we represent offloading as a Task Scheduling Problem (TSP). This latter is a Multi-Objective Optimization (MOO) problem where the goal is to find the best schedule for processing mobile source tasks, while minimizing both the average processor energy consumption and the average task processing time. Owing to the combinatorial nature of the problem, the TSP in MCC is known as NP-hard. To overcome this difficulty in practice, we adopt meta-heuristic search techniques as they offer a good trade-off between solution quality and scalability. More precisely, we introduce a new optimization approach, that we call Multi-objective Discrete Water Cycle Algorithm (MDWCA), to schedule tasks from mobile source nodes to processor resources in a hybrid MCC architecture, including public cloud, cloudlets, and mobile devices. To evaluate the performance of our proposed approach, we conducted several comparative experiments on many generated TSP instances in MCC. The simulation results show that MDWCA outperforms the state-of-the-art optimization algorithms for several quality metrics.

Data-driven applications often change over time by considering new features to improve predictive accuracy. Re-training a model from scratch for every change loses the learned knowledge and is very time-consuming. To fill the big literature gap, we devise an incremental feature learning algorithm using constructive neural networks to include new groups of features gradually and without re-learning from scratch. The algorithm dynamically constructs the final model by determining the optimal network topology leading to the best performance. We demonstrate our algorithm’s efficacy through a regression problem by evaluating the sequential models obtained after extending the feature space incrementally, using different feature rankings. We also assess our algorithm without feature grouping and with the non-incremental learning version.

One key for improving hate speech detection performance is to have a textual training corpus that is vast and confidently labeled. This paper develops a semi-supervised learning approach with self-training to benefit from the abundant amount of social media content and develop a robust hate speech classifier for future predictions. The classifier is self-trained iteratively using the most confident pseudo labels obtained from a large-scale unlabelled Twitter corpus. We demonstrate our approach’s efficacy and the high quality of the produced supervised hate speech dataset through experiments.

Abstract For detecting malicious bidding activities in e‐auctions, this study develops a chunk‐based incremental learning framework that can operate in real‐world auction settings. The self‐adaptive framework first classifies incoming bidder chunks to counter fraud in each auction and take necessary actions. The fraud classifier is then adjusted with confident bidders' labels validated via bidder verification and one‐class classification. Based on real fraud data produced from commercial auctions, we conduct an extensive experimental study wherein the classifier is adapted incrementally using only relevant bidding data while evaluating the subsequent adjusted models' detection and misclassification rates. We also compare our classifier with static learning and learning without data relevancy.

Our study explores offensive and hate speech detection for the Arabic language, as previous studies are minimal. Based on two-class, three-class, and six-class Arabic-Twitter datasets, we develop single and ensemble CNN and BiLSTM classifiers that we train with non-contextual (Fasttext-SkipGram) and contextual (Multilingual Bert and AraBert) word-embedding models. For each hate/offensive classification task, we conduct a battery of experiments to evaluate the performance of single and ensemble classifiers on testing datasets. The average-based ensemble approach was found to be the best performing, as it returned F-scores of 91%, 84%, and 80% for two-class, three-class and six-class prediction tasks, respectively. We also perform an error analysis of the best ensemble model for each task.

Fraud detection systems aim to process a massive amount of data at high speed. To address the issues of data scalability, we introduce a chunk-based incremental classification approach based on a neural network (MLP) and a memory model to tackle the stability-plasticity dilemma. The incremental approach adapts the fraud model sequentially with incoming data chunks and retains past chunks a little more. We employ a large-scale credit-card fraud dataset that we organize into initial and incremental chunks for training and testing. Using data sampling, we solve the data skew problem, a critical issue in fraud detection. After each incremental phase, we evaluate the performance of the adjusted MLP classifier using the testing chunk. The experimental results demonstrate the effectiveness and efficiency of our incremental method and its superiority to the non-incremental MLP.

Constraint optimization consists of looking for an optimal solution maximizing a given objective function while meeting a set of constraints. In this study, we propose a new algorithm based on mushroom reproduction for solving constraint optimization problems. Our algorithm, that we call Mushroom Reproduction Optimization (MRO), is inspired by the natural reproduction and growth mechanisms of mushrooms. This process includes the discovery of rich areas with good living conditions allowing spores to grow and develop their own colonies. Given that constraint optimization problems often suffer from a high-time computation cost, we thoroughly assess MRO performance on well-known constrained engineering and real-world problems. The experimental results confirm the high performance of MRO, comparing to other known metaheursitcs, in dealing with complex optimization problems.

We introduce a new nature-inspired optimization algorithm namely Mushroom Reproduction Optimization (MRO) inspired and motivated by the reproduction and growth mechanisms of mushrooms in nature. MRO follows the process of discovering rich areas (containing goabod living conditions) by spores to grow and develop their own colonies. We thoroughly assess MRO performance based on numerous unimodal and multimodal benchmark functions as well as engineering problem instances. Moreover, to further investigate on the performance of the proposed MRO algorithm, we conduct a useful statistical evaluation and comparison with well known meta-heuristic algorithms. The experimental results confirm the high performance of MRO in dealing with complex optimization problems by discovering solutions with better quality.

Constraint Satisfaction Problems are regarded as NP-Complete problems which solving them with systematic methods requires exponential time. Firefly algorithm is a nature inspired algorithm which has been successfully applied to different combinatorial problems. This paper presents a new Discrete Firefly Algorithm for Solving Constraint Satisfaction problems (CSPs) and investigates its applicability for dealing with such problems. Performance of the proposed method has been assessed through extensive experiments on CSP instances generated by Model RB which is a standard mean for generating CSPs with different tightness. Results of the experiments in comparison with other methods including classical methods and other metaheuristic methods clearly demonstrate the significant performance of proposed discrete firefly algorithm in dealing with CSPs.

Online auctions created a very attractive environment for dishonest moneymakers who can commit different types of fraud. Shill Bidding (SB) is the most predominant auction fraud and also the most difficult to detect because of its similarity to usual bidding behavior. Based on a newly produced SB dataset, in this study, we devise a fraud classification model that is able to efficiently differentiate between honest and malicious bidders. First, we label the SB data by combining a hierarchical clustering technique and a semi-automated labeling approach. To solve the imbalanced learning problem, we apply several advanced data sampling methods and compare their performance using the SVM model. As a result, we develop an optimal SB classifier that exhibits very satisfactory detection and low misclassification rates.

Key Performance Indicators (KPIs) are used to inspect the performance and progress of businesses. This study introduces a new, integrated approach to manage KPIs in the context of decentralized information efficiently and to address the visual and managerial gaps existing in companies. The proposed Business Indicator Management (BIM) system is essential for any businesses to meet their needs in terms of information availability and agility as well as time efficiency and quality of the decision-making task. Thanks to BIM, executives are now able to obtain real-time information and analysis of the actual situation of their businesses, thus increasing their productivity. Today, no companies have yet this type of managing KPIs. Based on a detailed case study with a big-scale corporation, we thoroughly assess the effectiveness of BIM according to the system usability, data agility and decision making efficiency.

Exploration and exploitation are two strategies used to search the problem space in Evolutionary Algorithms (EAs). To significantly increase the performance of these optimization techniques in terms of the solution optimality is to strike the right balance between exploration and exploitation. Firefly is one of the most favored EAs. In this study, we introduce an entire fuzzy system to tune dynamically the firefly parameters in order to keep the exploration and exploitation in balance in each of the searching steps. A serious concern of EAs is to be stuck in local optimum solutions. The proposed fuzzy controller helps the firefly algorithm to converge to the optimal solution and escape from local optimums. To evaluate the efficiency of the fuzzy-based firefly algorithm, we conduct experiments on a set of high dimensional benchmark functions. The goal here is to compare the new firefly method with the standard firefly and well-known nature-inspired optimization algorithms. The results of the experiments show the superiority of the proposed Fuzzy firefly algorithm over the standard one.

This study introduces a new type of Combinatorial Reverse Auction (CRA), products with multi-units, multi-attributes and multi-objectives, which are subject to buyer and seller constraints. In this advanced CRA, buyers may maximize some attributes and minimize some others. To address the Winner Determination (WD) problem in the presence of multiple conflicting objectives, we propose an optimization approach based on genetic algorithms. To improve the quality of the winning solution, we incorporate our own variants of the diversity and elitism strategies. We illustrate the WD process based on a real case study. Afterwards, we validate the proposed approach through artificial datasets by generating large instances of our multi-objective CRA problem. The experimental results demonstrate on one hand the performance of our WD method in terms of three quality metrics, and on the other hand, its significant superiority to well-known heuristic and exact WD techniques that have been defined for simpler CRAs.

Simulation of complex Lotos specifications is not always efficient due to the space explosion problem of their corresponding transition systems. To overcome this difficulty in practice, we present in this paper a novel approach which integrates constraint propagation techniques into the Lotos specifications. These solving techniques are used to reduce the size of the search space before and during the search for a solution to a given combinatorial problem under constraints. In order to do that, we first tackle the challenging task of describing combinatorial problems in Lotos using the Constraint Satisfaction Problem (CSP) framework. In this regard, we provide two generic Lotos templates for describing CSPs and temporal CSPs (CSPs involving temporal constraints). To evaluate the time performance of the framework we propose, we have conducted several experimental tests on instances of the N-Queens, the machine scheduling and randomly generated CSPs. The results of these experiments are promising and demonstrate the efficiency of Lotos simulation when CSP techniques are integrated.

This paper presents a tool, namely Multi-Language Information Searching (MLIS), for a meaning-oriented search. MLIS allows users to access the right information and provides the search results with knowledge from different cultures and languages. MLIS takes advantage of agent technology as well as web services to enhance the quality of existing search engines in terms of accessibility, usability and flexibility. Their are several advantages in using MLIS, including: MLIS simultaneously performs in different languages translation and search activities as background processes hidden from users; it provides a friendly graphical user interface that instantly displays the search results in separated tabs, categorizing them according to languages chosen by users; it is proposed with a flexible architecture to automatically create multiple web-service agents based on users' preferences; it is service-independent and can integrate, for the same session, several web services from different service providers.

There is an increasing need for high-quality software components. Reusable components and formal specifications are two complementary and promising approaches to achieve this goal. One method for enhancing the reusability of existing components is generalization that creates generic components by parameterizing specific ones. Generalization and instantiation are two methods related respectively to the development for reuse and development with reuse. Generalization, that is the abstraction of existing components, identifies commonalities across a class of entities, while instantiation customizes the general properties under different circumstances. In this paper, we present several generalization and instantiation algorithms for algebraic specifications. A major difficulty during the generalization process is determining the appropriate level of generality. Highly specific components have little chance of being reused. Meanwhile, if a component is too general, its reuse might also be hard. Therefore, we introduce a novel method based on the categorized constructors to control the level of abstraction in generic components with the goal of producing effective reusable components. Through a medium-scale example, the generalization and instantiation operations are illustrated in detail.

Lotos is the ISO formal specification language for describing and verifying concurrent and distributed systems. The simulation or execution of complex Lotos specifications is, however, not always efficient due to the space explosion problem of their corresponding transition systems. To overcome this difficulty in practice, we propose in this paper the integration of constraint propagation techniques into the Lotos simulation. Indeed, constraint propagation techniques are very powerful for solving hard discrete combinatorial problems. Experimental tests, we have conducted on the simulation of several specified combinatorial problems, demonstrate the efficiency of integrating constraint propagation into Lotos simulation

Agent interaction protocols (AIP) design is one of the principal issues for building multi-agent systems. Indeed, the construction of AIP should integrate theories, methodologies and tools. We propose in this paper a unifying framework that provides a generic agent architecture to be reused as well as a methodology to construct and refine AIP specifications in an incremental way. This framework is based on the highly expressive formal language Lotos and its related technologies, such as finite state machines and temporal logics. Hence, the proposed framework also facilitates formal validation and verification of AIP specifications using rigorous tools. We argue that there are three layers of semantics of Lotos specifications that can improve Lotos expressivity in describing agent interaction. Therefore, this framework can describe almost all aspects of agent interaction and at different abstraction levels. In addition, we demonstrate how to generate an online auction protocol from the generic framework, and how to validate and verify this protocol.