Efficient real-time resource allocation for electric vehicle (EV) charging in Fog computing environments demands fast and intelligent decision-making under strict quality-of-service constraints. Traditional metaheuristics like genetic algorithms and differential evolution yield high-quality solutions but incur prohibitive computational costs, limiting their applicability in real-time systems. This paper introduces the Committee-Based Active Learning Surrogate-Assisted Particle Swarm Optimization (QBC-SA-PSO) framework, which combines multiple surrogate models with a Query by Committee (QBC) strategy to intelligently approximate fitness evaluations. By balancing exploration and exploitation, the framework drastically reduces the need for expensive exact simulations while maintaining near-optimal solution quality. Experimental validation on EV charging datasets demonstrates that PSO-SA-QBC converges within only 33 iterations, achieving a 66% reduction compared to traditional simulation techniques replaced with exact fitness evaluations, while preserving over 99% solution quality.

The rapid adoption of electric vehicles (EVs) raises critical challenges in managing charging stations, requiring efficient resource allocation to balance demand, optimize energy use, and maintain grid stability. This paper proposes FL-PSO, a hybrid strategy combining Federated Learning (FL) and Particle Swarm Optimization (PSO) for resource management in fog-cloud computing environments. FL-PSO is applied to EV charging stations and benchmarked against standard PSO, Genetic Algorithm (GA), and Cat Swarm Optimization (CSO). Evaluation is conducted using Quality of Service (QoS) metrics, including latency, energy consumption, and load balancing. Results show that FL-PSO achieves a fitness value of 0.1489, reduced energy consumption (0.2644 W), and improved load balancing (4.29), consistently outperforming baseline methods under both fixed weights and optimal QoS trade-offs. By leveraging the synergy between FL and PSO, FL-PSO provides a scalable and intelligent solution for distributed fog-cloud systems supporting smart EV charging infrastructures.

Sentiment analysis, a key application of natural language processing (NLP) and machine learning, enables the systematic extraction and interpretation of subjective content from large textual datasets. Using methods such as lexical analysis, supervised learning, and advanced deep learning models (e.g., BERT), organizations can classify sentiments (positive, negative, neutral), detect emotional nuances, and reveal emerging themes in public discussions. Sentiment analysis has numerous applications: companies monitor brand perception and refine marketing strategies, governments assess public reaction to policies, and financial experts predict market trends based on social sentiment. However, challenges such as linguistic ambiguity, cultural nuances, and data noise require continued advances in NLP to improve accuracy and scalability. This paper focuses on refining DistilBERT, a streamlined and faster variant of BERT, by enhancing its feature extraction capabilities. Our approach improves the model’s ability to detect subtle contextual cues, leading to more precise sentiment classification. Using deep learning techniques, we applied this optimized DistilBERT model to large-scale sentiment analysis tasks, achieving 90% accuracy. This result not only highlights the effectiveness of fine-tuned LLMs in sentiment classification but also surpasses prior models in both speed and performance when processing real-world sentiment data.

Effective and precise passage retrieval is crucial in contemporary information retrieval systems, particularly in light of the exponential expansion of digital document collections. Conventional approaches relying on lexical matching frequently struggle to capture semantic subtleties, often overlooking passages that are semantically pertinent but lexically divergent from user queries. Recent advancements in transformer-based models, such as BERT and its variants, have substantially enhanced retrieval performance by facilitating a deeper semantic comprehension. Nonetheless, the computational demands of these models present scalability challenges when applied to large datasets. In this work, we propose an innovative passage retrieval framework that seamlessly integrates lexical-semantic relevance with contextual information, markedly boosting retrieval precision and coherence. Our methodology unfolds in three stages: first, a Cross-Encoder BERT model assesses semantic relevance by analyzing fine-grained interactions between queries and passages. Second, a Bi-Encoder architecture generates initial embeddings, which are subsequently refined using a Graph Attention Network (GAT) to incorporate structural and contextual relationships among passages. Third, a dynamic reranking mechanism improves ranking by merging semantic relevance scores with contextual similarity evaluations, ensuring that retrieved passages are both contextually coherent and semantically precise. Evaluations conducted on benchmark datasets against state-of-the-art baselines reveal that our hybrid approach enhances retrieval accuracy and coherence. Furthermore, by integrating graph-based embedding refinements, our model mitigates the scalability constraints typically associated with Cross-Encoder architectures.

The paper offers a cutting-edge technique that combines deep learning and generative AI to analyze social media networks and forecast user interaction with purchases. However, there is still a big problem: the lack of profile information for inactive members is a barrier. To forecast a user’s involvement with a brand, we provide a novel method that blends deep learning models and generative AI. The user experience, engagement, recommendation systems, content discovery, and revenue potential are all enhanced by this approach. In particular, we use Pre-trained Language Models (PaLMs), Variational Autoencoders (VAEs), Generative Adversarial Networks (GANs), and a Convolutional Neural Network (CNN)-based model to categorize and forecast purchasers. We assess our algorithms using metrics such as accuracy, precision, recall, and F1-score on the Kaggle Social Network dataset, and we juxtapose our performance metrics with those of a novel technique. The results demonstrate the effectiveness of our proposed purchaser-prediction strategy.

As the demand for real-time stream processing grows in cloud and edge environments, the ever-changing nature of stream applications (SA) and the unpredictability of resource requirements make it essential to analyze, comprehend, and predict both user-submitted tasks and computing resources within the given distributed and heterogeneous infrastructures. Thus, optimizing resource utilization becomes paramount. This paper presents a novel method, ML_WPStreamCloud, leveraging stream workload analysis, stream task clustering, and resource classification to enhance SA offloading in edge and cloud environments. Aiming to offload tasks to suitable resources Efficiently, ML WPStreamCloud profiles SA workloads based on task profiling using the k-means algorithm. We evaluate the clustering results of state-of-the-art unsupervised clustering algorithms (e.g., K-means, MeanShift, Agglomerative) and find that k-means is more closely aligned with actual task labels, demonstrating superior performance. It outperforms other algorithms with higher silhouette scores (≈ 0.34), lower Davies-Bouldin index (≈ 0.95), and higher Calinski-Harabasz index (≈ 2511), respectively. The proposed method ML WPS treamCloud adeptly overcomes the challenges posed by diverse workloads and resource environments, offering a promising solution for the efficient scheduling of streaming applications. The simulation results with various performance parameters show the effectiveness of ML WPS treamCloud over baseline algorithms.

The needs of decision-makers vary greatly depending on their specialties, expertise and knowledge and therefore they do not visualize decision-making data in the same way. Everyone can have their preferences and prerequisites. The multidimensional model, made up of a set of facts analyzed according to different dimensions, is intended for a set of decision-makers and includes all the indicators and axes of analysis of a particular domain. They do not consider the preferences of each decision maker. A personalization phase of this diagram according to the preferences of these decision-makers will improve the decision-making process. In addition, the limited capacity of the devices used (smartphones, tablets, etc.) for consulting the results of OLAP queries presents another challenge for users. In this context, we propose to express the preferences in the form of visualization constraints which will be applied to the schema to represent the result in an incremental way.

The sentiment mining field (also known as opinion mining, opinion extraction, sentiment analysis (SA), sentiment extraction, and so on) has seen significant growth in academia. Researchers have experimented with a variety of approaches to automate SA and other fields in the fields of machine learning (ML), data mining, and natural language processing. Our research aims to develop a model that extracts and categorizes words from a specific text. In this study, we used TF-IDF to select 500 to 20,000 words with vector. After pre-processing and frequency-dependent word extraction, constructs are generated. Four Naive Bayes models (complement, multinomial, Bernoulli, and Gaussian) were used. The kappa scale, precision and accuracy scores, and F1 score were used to evaluate the proposed model. The Naive Bayes multinomial system produced the most accurate results, with an accuracy rate of 86.46 percent, according to the findings.

Decision makers do not evaluate all decision facts in the same way since their requirements differ greatly depending on their specialty, levels of knowledge, and other factors. Everyone has the right to have their own tastes and needs. The multidimensional model is intended for a group of decision-makers and consists of a set of data studied according to different dimensions. It contains all the indicators and analysis axes for a particular domain. They do not take into account the preferences of each decision-maker. A personalization phase of this diagram according to the preferences of these decision-makers would enrich the decision-making process. Another problem for consumers is the inability of mobile devices (smartphones, tablets, etc.) to display the results of OLAP queries. This paper presents a new approach in which preferences are expressed as visualization constraints that are applied sequentially to the schema to represent the result. This approach makes it possible to adapt the output to the profile choices while keeping in mind the constraint of visualization. Our strategy is based on the personalization of concepts and multidimensional data using a quantitative approach. This model interprets new preferences not expressed by the user based on the hierarchical structure of the data in the warehouse. We implemented our approach by considering three standard metrics, namely precision, recall, and F-measure. The results we obtained are very encouraging and confirm the feasibility of our proposed approach.

Indexing techniques are essential for efficiently storing and retrieving big data. Traditional indexing methods may not be suitable for uncertain environments due to their reliance on complete and accurate data. In such cases, probabilistic indexing techniques can be employed. These techniques assign probabilities to different values or attributes of the data, allowing for efficient retrieval even in the presence of uncertainty. Examples of probabilistic indexing techniques include inverted indexes with probabilistic weights and Bayesian networks. Handling uncertainty is a critical aspect of indexing big data in an uncertain environment. Our main objective in this paper is to propose a new method of semantic indexing of data in an uncertain environment. Basically, our proposed method is composed of two main parts: (i) a first part consists in processing the uncertain data in a textual document and, (ii) the second part consists in determining a new model of uncertain semantic indexing.We developed a series of experiments to evaluate our proposed solution by comparing it to the approaches studied in the literature. The results found showed that our solution is more efficient.

In recent years, social networks have become the most exploited sources of information, such as Facebook, Instagram, LinkedIn, and Twitter, have been considered to be the main sources of non-credible information. The presence of false information in these social networks has a very negative impact on the credibility of conversations. In this article, we propose a new approach, called CREDEEP (Conversational credibility based on deep learning). CREDEEP is based on: (i) the combination of post and user features in order to detect credible and not credible conversations; (ii) the integration of multi-dense layers to represent features more deeply and to improve the results. In order to study the performance of our approach, we have used the standard PHEME dataset. We compared our approach with the main approaches we have studied in the literature. The obtained results confirm the performance of our model in terms of precision, recall, and F1-measure.

“Uncertainty” is related to working with inaccurate data, imprecise and incomplete information, and unreliable results that can lead to irrational decisions. Several approaches to managing uncertain data on the Web have been proposed in the literature to resolve this problem. These approaches have failed to find the solution to this problem with accuracy and performance. Our study aims to propose a new probabilistic approach to manage Web information resources in an uncertain large-scale cloud environment. Our approach is based on three main steps: (1) modelling uncertain Web resources, (2) computing HTTP request model, and (3) interpretation and evaluation of uncertain Web resources in a context of classic hypertext navigation. The experimental study shows that the analysis of the execution time necessary for the composition of the services, by our approach, is negligible, compared to that of the other studied approaches. The algorithm that deals with the impact of the variation in the number of nodes, which we have proposed, has also been evaluated and checks all the possibilities in polynomial time and can adapt to many possibilities of multiplexing of values.

The rapid pace of technological progress has led to an increasing growth in the volume of digital data circulating on servers and on the web. This has contributed to the birth of the concept of Big Data. Simply put, this concept refers to the huge amount of information on the Internet; yet it also reveals the heterogeneity and complexity of such data. Therefore, analyzing these data, especially unstructured data, has become important since they can be used in many areas such as company management, health, smart city. In order to analyze these data, novel efficient tools are required as the current ones are not effective enough. This paper surveys the most frequently used tools and platforms for Big Data analysis with due emphasis on Machine Learning-based models. The results of this study provide in-depth knowledge of Big Data analytics applications related to machine learning that can contribute to the innovation and development of big data analytics platforms. Moreover, it helps to choose the right tools to ensure the best performance for designing an analytics system.

The lightweight description logic (DL-lite) represents one of the most important logic specially dedicated to applications that handle large volumes of data. Managing inconsistency issues, in order to effectively query inconsistent DL-Lite knowledge bases, is a topical issue. Since assertions (ABoxes) come from a variety of sources with varying degrees of reliability, there is confusion in hierarchical knowledge bases. As a consequence, the inclusion of new axioms is a main factor that causes inconsistency in this type of knowledge base. Often, it is too expensive to manually verify and validate all assertions. In this article, we study the problem of inconsistencies in the DL-Lite family and we propose a new algorithm to resolve the inconsistencies in prioritized knowledge bases. We carried out an experimental study to analyze and compare the results obtained by our proposed algorithm, in the framework of this work, and the main algorithms studied in the literature. The results obtained show that our algorithm is more productive than the others, compared to standard performance measures, namely precision, recall and F-measure.

Software mutation is a widely used technique of software testing that consists in generating variants of a base program by applying standard modifications to its source code. One of the main obstacles in the use of software mutations is the existence of equivalent mutants, i.e. mutants whose behavior is indistinguishable from the base program, even though their source code is distinct. Despite several decades of research, the identification of equivalent mutants remains an open problem. Rather than attempting to identify individual mutants that are equivalent to the base, we argue that it is often sufficient to estimate the number of equivalent mutants; also, we argue that the number of equivalent mutants depends on two factors that must be considered in the estimation effort, namely the base program and the mutation operators that are used; in this paper, we explore the impact of mutation operators on the number of equivalent mutants.

We conduct an experimental study of four algorithms for the Weighted Graph Matching Problem (WGMP) for directed graphs. The first algorithm is based on Umeyama's Eigen-decomposition approach and provides nearly an optimum solution by means of Hermitian matrices deduced from the adjacency matrices of pairs of directed weighted graphs. The second algorithm is based on Almohamad's Symmetric ploynomial transform approach. The Symmetric polynomial transform is applied to map input data representing polynomial roots into a set of coefficients that are invariant under permutation of the roots. Assuming this transformation, the weights of two nodes can be compared one to one via their resulted invariant coefficients. The third algorithm is inspired from Almohamad's Linear programming approach where the WGMP is initially formulated in a non linear problem and then transformed into a Linear one formulated in L1 norm. In the two first algorithms, the optimum match is given by the Hungarian method when a permutation exists between each two nodes. The fourth algorithm is an improved version of the first one that gives exact results only for graphs satisfying certain conditions.

This paper presents a Configurable Cloud Service Discovery and Selection System (C2SDS2) that aims to guide Cloud users in retrieving configurations of IaaS Cloud resources over the Internet. The C2SDS2 takes into account both user functional and non-functional requirements in the retrieval and selection process. In this work, configurable services are designed as directed Cloud extended feature graphs inspired by graph structures and feature models. The discovery-based matching is performed in two steps. In the first step, the structural matching is performed by adapting two heuristics: (1) Hungarian and (2) VG (Volgenant-Jonker) which is an improved Hungarian algorithm. In the second step, the QoS matching and ranking are achieved using three different methods of directed weighted graph matching namely the Eigen-decomposition, the Symmetric polynomial transform and the Linear programming methods. We show the efficiency and effectiveness of our system through an experimental study conducted on a configurable IaaS services collection. The experiment results show the performance and the efficiency of the algorithms combinations.

Nowadays, due to the growth of technology, there is a mass production of data (of large volume), available in a digital form. These currently available data are not unified but appear in different formats and types. The diversity of the data is based on the type of information, they contain, such as text, image, video and audio documents and also on their sources, such as data from sensors (high variety). In addition, with the expansion of the Internet and the World Wide Web, the majority of these data become the publicity available for a wide range of users (at high speed). The main objective of this work is to propose an efficient Big Uncertainty Web Data Services Indexing Model able to reasoning in uncertain data environment. More concretely, the proposed approach is based on two main phases: the first one consists on processing uncertain data in the syntactic indexing phase and the second one consists on the semantic indexing phase. These two phases are presented as two algorithms syntactic and semantic.

The multi-tenancy aware discovery of configurable Cloud services is one of the most important and difficult issues, because of multiplicity and non-standardization of their description in the Cloud. In this paper, relying on a feature model based specification of configurable WSDL services, we develop a multi-tenancy aware approach for their discovery. Our approach empowers multiple tenants to discover their desired service configured variants, considering individual variations. To do so, we reduce the problem of configurable matching to a tree matching problem and we adapt existing algorithms for this aim. The experimental results show the feasibility of our approach.

User generated content on the microblogging social network Twitter continues to grow with significant amount of information. The semantic analysis offers the opportunity to discover and model latent interests’ in the users’ publications. This article focuses on the problem of uncertainty in the use rs’ publications that has not been previously treated. It proposes a new approach for users’ interest discovery from uncertain information that augments traditional methods using possibilistic logic. The possibility theory provides a solid theoretical base for the treatment of incomplete and imprecise information and inferring the reliable expressions from a knowledge base. More precisely, this approach used the product-based possibilistic network to model knowledge base and discovering possibilistic interests. DBpedia ontology is integrated into the interests’ discovery process for selecting the significant topics. The empirical analysis and the comparison with the most known methods proves the significance of this approach.

In this article, we propose a new approach for indexing biomedical documents based on a possibilistic network that carries out partial matching between documents and biomedical vocabulary. The main contribution of our approach is to deal with the imprecision and uncertainty of the indexing task using possibility theory. We enhance estimation of the similarity between a document and a given concept using the two measures of possibility and necessity. Possibility estimates the extent to which a document is not similar to the concept. The second measure can provide confirmation that the document is similar to the concept. Our contribution also reduces the limitation of partial matching. Although the latter allows extracting from the document other variants of terms than those in dictionaries, it also generates irrelevant information. Our objective is to filter the index using the knowledge provided by the Unified Medical Language System®. Experiments were carried out on different corpora, showing encouraging results (the improvement rate is +26.37% in terms of main average precision when compared with the baseline).

Hidden Markov Models (HMM) have been widely used in classification tasks. Despite their efficiency in stochastic sequences labeling, they are overwhelmed by imperfect quality of used data in the learning and inference processes. In this paper, we try to evaluate the contribution of possibilistic theory in creating sequences of observations used by HMM models. Experimental results show that observation sequences, obtained by possibilistic reasoning significantly, improve the performance of HMM in the recognition of online e-learning activities.

In this work, we propose a new stemmer algorithm to indexing unstructured Document. It can detect the most relevant words in an unstructured document. This algorithm is based on two main modules: the first module ensures the processing of compound words and the second allows the detection of the endings of the words that have not been taken into consideration by the approaches presented in literature. The proposed algorithm allows the detection and removal of suffixes and enriches the basis of suffixes by eliminating the suffixes of compound words. We have experienced our algorithm on a standard basis of terms and the results show the remarkable effectiveness of our algorithm compared to others presented in related works.

This paper proposes a new Information Retrieval Model based on Possibilistic Networks. The model structure integrates most relevant term to term dependence relationships. The approach used to extract the set of these dependencies focuses on local dependencies between terms within each document. The relevance of a document to a query is interpreted by two degrees: the necessity and the possibility. The necessity degree evaluates the extent to which a document is relevant to a query, whereas the possibility degree evaluates the reasons of eliminating irrelevant documents. These two measures are also used for quantifying terms-terms links and terms-documents links. Experiments carried out on three standard document collections show the effectiveness of the model.

This paper proposes a new approximate model for biomedical concept extraction. This model is based on possibilistic network, statistical computing and semantic proximity. The possibilistic network is used for representing the MeSH structure in order to select the relevant concepts for a biomedical text. Moreover, we propose an enrichment model of the MeSH thesaurus by the identification of the semantic relations between concepts. The results of the extraction model serve to mapping a query in an information retrieval process. And, to prove the significance of our model in the Information Retrieval context, we used a vector model and the OHSUMED collection.