Machine learning systems deployed in dynamic environments frequently operate under non-stationary data distributions, where controlled distribution shift can progressively degrade predictive performance. However, many widely used tabular benchmark datasets lack explicit temporal structure, limiting reproducible evaluation of drift adaptation methods. This work proposes a cluster-induced distribution shift simulation framework that transforms static tabular datasets into controlled evolving data streams through structured perturbations across feature-space partitions. Using this framework, six adaptation strategies are systematically evaluated: static learning, sliding-window retraining, global ADWIN retraining, cluster-local ADWIN retraining, random subspace drift detection, and feature-partitioned drift detection. Experiments are conducted on five benchmark datasets covering both classification and regression tasks using diverse predictive model families, including linear models, k-Nearest Neighbours, tree ensembles, boosting methods, and adaptive online learners. The results show that sliding-window retraining often achieves the strongest predictive performance but incurs substantial computational and labelling cost due to frequent global updates. Global ADWIN retraining reduces update frequency while preserving competitive accuracy. In contrast, the proposed cluster-local adaptation strategy consistently achieves a stronger balance between predictive performance and computational efficiency across both classification and regression settings. Under structured distribution-shift scenarios, the method reduces retraining effort by up to 75% while maintaining competitive predictive performance relative to continuously adaptive baselines. For nonlinear regression models, cluster-local adaptation preserves competitive R2 performance while substantially lowering update frequency and training effort. The proposed framework provides a reproducible benchmark for evaluating adaptation under controlled distribution shift on static tabular datasets and demonstrates that cluster-aware drift/shift detection constitutes an effective and computationally efficient alternative to uniform retraining strategies under heterogeneous distributional shift conditions.

The rapid spread of multimodal fake news (FN) on Online Social Networks (OSNs) threatens digital information ecosystems, particularly in low-resource languages. Existing multimodal fake news detection (FND) methods are largely limited to high-resource settings, restricting their global applicability. We propose an M&M-RAG, a Multilingual & Multimodal Retrieval-Augmented Generation framework, that leverages Large Vision-Language Models (LVLMs) and Large Language Models (LLMs) to verify news claims across English, Chinese and Urdu. M&M-RAG integrates real-time multilingual evidence retrieval, language-aware prompting, and cross-modal reasoning for fact verification. We further propose Multi-Ax-to-Grind Urdu, the first large-scale, multi-domain multimodal benchmark for FND in Urdu. Experiments on typologically diverse monolingual multimodal datasets demonstrate that M&M-RAG achieves state-of-the-art (SOTA) performance, with 94.6% accuracy and 94.2% F1 score, surpassing models such as SpotFake, MPFN, MMCFND, and Semi-FND. The proposed framework remains robust in zero-shot and cross-lingual scenarios under frozen-model inference without task-specific fine-tuning. The results underscore the scalability and interpretability of LVLM-based approaches for combating multimodal misinformation, particularly in under-represented and typologically diverse languages.

Abstract Android devices have evolved to offer a diverse array of services, spanning applications related to banking, business, health, and entertainment. The widespread adoption of Android devices, coupled with the open-source architecture of the Android operating system, has rendered them a prime target for malicious actors. Among the most perilous threats are Android botnets, which enable malicious actors, often referred to as botmasters, to exert remote control for the execution of destructive attacks. Android botnets have huge potential to be an emerging threat to mobile device security. In this paper, we focus on detecting evolving Android botnets and introduce a new dataset of 3458 apps, represented by 455 permission-based features. We propose an improved multilayer perceptron neural network for zero-day botnet detection. Our methodology, in this way, achieves an accuracy of 98.5%, thus outperforming traditional classifiers. It has a lot of functionality and is based on the neural network approach, making it able to identify slight botnet behaviours in order to improve Android security.

Abstract Android malware has become an increasingly important threat to individuals, organizations, and society, posing significant risks to data security, privacy, and infrastructure. As malware evolves in sophistication and complexity, the detection and mitigation of these malicious software instances have become more challenging and time consuming since the required number of features to identify potential malware can be very high. To address this issue, we have developed an effective feature selection methodology for malware detection in Android. The critical concern in the field of malware detection is the complexity of algorithms and the use of features that are used to detect malware. The present paper delivers a methodology for pre-processing datasets to select the most optimal features that will allow detecting malware, while maintaining very high accuracy. The proposed methodology has been tested on two real world datasets and the results indicate that the number of features is significantly reduced from 489 to between 19 and 28 for the first dataset and from 9503 to between 9 and 27 for the second dataset, whilst the accuracy is maintained as if all features were used.

The existing solutions related to local differential privacy (LDP) in multi-layer networks for edge computing scenarios present several limitations in both key-value data heavy hitter identification and related frequency and mean estimation tasks. First, existing LDP approaches cannot effectively use edge nodes to improve their utility/performance. Secondly, there are many network transmission tasks in edge computing, which have relatively high requirements for communication and storage costs. Furthermore, the traditional privacy budget allocation cannot attain the best utilization. To solve the above problems, we propose MLPKV, a local differential multi-layer private key-value data collection scheme for edge computing, structured into three phases: dimensional reduction, padding-length estimation, and estimation. An improved EC-OLH algorithm is used to offload the computing efforts related to aggregation and estimation to edge nodes for achieving greater efficiency. In the dimensional reduction phase, a candidate set is generated to prune the domain of original data, which improves the estimation. In addition, our method groups users for completing the tasks in each phase to avoid additional errors caused by dividing the privacy budget, and proposes a new user division with an optimal grouping ratio. Finally, the proposed method was implemented in a proof-of-concept prototype system. We compare MLPKV with baseline methods such as PrivKV and PCKV. Experimental results on both synthetic and real-world datasets show that our method achieves better utility for heavy hitter identification, frequency, and mean estimations than other state-of-the-art mechanisms. For small data sets, our approach also provides high-accuracy estimation with a low privacy budget.

Abstract It is well-known that numerical weather prediction (NWP) models require considerable computer power to solve complex mathematical equations to obtain a forecast based on current weather conditions. In this article, we propose a novel lightweight data-driven weather forecasting model by exploring temporal modelling approaches of long short-term memory (LSTM) and temporal convolutional networks (TCN) and compare its performance with the existing classical machine learning approaches, statistical forecasting approaches, and a dynamic ensemble method, as well as the well-established weather research and forecasting (WRF) NWP model. More specifically Standard Regression (SR), Support Vector Regression (SVR), and Random Forest (RF) are implemented as the classical machine learning approaches, and Autoregressive Integrated Moving Average (ARIMA), Vector Auto Regression (VAR), and Vector Error Correction Model (VECM) are implemented as the statistical forecasting approaches. Furthermore, Arbitrage of Forecasting Expert (AFE) is implemented as the dynamic ensemble method in this article. Weather information is captured by time-series data and thus, we explore the state-of-art LSTM and TCN models, which is a specialised form of neural network for weather prediction. The proposed deep model consists of a number of layers that use surface weather parameters over a given period of time for weather forecasting. The proposed deep learning networks with LSTM and TCN layers are assessed in two different regressions, namely multi-input multi-output and multi-input single-output. Our experiment shows that the proposed lightweight model produces better results compared to the well-known and complex WRF model, demonstrating its potential for efficient and accurate weather forecasting up to 12 h.

Internet of Things (IoT) refers to the complex systems generated by the interconnections among widely available objects. Such interactions generate large networks, whose complexity needs to be addressed to provide suitable computationally efficient approaches. In this article, we propose a distributed local community detection algorithm based on specific properties of community center expansions (DLCD-CCE) for large-scale complex networks. The algorithm is evaluated via a prototype system, based on Spark, to verify its accuracy and scalability. The results demonstrate that compared to the typical local community detection algorithms, DLCD-CCE has better accuracy, stability, and scalability, and effectively overcomes the problem that existing algorithms are sensitive to the location of initial seeds.

Abstract Non-predictive or inaccurate weather forecasting can severely impact the community of users such as farmers. Numerical weather prediction models run in major weather forecasting centers with several supercomputers to solve simultaneous complex nonlinear mathematical equations. Such models provide the medium-range weather forecasts, i.e., every 6 h up to 18 h with grid length of 10–20 km. However, farmers often depend on more detailed short-to medium-range forecasts with higher-resolution regional forecasting models. Therefore, this research aims to address this by developing and evaluating a lightweight and novel weather forecasting system, which consists of one or more local weather stations and state-of-the-art machine learning techniques for weather forecasting using time-series data from these weather stations. To this end, the system explores the state-of-the-art temporal convolutional network (TCN) and long short-term memory (LSTM) networks. Our experimental results show that the proposed model using TCN produces better forecasting compared to the LSTM and other classic machine learning approaches. The proposed model can be used as an efficient localized weather forecasting tool for the community of users, and it could be run on a stand-alone personal computer.

Abstract The dynamic travelling salesman problem (DTSP) is a natural extension of the standard travelling salesman problem, and it has attracted significant interest in recent years due to is practical applications. In this article, we propose an efficient solution for DTSP, based on a genetic algorithm (GA), and on the one-by-one revision of two sides (GORTS). More specifically, GORTS combines the global search ability of GA with the fast convergence feature of the method of one-by-one revision of two sides, in order to find the optimal solution in a short time. An experimental platform was designed to evaluate the performance of GORTS with TSPLIB. The experimental results show that the efficiency of GORTS compares favourably against other popular heuristic algorithms for DTSP. In particular, a prototype logistics system based on GORTS for a supermarket with an online map was designed and implemented. It was shown that this can provide optimised goods distribution routes for delivery staff, while considering real-time traffic information.

Industry 4.0 has drawn considerable attention from industry and academic research communities. The recent advances in Internet of Things (IoT), Big Data analytics, sensor technology, and artificial intelligence have led to the design and implementation of novel approaches to take full advantage of data-driven solutions applicable to Industry 4.0. With the availability of large datasets, it has become crucially important to identify the appropriate amount of relevant information, which would optimize the overall analysis of the corresponding systems. In this article, specific properties of dynamically evolving data systems are introduced and investigated, which provide framework to assess the appropriate amount of representative information.

Combinatorial optimization problems are typically NP-hard, due to their intrinsic complexity. In this paper, we propose a novel chaotic particle swarm optimization algorithm (CS-PSO), which combines the chaos search method with the particle swarm optimization algorithm (PSO) for solving combinatorial optimization problems. In particular, in the initialization phase, the priori knowledge of the combination optimization problem is used to optimize the initial particles. According to the properties of the combination optimization problem, suitable classification algorithms are implemented to group similar items into categories, thus reducing the number of combinations. This enables a more efficient enumeration of all combination schemes and optimize the overall approach. On the other hand, in the chaos perturbing phase, a brand-new set of rules is presented to perturb the velocities and positions of particles to satisfy the ideal global search capability and adaptability, effectively avoiding the premature convergence problem found frequently in traditional PSO algorithm. In the above two stages, we control the number of selected items in each category to ensure the diversity of the final combination scheme. The fitness function of CS-PSO introduces the concept of the personalized constraints and general constrains to get a personalized interface, which is used to solve a personalized combination optimization problem. As part of our evaluation, we define a personalized dietary recommendation system, called Friend, where CS-PSO is applied to address a healthy diet combination optimization problem. Based on Friend, we implemented a series of experiments to test the performance of CS-PSO. The experimental results show that, compared with the typical HLR-PSO, CS-PSO can recommend dietary schemes more efficiently, while obtaining the global optimum with fewer iterations, and have the better global ergodicity.

The multiple-travelling salesman problem (MTSP) is a computationally complex combinatorial optimisation problem, with several theoretical and real-world applications. However, many state-of-the-art heuristic approaches intended to specifically solve MTSP, do not obtain satisfactory solutions when considering an optimised workload balance. In this article, we propose a method specifically addressing workload balance, whilst minimising the overall travelling salesman’s distance. More specifically, we introduce the two phase heuristic algorithm (TPHA) for MTSP, which includes an improved version of the K -means algorithm by grouping the visited cities based on their locations based on specific capacity constraints. Secondly, a route planning algorithm is designed to assess the ideal route for each above sets. This is achieved via the genetic algorithm (GA), combined with the roulette wheel method with the elitist strategy in the design of the selection process. As part of the validation process, a mobile guide system for tourists based on the Baidu electronic map is discussed. In particular, the evaluation results demonstrate that TPHA achieves a better workload balance whilst minimising of the overall travelling distance, as well as a better performance in solving MTSP compared to the route planning algorithm solely based on GA.

Summary In this paper, we introduce a novel method to define semi‐automatically fuzzy partition rules to provide a powerful and accurate insight into cardiac arrhythmia. In particular, we define a text mining approach applied to a large dataset consisting of the freely available scientific papers provided by PubMed. The information extracted is then integrated with expert knowledge, as well as experimental data, to provide a robust, scalable and accurate system, which can successfully address the challenges posed by the management and assessment of big data in the medical sector. The evaluation we carried out shows an accuracy rate of 93% and interpretability of 0.646, which clearly shows that our method provides an excellent balance between accuracy and system transparency. Furthermore, this contributes substantially to the knowledge discovery and offers a powerful tool to facilitate the decision‐making process. Copyright © 2015 John Wiley & Sons, Ltd.

Networks are widely used to model a variety of complex, often multi-disciplinary, systems in which the relationships between their sub-parts play a significant role. In particular, there is extensive research on the topological properties associated with their structure as this allows the analysis of the overall behaviour of such networks. However, extracting networks from structured and unstructured data sets raises several challenges, including addressing any inconsistency present in the data, as well as the difficulty in investigating their properties especially when the topological structure is not fully determined or not explicitly defined. In this paper, we propose a novel method to address the automated identification, assessment and ranking of the most likely structure associated with networks extracted from a variety of data sets. More specifically, our approach allows to mine data to assess whether their associated networks exhibit properties comparable to well-known structures, namely scale-free, small world and random networks. The main motivation is to provide a toolbox to classify and analyse real-world networks otherwise difficult to fully assess due to their potential lack of structure. This can be used to investigate their dynamical and statistical behaviour which would potentially lead to a better understanding and prediction of the properties of the system (s) they model. Our initial validation shows the potential of our method providing relevant and accurate results.