The integration of fully homomorphic encryption (FHE) in federated learning (FL) has led to significant advances in data privacy. However, during the aggregation phase, it often results in performance degradation of the aggregated model, hindering the development of robust representational generalization. In this work, we propose a novel multimodal quantum federated learning framework that utilizes quantum computing to counteract the performance drop resulting from FHE. For the first time in FL, our framework combines a multimodal quantum mixture of experts (MQMoE) model with FHE, incorporating multimodal datasets for enriched representation and task-specific learning. Our MQMoE framework enhances performance on multimodal datasets and combined genomics and brain MRI scans, especially for underrepresented categories. Our results also demonstrate that the quantum-enhanced approach mitigates the performance degradation associated with FHE and improves classification accuracy across diverse datasets, validating the potential of quantum interventions in enhancing privacy in FL.

In recent years, the development of computer vision and deep learning in general has significantly impacted various domains. As (semi-) autonomous driving and advanced driving assistance systems gain popularity, the demand for robust and accurate traffic sign classification solutions continues to rise. The task of classifying traffic signs faces a special challenge in the European Union, where cars often cross national borders and traffic signs differ from country to country. We propose a method to augment a training dataset with illustrative drawings to improve the precision of traffic sign classification models. Our results suggest that illustrative drawings improve precision. For example, the precision of Estonian traffic sign classification models increased from 0.92000 to 0.92667 when using illustrative images of Estonian traffic signs, and even using illustrative images of German traffic signs resulted in an improvement (0.92222). Similarly, the precision of German traffic sign classification models increased from 0.78889 to 0.83111 with German illustrations and 0.83556 with Estonian illustrations. Moreover, the results suggest that models trained on optimal ratio of real traffic signs and illustrative images also improve F1, recall and precision of predictions.

The advancement of the semantic web and Linked Open Data (LOD) cloud has led to the creation and integration of various knowledge bases defined by ontologies. A significant challenge within the LOD paradigm is identifying resources that refer to the same real-world object to enable large-scale data integration and sharing. In this context, instance matching has emerged as a key solution, linking co-referent instances from heterogeneous data sources using owl:sameAs links. Traditional approaches focus on schema-level matching but often fail to address property-level heterogeneity. Moreover, given the large scale of instances, examining all possible instance pairs is impractical. This paper proposes a scalable and efficient instance-matching approach using MongoDb (Humongous database) and Lucene. MongoDb stores instances at any scale and Lucene uses inverted indexes to identify matching candidates. Experiments on the instance matching track from the Ontology Alignment Evaluation Initiative (OAEI’2022) show that our approach matches the F-measure score of RE-Miner, the top performer in OAEI’2020, while surpassing all other participants in OAEI’2020, 2021 and 2022. Additionally, it operates 17 times faster than RE-Miner, four times faster than Lily and 15 times faster than LogMap, the fastest in OAEI’2020, 2021 and 2022, respectively. Moreover, we evaluate our approach on other knowledge bases from OAEI’2010. Once again, our approach gets highly competitive resuts compared to state-of-the-art approaches.

Multiple objectives have emerged in tuning protein-protein interaction (PPI) networks, such as identifying cross-species network similarities and predicting protein complexes and functions. Despite the proliferation of tuning methodologies, challenges remain in balancing accuracy and efficiency. In this paper, we introduce GA2Vec, a novel approach for globally aligning multiple PPI networks using genetic algorithms in a many-to-many fashion. GA2Vec leverages vector embeddings of protein sequences from ProtBERT, ESM-2, and ProtT5-XL-UniRef50 to reconstruct weighted PPI networks, incorporating functional similarity through Gene Ontology (GO) term embeddings derived from the Anc2vec method. We employ four community detection algorithms to generate candidate clusters from the weighted graph, serving as initial solutions for the genetic algorithm. The genetic algorithm optimizes network alignment by refining these clusters using a fitness function based on similarity scores from pre-trained embeddings and GO terms, achieving a robust global network alignment. We demonstrate the effectiveness of our method through experiments on eukaryotic, prokaryotic, SARS-CoV, and virus-host biological networks. It achieves robust alignment between SARS-CoV-2 and SARS-CoV-1 PPI networks, balancing $F1$, cluster interaction quality ($CIQ$), internal cluster quality ($ICQ$), consistent clusters, and $sensitivity$, with scores reflecting its adaptability to diverse biological contexts.

Point of Interest (POI) recommender systems (RSs) play a primary role in improving Location-based Social Networks’ user experience. This paper studies the potential usefulness of serendipity in POI recommendations. We first introduce a new POI RS, called Discovery, that attempts to improve the accuracy-serendipity trade-off. The proposed RS aims to recommend POIs that provide a pleasant surprise, allowing users to discover new venues known as serendipitous POIs. We then look closely at how serendipity affects the quality of POI suggestions by contrasting the outcomes of Discovery with those of three cutting-edge non-serendipitous POI RSs. We use two real-world datasets—Foursquare and Flickr—along with a variety of metrics to test our ideas. These include (i) accuracy, which checks the precision, recall, and f-measure of Top-N recommendations; and (ii) beyond-accuracy, which checks the categorical and geographical diversity, explainability, and coverage in terms of POIs. The reported experimental observations show that serendipity boosts POI recommendation accuracy and favors geographically proximate and explainable POIs. However, standard POI baselines outperform Discovery in terms of categorical diversity and coverage.

Missing values in datasets pose a significant challenge, often leading to biased analyses and suboptimal model performance. This study shows a way to fill in missing values using Denoising AutoEncoders (DAE), a type of artificial neural network that is known for being able to learn stable ways to represent data. The observed data are used to train the DAE, and then they are used to fill in missing values. Extensive tests on different image datasets, taking into account different mechanisms of missing data and percentages of missingness, are used to see how well this method works. The results of the experiments show that the DAE-based imputation works better than other imputation methods, especially when it comes to handling informative missingness mechanisms.

The electric vehicle (EV) market is experiencing substantial growth, and it is anticipated to play a major role as a replacement for fossil fuel-powered vehicles in transportation automation systems. Nevertheless, as a rule of thumb, EVs depend on electric charges, where appropriate usage, charging, and energy management are vital requirements. Examining the work that was done before gave us a reason and a basis for making a system that forecasts the real-time availability of electric vehicle charging stations that uses a scalable prediction engine built into a server-side software application that can be used by many people. The implementation process involved scraping data from various sources, creating datasets, and applying feature engineering to the data model. We then applied fundamental models of machine learning to the pre-processed dataset, and subsequently, we proceeded to construct and train an artificial neural network model as the prediction engine. Notably, the results of our research demonstrate that, in terms of precision, recall, and F1-scores, our approach surpasses existing solutions in the literature. These findings underscore the significance of our approach in enhancing the efficiency and usability of EVs, thereby significantly contributing to the acceleration of their adoption in the transportation sector.

Because of the spread of epidemics and diseases in many countries around the world, news consumption from online sources has substantially increased. These news stories written in Arabic are about more than one topic, which is interesting for the multi-label classification paradigm. Furthermore, the recent studies based on multi-label Arabic text classification deal with news articles, which are rather long texts. Thus, we put forward a large dataset of concise Arabic news based basically on the Corona virus, namely Hana-MC, which has been built from various news portals. We conducted a comparative study using several multi-label classification approaches, including algorithm adaptation, problem transformation, and ensemble methods. Experimental results showed that the Ensemble Method RAKELD with the Random Forest base classifier obtained the best accuracy score.

To improve the accuracy of an RNN when processing sparse and irregular multivariate clinical time series, we introduce two stacked deep learning models built on top of it, namely Padd-GRU and Alignment-driven Neural Network (ALNN). The Padd-GRU performs data-driven padding and imputation to obtain equal-length univariate and fill-in missing values, respectively. Then, the ALNN component transforms the resulting padded irregular multivariate clinical time series into a pseudo-aligned (or pseudo-regular) latent multivariate time series. We use the MIMIC-3 and PhysioNet databases to evaluate and compare our model to the state-of-the-art models on the mortality prediction task.

The electric vehicle (EV) market is experiencing substantial growth, and it is anticipated to play a major role as a replacement for fossil fuel-powered vehicles in transportation automation systems. Nevertheless, as a rule of thumb, EVs depend on electric charges, where appropriate usage, charging, and energy management are vital requirements. Examining the work done before gave us a reason and a basis for making a system that forecasts the real-time availability for EV charging stations that uses a scalable prediction engine built into a server-side software application that many people can use. The implementation process involved scraping data from various sources, creating datasets, and applying feature engineering to the data model. We then applied fundamental machine learning models to the pre-processed dataset, and subsequently, we implemented an ensemble model that combines the strengths of both Random Forest (RF) and Artificial Neural Network (ANN). This approach leverages the RF’s resilience to overfitting and its ability to handle diverse data while benefiting from the ANN’s capacity to capture complex non-linear relationships. The resulting ensemble model demonstrates significant improvements in precision, recall, and F1-score compared to individual models, making it a valuable contribution to the transportation sector.

In urban environments, bicycle-sharing emerges as an eco-friendly solution, yet inherent imbalances in bicycle rents and returns necessitate systematic rebalancing, posing a global challenge. This underscores the crucial role of forecasting in optimizing bicycle allocation across diverse docks. Despite the commendable goals of reducing carbon emissions and promoting public health, efficient rebalancing remains elusive, emphasizing the need for forecasting to enhance overall system efficiency. User demand in public bicycle-sharing systems presents a primary challenge, influenced by commuting patterns and topographical conditions, leading to critical spatial incongruities. Integration of robust demand prediction mechanisms becomes essential, strategically overcoming challenges and ensuring seamless bicycle-sharing system operation. Our solution proactively addresses disruptions by forecasting user demand and employing manual redistribution based on Long Short-Term Memory (LSTM) models. Empirical validation attests to its efficiency and accuracy, showcasing versatility. The system seamlessly integrates frameworks for forecast delivery to applications, ensuring robustness and high availability through meticulous dataset consumption.

Following the publication of the original article [1], the authors identified errors in calculating the percentages of improvement and deterioration among DTIOG variants in the Results and discussion section. The changes have been highlighted with bold typeface and are shown in Additional file 1. The original article [1] has been corrected.

Abstract In many fields, like aspect category detection (ACD) in aspect‐based sentiment analysis, it is necessary to label each instance with more than one label at the same time. This study tackles the multilabel classification problem in the ACD task for the Arabic language. For this purpose, we used Arabic hotel reviews from the SemEval‐2016 dataset, comprising 13,113 annotated tuples provided for training (10,509) and testing (2,604). To extract valuable information, we first propose specific data preprocessing. Then, we suggest using the dynamic weighted loss function and a data augmentation method to fix the problem with this dataset's imbalance. Using two possible approaches, we develop new ways to find different categories of things in a review sentence. The first is based on classifier chains using machine learning models. The second is based on transfer learning using pretrained AraBERT fine‐tuning for contextual representation. Our findings show that both approaches outperformed the related works for ACD on the Arabic SemEval‐2016. Moreover, we observed that AraBERT fine‐tuning performed much better and achieved a promising ‐score of .

The rapid evolution of digital technologies has given rise to the concept of Digital Twin, a dynamic, virtual representation of physical systems, processes, and environments. This special issue delves into the transformative potential of Digital Twins in the realm of future networks and emerging Internet of Things (IoT) applications. By integrating advanced simulation, real-time data analytics, and machine learning, Digital Twins offer unprecedented opportunities for optimizing network performance, enhancing predictive maintenance, and enabling smarter IoT solutions. The articles in this issue explore a variety of topics, including the development and implementation of Digital Twins for next-generation communication networks, the role of artificial intelligence in enhancing the fidelity and utility of Digital Twins, and the application of these technologies in diverse IoT domains such as smart cities, healthcare, industrial automation, and environmental monitoring. Emphasis is placed on innovative methodologies, case studies, and experimental results that highlight the practical benefits and challenges associated with deploying Digital Twins in real-world scenarios. Through this special issue, we aim to provide a comprehensive overview of the current state of research and development in Digital Twins, underscore the technological advancements driving their adoption, and discuss future directions and open research questions. This collection of works serves as a valuable resource for researchers, practitioners, and policymakers interested in harnessing the power of Digital Twins to revolutionize network infrastructures and IoT ecosystems.

This paper introduces the Hyper-5G research project, which aims at developing and evaluating an experimental proof of concept of a cross-Atlantic Network Digital Twin for the future wireless mobile 5G and beyond (B5G). Hyper-5G project brings innovative capabilities to allow distributed twins to replicate the 5G IoT network infrastructure digitally. Hyper-5G project interconnects two geographically distributed edge-cloud infrastructures, i.e., Grid5000 in Europe and Chameleon cloud in the US, to assess the feasibility of deploying new 5G IoT services using the twin. Hyper-5G offers an open European platform to experiment with different IoT scenarios, ranging from smart agriculture to healthcare, connected cars, etc. In the USA, Hyper-5G deploys the DT Hub inside the Chameleon cloud, connected to CHI-Edge IoT testbed, to enable emulating real-world IoT scenarios such as connected robots, smart cities, and smart grids, etc.

Posting sarcastic comments on hotel opinion websites became a common trend. Therefore, sarcasm detection becomes of primary importance, mainly since sarcasm can flip the review's polarity. To our knowledge, no work related to detecting sarcasm exists in the hospitality industry. The lack of a labeled dataset is an utmost issue for researchers. We propose a new approach based on the domain adaptation using the self-training technique for the semi-supervised learning of sarcastic Arabic hotel reviews. We use AraBERT for contextual embedding and machine learning models for self-training classification. The obtained accuracy reaches 91.1%.

Asymptotically consistent estimators of a treatment effect under many potential confounders became possible with the latest advancements in doubly-robust causal inference models (e.g., Double ML). In this study, we propose SAFE-TH framework to estimate and explain the heterogeneous treatment effect with partial dependence plots and report it under a reduced hypothesis space of interest. We analyze a shift in accessibility to credit for small to medium enterprises (SMEs) during the first months of the COVID-19 pandemic. Utilizing the proposed framework can improve the interpretability of CATE models by identifying and providing confidence intervals for regions of heterogeneity.

BACKGROUND: The pharmaceutical field faces a significant challenge in validating drug target interactions (DTIs) due to the time and cost involved, leading to only a fraction being experimentally verified. To expedite drug discovery, accurate computational methods are essential for predicting potential interactions. Recently, machine learning techniques, particularly graph-based methods, have gained prominence. These methods utilize networks of drugs and targets, employing knowledge graph embedding (KGE) to represent structured information from knowledge graphs in a continuous vector space. This phenomenon highlights the growing inclination to utilize graph topologies as a means to improve the precision of predicting DTIs, hence addressing the pressing requirement for effective computational methodologies in the field of drug discovery. RESULTS: The present study presents a novel approach called DTIOG for the prediction of DTIs. The methodology employed in this study involves the utilization of a KGE strategy, together with the incorporation of contextual information obtained from protein sequences. More specifically, the study makes use of Protein Bidirectional Encoder Representations from Transformers (ProtBERT) for this purpose. DTIOG utilizes a two-step process to compute embedding vectors using KGE techniques. Additionally, it employs ProtBERT to determine target-target similarity. Different similarity measures, such as Cosine similarity or Euclidean distance, are utilized in the prediction procedure. In addition to the contextual embedding, the proposed unique approach incorporates local representations obtained from the Simplified Molecular Input Line Entry Specification (SMILES) of drugs and the amino acid sequences of protein targets. CONCLUSIONS: The effectiveness of the proposed approach was assessed through extensive experimentation on datasets pertaining to Enzymes, Ion Channels, and G-protein-coupled Receptors. The remarkable efficacy of DTIOG was showcased through the utilization of diverse similarity measures in order to calculate the similarities between drugs and targets. The combination of these factors, along with the incorporation of various classifiers, enabled the model to outperform existing algorithms in its ability to predict DTIs. The consistent observation of this advantage across all datasets underlines the robustness and accuracy of DTIOG in the domain of DTIs. Additionally, our case study suggests that the DTIOG can serve as a valuable tool for discovering new DTIs.

In the last few decades, Deep Neural Networks (DNNs) has shown outstanding performance in speech recognition applications. We demonstrate that the improved accuracy obtained by Deep Convolutional Neural Network (DCNN) arose from their capacity to extract discriminative representations which are robust to various sources of variability in speech signals. By this study, we propose a new algorithm, named Scattering Transform-Deep Convolutional Neural Network CNN: ST-DCNN to identify normal and pathological voices. The effectiveness of advances in speech features have been proven to be the root for an efficient pathological voices classification. The proposed algorithm involved two stages: First, scatter wavelet features are extracted. Then, DCNN is used to classify the voices samples. We evaluated the robustness of the proposed system in silent environments. The experimental results indicates that it achieves better performance with scattering wavelet and DCNN with the clean data within 99.62 % of recognition rate.

This study investigated the effectiveness of using the Kahoot! game in developing the cognitive achievement and direction of students of pharmacy at Alasmarya Islamic University, Libya. The study design is based on action research. Kahoot! was implemented once at the end of each of three units. The study sample consisted of 30 female students from the first year of university at the Pharmacy Science College in Libya. The students were selected randomly, and studied using Kahoot! technology. For the quantitative part of the study, data were collected through a 20-item questionnaire on 20 participants and 15 participants to gather information on students’ perceptions about this application by the interview. Results indicate that the students were able to engage actively in the chemistry lessons and learned the unit on molecular weights effectively, leading them to enjoy basic chemistry, and the surveys allow for anonymous classroom participation, which further engages all students. The results hold implications for the development of more efficient, effective in educational process.

This research proposal explores how citizen-centered learning and career advancement can benefit from artificial intelligence, occupational classification frameworks, and the concept of proactive services. In the current literature, there are a lot of machine learning methods used in various job and training recommendation systems to tackle scientific or real-life problems. However, only a few use the existing occupational classifications to classify job and training advertisements or enable cross-regional labor mobility. Additionally, the quality of public e-services regarding labour market services in each European country varies greatly. For example, even though Estonia is typically referred to as being at the forefront of public service digitization and automation, it has not implemented machine learning methods to match job offers or training with candidates. The matching process between vacancy and job seeker is currently carried out with outdated International Standard Classification of Occupations (ISCO) codes, altered to the Estonian Unemployment Insurance Fund's needs. The ISCO code is assigned to each job vacancy by a company and job wish by a citizen manually. Such functionality is rigid and requires the users to define an accurate ISCO code. Even when the filled-in CV consists of detailed previous work experience and educational background, if the ISCO code is not accurate, the e-service will not help the citizen find a new job. Moreover, in today's public employment service portals, there is typically no option to insert specific skills that a citizen has to receive an increased number of accurate job or training recommendations. Consequently, despite many well-established frameworks dealing with competencies and occupations, citizen-centered public services supporting upskilling and finding a new job are inefficient. In this paper, we put forth a research roadmap for investigating how to enable a technical ecosystem using occupational classification frameworks so that citizens, both employed and unemployed, can receive proactive recommendations about upcoming training events and job vacancies. Such a system should be tailored to support citizen life events. For example, it could consider citizens’ previous work and educational background to help with retraining, upskilling, or changing one's career path. Therefore we have initiated a project in collaboration with the Estonian Unemployment Insurance Fund, the Estonian Qualifications Authority, other Estonian public organizations, and partner universities in Latvia and Finland. The research is planned as an action design research to design an artificial intelligence-enabled Virtual Competence Assistant (VCA) for the EU labour market.

Today’s common consensus is that the world’s most valuable resource is no longer oil but data. But, like oil, data is a source of pollution, mainly caused by the processing of extensive amounts of data. Thus, providers of data storage and processing solutions are at the heart of the debate on green computing. These solutions shall satisfy at the same time two conflictual non-functional requirements (NFRs): (i) the performance of analytical queries and (ii) the reduction of the energy consumption. These NFRs are strongly connected to query processors. Contrary to the first NFR, which has been widely studied by academia and industry, the second one does not get the same attention. The current works dealing with query processors’ energy efficiency (EE) are mainly focused on logical optimizations of database operations. However, nobody can deny that the satisfaction of the first NFR passes necessarily through physical optimizations such as indexes. Based on this discussion, we highly recommend the usage of green physical optimizations by existing and ongoing query processors. To promote and defend our vision of a green World, we propose in this paper to study the problem of selecting Bitmap Join Indexes that balance our NFRs. Because of its hardness, we first introduce a pruning strategy that eliminates non-relevant indexable attributes. Second, an approach for selecting indexes includes a hypergraph structure to manage the large search space of our problem and a Skyline operator to find the compromise between these NFRs. Third, we conduct intensive experiments to assess the impact of our proposal on our studied NFRs.

Travel time estimation is an important dynamic measure in developing mobility on the road navigation services of Intelligent Transportation System (ITS). The key challenge is how to accurately assess the time required for a given path that is extensively varied and affected by a wealthy number of spatial, temporal, and road conditions factors. However, former works have focused on capturing the local trajectory patterns for reducing the model's accuracy. In this paper, we introduce a novel approach called Zone Embedding and Deep Neural Network-based Travel Time Estimation Approach (ZED-TTE). The main originality of the latter is that it summarizes the road network into several meaningful zones for extracting global spatial correlations and temporal dependencies. Thus, it has a better overview of the global picture to efficiently gauge the travel time for the full path, by directly providing a source and a destination without intermediate trajectory points involving some road external conditions. Experiments carried out on two large-scale real-world taxi trips datasets show that the proposed approach sharply outperforms the state-of-the-art models.

Formal Concept Analysis (FCA) is a popular method for knowledge discovery and data mining in binary data. A shortcoming of FCA is the huge number of formal concepts that may be drawn from formal contexts (binary data tables) of moderate size or larger. A strategy to deal with this shortcoming is to extract a subset of formal concepts that cover the context either fully or partially. We compare different methods for generating full and partial concept cover, present a simple Greedy Coverage (GC) method, and show that it is an efficient option, especially for generating partial concept cover.

Even though several systems for Air Quality (AQ) monitoring have been in existence for over a decade, a research model for Knowledge Management (KM) of AQ data has to be created in order to enhance the decision-making and organize the air quality data collected from the Internet of Things (IoT) consumer devices. This model should be made more performant by ensuring greater flexibility and interoperability between devices and emerging technologies. In this context, we propose an approach for representing Data WareHouse (DWH) schema based on an ontology that captures the multidimensional knowledge of tools, techniques, and technologies used for novel AQ systems. This enhances decision-making by coping with potential problems such as data sources heterogeneity and covering the various phases of the decision-making life cycle.

Summary Roadside units (RSUs) have a crucial role in maintaining vehicular ad hoc networks (VANETs) connectivity and coverage, especially, for applications gathering or disseminating nonsafety information. In big cities with complex road network topology, a huge number of costly RSUs must be deployed to collect data gathered by all moving vehicles. In this respect, several research works focusing on RSUs deployment have been proposed. The thriving challenge would be to (1) reduce the deployment cost by minimizing as far as possible the number of used RSUs and (2) to maximize the coverage ratio. In this paper, we introduce a spatiotemporal RSU deployment framework including three methods, namely, SPaCov, SPaCov+, and HeSPic. SPaCov starts by mining frequent mobility patterns of moving vehicles from their trajectories; then, it computes the best RSU locations that cover the extracted patterns. Nonetheless, SPaCov+ extracts the frequent mobility patterns as well as the rare ones to enhance the coverage ratio. HeSPiC is a budget‐constrained spatiotemporal coverage method that aims to maximize the coverage ratio subject to a budget constraint, which is defined in terms of RSU number. Performed simulations highlight the efficiency and the effectiveness of the proposed RSU deployment framework in terms of coverage ratio, deployment cost, network latency and overhead.

Breast-cancer (BC) is the most common invasive cancer in women, with considerable death. Given that, BC is classified as a hormone-dependent cancer, when it collides with pregnancy, different questions may arise for which there are still no convincing answers. To deal with this issue, two new frameworks are proposed within this paper: CoRaM and Dist-CoRaM. The former is the first unified framework dedicated to the extraction of a generic basis of Correlated-Rare Association rules from gene expression data. The proposed approach has been successfully applied on a breast-cancer Gene Expression Matrix (GSE1379) with very promising results. The latter, the Dist-CoRaM approach, is a big-data processing based on Apache spark framework, dealing with correlation mining from micro-array pregnancy associated breast-cancer assays (PABC) data. It is successfully applied on the (GSE31192) gene expression matrix (GEM). The correlated patterns of gene-sets shed light on the fact that PABC exhibits heightened aggressiveness compared to cancers for Non-PABC women. Our findings suggest that higher levels of estrogen and progesterone hormones, unfortunately, are very keen to the increase of the tumor aggressiveness and the proliferation of the cancer.

In big cities, with heavy traffic, tracking vehicle movements undoubtedly generates big data, which can be cumbersome for real time analysis. In the other hand, the process of roadside unit (RSUs) deployment in Vehicular Ad-hoc Networks (VANETs) mainly relies on the analysis of vehicles trajectories, which is considered as big sequential data. The complexity comes from (i) the number of transactions recorded at each time and (ii) the redundancy of transactions in the database. In this respect, several research works focusing on RSUs deployment have been proposed. The thriving challenge would be to (i) reduce the deployment cost by minimizing as far as possible the number of used RSUs; and (ii) to maximize the coverage ratio. In this paper, we introduce a spatio-temporal RSU deployment method consisting of two main parts namely ”finding representative transactions from the sequential database” and ”computing coverage”. The first one identifies the representative transactions, a.k.a transactions having a high utility, from a large sequential database. The second part computes a minimal set of junctions, where RSUs have to be ideally placed. The performed experiments show that our scheme outperforms its competitors in terms of cost, time performance as well as coverage ratio.

The growing demand and the diverse traffic patterns coming from various heterogeneous Internet of Things (IoT) systems place an increasing strain on the IoT infrastructure at the network edge. Different edge resources (e.g. servers, routers, controllers, gateways) may illustrate different execution times and energy consumption for the same task. They should be capable of achieving high levels of performance to cope with the variability of task handling. However, edge nodes are often faced with issues to perform optimal resource distribution and energy-awareness policies in a way that makes effective run-time trade-offs to balance response time constraints, model fidelity, inference accuracy, and task schedulability. To address these challenging issues, in this paper we present a SDN-based dynamic task scheduling and resource management Deep Reinforcement Learning (DRL) approach for IoT traffic scheduling at the network edge. First, we introduce the architectural design of our solution, with the specific objective of achieving high network performance. We formulate a task assignment and scheduling problem that strives to minimize the network latency while ensuring energy efficiency. The evaluation of our approach offers better results compared against both deterministic and random task scheduling approaches, and shows significant performances in terms of latency and energy consumption.

Scene text detection is one of the raising aspects of Information and Communications Technology (ICT) field used by individuals in our daily life. Therefore, textual information detection aim to determinate text line coordinates in two fields: printed documents and real-world scenes images. One of the key points to the success of detecting text on both accuracy and precision values for printed documents is deep learning. However, real-world scenes images still face Challenges from recognizing text from the remaining shapes. In this paper, we propose a deep Fully Convolutional Networks (FCN) multi-oriented system to real-time localized text through an end-to-end trainable single network. For training and evaluating stages, we have used the freely available Arabic-Text-in-Video (AcTiV) dataset.

The accuracy and relevance of Business Intelligence & Analytics (BI&A) rely on the ability to bring high data quality to the data warehouse from both internal and external sources using the ETL process. The latter is complex and time-consuming as it manages data with heterogeneous content and diverse quality problems. Ensuring data quality requires tracking quality defects along the ETL process. In this paper, we present the main ETL quality characteristics. We provide an overview of the existing ETL process data quality approaches. We also present a comparative study of some commercial ETL tools to show how much these tools consider data quality dimensions. To illustrate our study, we carry out experiments using an ETL dedicated solution (Talend Data Integration) and a data quality dedicated solution (Talend Data Quality). Based on our study, we identify and discuss quality challenges to be addressed in our future research.

In order to enhance the business process, the Bid Process Information System (BPIS) should be made more performant by ensuring greater flexibility and interoperability between devices. Moreover, the specication of this system has to deal with “three fit” problems. To this end, four dimensions have been identified in order to cope with potential failures: operational, organizational, decision-making, and cooperative dimensions. In this paper, we focus on the decision-making dimension of the BPIS and propose an approach for representing data warehouse schema based on an ontology that captures the multidimensional bid-knowledge.

This research paper starts from the belief that digital government is a systemic process and that by studying the technical architecture it is possible to gain a better understanding of its behavior. Drawing on insights from system dynamics, systems theory, and complex adaptive systems, the paper hypothesizes that by including and reinforcing feedback loops in the architecture, it is possible to obtain significant performance gains. For the sake of testing the hypothesis, and to answer the research question: "How can Systems Theory be used to understand the digital government infrastructure of a country?" the case of Estonia is presented. As a result of the research, it is found that the presence of reinforcement loops in the technical architecture leads to performance gains, that it is possible to observe complex behavior in digital government infrastructure, and demonstrates that systems thinking may be used to analyze digital government systems.

Binary data arise naturally in many fields including shopping carts, pass-fail tests, social networks etc. Descriptive data mining aims to discover a concise set of general patterns in these possibly noisy data. An important tool for describing binary data is Formal Concept Analysis (FCA) which describes the data through formal concepts. As the full lattice of formal concepts can become large even when dealing with relatively modest amounts of data there are several methods to reduce the number of concepts used to describe the data: selecting a subset of "interesting" concepts, finding a subset of concepts that cover the data fully etc. In this paper we apply a novel method of concept chain coverage generation to service use data of a telecommunications company. Concept chain coverage aims to cover the data not with single concepts but with chains of related concepts. The aim is not the full coverage but high enough coverage through a concise set of concept chains. We show that a relatively modest set of concept chains (4 to 10) can describe most of the data and that the performance of the algorithm is very acceptable for this case study.

Recently, we have witnessed the advent of personal manufacturing, where home users, small, medium, and Fortune 500 enterprises use devices such as 3D printers, CNC mills, and robotics to manufacture products locally. We have been developing a digital ecosystem of personal manufacturing for the last seven years. This ecosystem is currently used or being tried by 111 Fortune 2000 enterprises. In this paper, we focus on the creation of the cloud-based manufacturing operating system, 3DPrinterOS, to address an evolving critical problem of personal manufacturing. We introduce a novel software ecosystem architecture to sustain a massive communication load of command, control, and telemetry data to and from millions of manufacturing machines and users. Our solution allows users to create and deploy their own applications into 3DPrinterOS cloud operating system. Our long term experiments show that over the last five years, 95, 000 users have generated over three million CAD designs and machine codes, and produced more than 1, 030, 000 physical parts on 32, 000 manufacturing machines in 100 countries. Short term experiments showed that, on average, it is five times faster to perform a 3D print using 3DPrinterOS.

Presents corrections to the paper, A novel computational approach for global alignment for multiple biological networks,” (Djeddi, W.E., et al), Trans. Comput. Biol. Bioinf., vol. 15, no. 6, pp. 2060–2066, Nov./Dec. 2018.

The ultimate goal of a Traffic Information System (TIS) consists in properly informing vehicles about road traffic conditions in order to reduce traffic jams and consequently CO2 emission while increasing the user comfort. Therefore, the design of an efficient aggregation protocol that combines correlated traffic information like location, speed and direction known as Floating Car Data (FCD) is of paramount importance. In this paper, we introduce a new TIS data aggregation protocol called Smart Directional Data Aggregation (SDDA) able to decrease the network overload while obtaining high accurate information on traffic conditions for large road sections. To this end, we introduce three levels of messages filtering: (i) filtering all FCD messages before the aggregation process based on vehicle directions and road speed limitations, (ii) integrating a suppression technique in the phase of information gathering in order to eliminate the duplicate data, and (iii) aggregating the filtered FCD data and then disseminating it to other vehicles. The performed experiments show that the SDDA outperforms existing approaches in terms of effectiveness and efficiency.

Biclustering is a popular technique to study gene expression data, especially to identify functionally related groups of genes under subsets of conditions. Nevertheless, most of the existing biclustering algorithms only focus on the positive correlations of genes. However, recent research shows that groups of biologically significant genes may exhibit negative correlations. Thus, we need a novel way to efficiently unveil such a type of correlations. We introduce, in this paper, a new algorithm, called the Negative Bicluster Finder (NBF). The sighting features of the NBF stands in its ability to discover the biclusters of negative correlations using the theoretical results provided by the Formal Concept Analysis. Exhaust experiments are carried out on three real-life datasets to assess the performance of the NBF. Our results prove the NBF's ability to statistically and biologically identify significant biclusters.

Handling imbalanced datasets is a challenging problem in Knowledge Discovery in Databases. Several associative classification approaches have been proposed in order to treat this kind of data. However, these approaches suffer from a major drawback which is the use of a unique single measure for the filtering and selection of rules. This could be misleading since each measure is constructed in order to select a specific category of rules. To overcome such drawback, we introduce, IARCID, a novel approach that consists in aggregating several measures during the rule selection phase using the fuzzy Choquet Integral in order to produce a global measure. This latter is shown to be suitable for the classification of Imbalanced datasets from different domains. The performance of IARCID is assessed on four datasets with reference to three evaluation metrics. Experimentations show that IARCID outperforms approaches which use one single measure for the rule selection, by offering good classification results with all the datasets and according to all the assessment measures used for.

Sentiment analysis of citations in scientific papers is a new and interesting research area. In this paper, we focus on the problem of automatic identification of positive and negative sentiment polarity of citations in scientific papers. In this work, we conducted empirical research to investigate the classification of positive and negative citations. It is based on word vectors as a feature space, to which the examined citation context was mapped to. In order to handle with the huge amount of data, we have implemented our proposed approach in a distributed manner according to MapReduce paradigm through the Hadoop framework.

Missing values are a common problem in most data quality research. Most existing imputation methods of missing values are often evaluated using some distance measure computed between the reference data and the imputed one. Another alternative to evaluate the quality of the imputation is to assess the classification accuracy. In this paper, we address the problem of the evaluation of missing values imputation methods showing that the “best” imputation method according to one criterion is not necessary the “best” according to other criteria. Unfortunately, evaluating an imputation method according to a single aspect is not sufficient. Nevertheless, it is possible that we are interested in the “best” method in terms of two or three aspects simultaneously. This paper proposes different criteria aggregation based on the idea of preserving the characteristics of the original data and considers the evaluation technique of the missing values imputation methods as a Multiple Criteria Decision Making (MCDM) problem. We use the TOPSIS method to implement this multiple criteria evaluation technique. Carried out experiments on benchmark datasets confirm the soundness of our approach.

Due to the rapid progress of biological networks for modeling biological systems, a lot of biomolecular networks have been producing more and more protein-protein interaction (PPI) data. Analyzing protein-protein interaction networks aims to find regions of topological and functional (dis)similarities between molecular networks of different species. The study of PPI networks has the potential to teach us as much about life process and diseases at the molecular level. Although few methods have been developed for multiple PPI network alignment and thus, new network alignment methods are of a compelling need. In this paper, we propose a novel algorithm for a global alignment of multiple protein-protein interaction networks called MAPPIN. The latter relies on information available for the proteins in the networks, such as sequence, function, and network topology. Our algorithm is perfectly designed to exploit current multi-core CPU architectures, and has been extensively tested on a real data (eight species). Our experimental results show that MAPPIN significantly outperforms NetCoffee in terms of coverage. Nevertheless, MAPPIN is handicapped by the time required to load the gene annotation file. An extensive comparison versus the pioneering PPI methods also show that MAPPIN is often efficient in terms of coverage, mean entropy, or mean normalized.

Over the past several years, social networking services or micro-blogs have become ubiquitously accessible anytime and contain users' opinions expressed in the form of short text messages. In this paper, we introduce a new automatic approach named FEmoRec for emotional context recognition from online social networks that applies a semantic similarity measure based on Multi-Layer Perceptron Neural Net Model. We rely on the assumption that a tweet may belong to many emotional categories with different membership degrees. We classify the tweet by computing an emotion vector that represents the tweet's fuzzy membership values to Ekman's emotion classes. Carried out experiments emphasize the relevance of our proposal, compared to other methods.

User preference discovery aims to detect the patterns of user preferences for various topics of interest or items such as movie genre or category. Preferences discovery is a crucial stage in the development of intelligent personalization systems. Although a variety of studies have been proposed in the literature addressing a wide range of applications such as recommender systems or personalized search, only a few of them have considered the management of imprecision in the representation of user and item features. This paper aims to address the above issue by using fuzzy sets. The paper proposes a general framework for preferences discovery through fuzzy sets and fuzzy models and it introduces a new algorithm for representing and discovering fuzzy user interest profile. Based on the results of the empirical evaluation, the proposed approach outperforms two well-known recommendation approaches in terms of well-known quality assessment metrics, namely: discounted cumulative gain, precision, recall, as well as F1-measure.

A majority of existing biclustering algorithms for microarrays data focus only on extracting biclusters with positive correlations of genes. Nevertheless, biological studies show that a group of biologically significant genes may exhibit negative correlations. In this paper, we propose a new biclustering algorithm, called NBic-ARM (Negative Biclusters using Association Rule Mining). Based on Generic Association Rules, our algorithm identifies negatively-correlated genes. To assess NBic-ARM’s performance, we carried out exhaustive experiments on three real-life datasets. Our results prove NBic-ARM’s ability to identify statistically and biologically significant biclusters.

International audience

The main objective of the VANET networks is to improve road safety as well as transport efficiency through the use of communications technology and the emergence of wireless devices at low cost. Thus, the design of an efficient dissemination protocol, that informs vehicles about interesting safety and nonsafety events, is of paramount importance. The thriving challenge would be to maximize the delivery ratio by avoiding as far as possible the broadcast storm problem. A scrutiny of the literature wealthy number of approaches highlights that all of them fail to fulfill with the "sine qua non" requirements that we introduce. In this paper and to palliate this shortage, we introduce a new infrastructure-less geocast approach, called DPMS (data Dissemination Protocol based on Map Splitting). The main thrust of DPMS stands in reaching a high delivery ratio as well as a high geocast precision by sending messages only to vehicles in the Zone of Relevance (ZOR) with a minimum overhead cost. Carried out experiments showed that DPMS outperforms its competitors in terms of effectiveness and efficiency.

Over the last decade, several ontology reasoners have been proposed to overcome the computational complexity of inference tasks on expressive ontology languages. Nevertheless, it is well-accepted that there is no outstanding reasoner that can outperform in all input ontologies. Thus, an algorithm selection problem have emerged in this field of study. In this paper, we describe first steps to develop a new system to provide user support when looking for guidance over ontology reasoners. Our main goal is to be able to automatically rank a set of candidate reasoners for any given ontology. Robustness standing for the ability of reasoner to correctly achieve a reasoning task within a fixed time limit is our primary ranking criterion. Our ranking method follows a meta-learning approach and applies bucket order rules. An extensive experiments covering over 2500 well selected real-world ontologies and six state-of-the-art of the most performing reasoners was carried out to provide enough data for the study. Our prediction and ranking results are encouraging, witnessing the potential benefits of the proposed approach.

The fast growing of on-line multimedia content have created the need to investigate new paradigms and techniques allowing to express how to index, retrieve and explore such contents. Indeed, nowadays, Movie becomes a predominant form of entertainment in human life. Most video websites such as YouTube and a number of social networks allow users to freely assign a rate to watched or bought videos or movies. In this paper, we introduce a movie rating recommendation approach based on the exploitation of the Hidden Markov Model (HMM). Specifically, we extend the HMM to include user's rating profiles, formally represented as triadic concepts. Carried out experiments emphasize the relevance of our proposal and open many thriving issues.

Trust systems represent a significant trend in decision support for social networks’ service provision. The basic idea is to allow users to rate each other even without being direct neighbours. In this case, the purpose is to derive a trust score for a given user, which could be of help to decide whether to trust other users or not. In this article, we investigate the properties of trust propagation within social networks, based on the notion of transitivity , and we introduce the TISoN model to generate and evaluate T rust I nference within online So cial N etworks. To do so, ( i ) we develop a novel TPS algorithm for T rust P ath S earching where we define neighbours’ priority based on their direct trust degrees, and then select trusted paths while controlling the path length; and, ( ii ) we develop different TIM algorithms for T rust I nference M easuring and build a trust network. In addition, we analyse existing algorithms and we demonstrate that our proposed model better computes transitive trust values than do the existing models. We conduct extensive experiments on a real online social network dataset, Advogato. Experimental results show that our work is scalable and generates better results than do the pioneering approaches of the literature.

In this paper, we introduce models for the optimization of the door-to-door freight transportation. The main thrust of these models stand in the allowance to forecast of freight amounts that will be transported daily considering capacity as well as time constraints (e.g. time availability of retailers and customers). We also have integrated real-world constraints to meet practical difficulties that may actually face transportation (e.g., fixed number of working hours of drivers, availability of vehicles). The transportation scheme is characterized by a consolidation center. We assumed, also, that freight transportation is mutualized. Thus, we have split the FDPTW (Pickup and Delivery Problem with Time Windows) into two problems: (i) is a Vehicle Routing Problem with Time Windows (VRPTW) with Pickup; (ii) is a VRPTW with Delivery. The proposed models are solved by LINGO. The output results are the optimal trucks in each model. Lastly, one of the elaborated models (VRPTW with delivery) is tested with Solomon's benchmark and for the other models we propose different numerical experiments to validate our contributions.

Reasoning with ontologies is one of the core tasks of research in Description Logics. A variety of reasoners with highly optimized algorithms have been developed to allow inference tasks on expressive ontology languages such as OWL (DL). However, unexpected behaviours of reasoner engines is often observed in practice. Both reasoner time efficiency and result correctness would vary across input ontologies, which is hardly predictable even for experienced reasoner designers. Seeking for better understanding of reasoner empirical behaviours, we propose to use supervised machine learning techniques to automatically predict reasoner robustness from its previous running. For this purpose, we introduced a set of comprehensive ontology features. We conducted huge body of experiments for 6 well known reasoners and using over 1000 ontologies from the ORE’2014 corpus. Our learning results show that we could build highly accuracy reasoner robustness predictive models. Moreover, by interpreting these models, it would be possible to gain insights about particular ontology features likely to be reasoner robustness degrading factors.

Microaggregation for Statistical Disclosure Control (SDC) has been shown to be an efficient method to hamper individual identification. Indeed, micro data are wrapped in such a way that can be published and mined without providing any private information that can be linked to specific individuals. In this respect, a microaggregation method would seek to lower the information loss resulting from this replacement process. The challenge is how to minimize the information loss during the microaggregation process. In this paper, we introduce a new algorithm, called AdMicro-FSOM for the multivariate microaggregation task. The main thrust of this algorithm stands in its handling fuzzy partition into a microaggregation method. The extensive carried out experiments show the obtention of low information loss, even when handling noisy data. In addition, the obtained results sharply outperform those obtained by the pioneering algorithms of the dedicated literature.

The problem of extracting the minimal transversals from a hypergraph is known to be particularly difficult. Given that the number of minimal transversals can be exponential even for moderately sized hypergraphs, we propose, in this paper, a concise representation of minimal transversals in order to optimize the computation time and we introduce a new algorithm, called IRRED-ENGINE, which extracts all the minimal transversals from a subset. Experiments carried out on several types of hypergraphs, showed that IRRED-ENGINE obtains very interesting results evaluated through a compactness measure. To illustrate the benefit of our approach, we show how our concise representation can be used to solve the dependency inference problem by computing a concise cover of functional dependencies.

During the last years, many works focused on the exploitation and the extraction of rare patterns. In fact, these patterns allow conveying knowledge on rare and unexpected events. They are hence useful in several application fields. Nevertheless, a m

In the belief function theory, several measures of uncertainty have been introduced. One of their possible use is unreliable source discounting before the fusion stage. Two different measures of uncertainty exist which are the intrinsic and extrinsic ones. The intrinsic measure makes it possible to assess the source's confusion whereas the extrinsic one measures the contradiction between sources. In this paper, we associate both measures in order to estimate the global reliability of a source. This method, named Generic Discounting Approach (GDA), is proposed in two different versions: Weighted GDA and Exponent GDA. Those reliability measures are integrated into a classifier. The method was tested, against to some pioneer approaches, on several UCI datasets as well as on an urban image classification problem and showed very encouraging results.

Decision making by considering multiple information sources could provide interesting results. For that reason, fusion formalisms were a major concern in the belief function community. In this context, the Belief function theory allows information fusion thanks to its combinations tools that it integrates. Nevertheless, belief function theory highlights a limit in the merging of contradictory (conflictual) sources. Many authors tackled this problem offering contributions in this field. Unfortunately, no proposed operator has distinguished by its adequacy regardless the type of handled sources. In this paper, we demonstrate the limits of some referenced works and we diagnostic the issues origin. We propose a conflict management approach based on an extra-information that guides the treatment. We also integrate a generic associative base borrowed from the data mining domain in order to apply the adequate conflict management.

Recently, an increasing number of information retrieval studies have triggered a resurgence of interest in redefining the algorithmic estimation of relevance, which implies a shift from topical to multidimensional relevance assessment. A key underlying aspect that emerged when addressing this concept is the aggregation of the relevance assessments related to each of the considered dimensions. The most commonly adopted forms of aggregation are based on classical weighted means and linear combination schemes to address this issue. Although some initiatives were recently proposed, none was concerned with considering the inherent dependencies and interactions existing among the relevance criteria, as is the case in many real‐life applications. In this article, we present a new fuzzy‐based operator, called iAggregator, for multidimensional relevance aggregation. Its main originality, beyond its ability to model interactions between different relevance criteria, lies in its generalization of many classical aggregation functions. To validate our proposal, we apply our operator within a tweet search task. Experiments using a standard benchmark, namely, Text REtrieval Conference Microblog, emphasize the relevance of our contribution when compared with traditional aggregation schemes. In addition, it outperforms state‐of‐the‐art aggregation operators such as the Scoring and the And prioritized operators as well as some representative learning‐to‐rank algorithms.

The main moan that can addressed to the pioneering approaches of fuzzy clustering stand in their approximate management of a noisy surroundings as well as their snugness dependency of an apriori determination of the number of clusters. The aim of this paper is twofold: First, we introduce a new algorithm, called PFKCN, based on neural network. This algorithm introduces both membership and typicality values, simultaneously, into the Kohonen Network clustering. Then, we tackle the problem of estimating the number of clusters, by using a multi level PFKCN based clustering algorithm, called Multi-PFKCN. The latter is able to find the optimal number of clusters by using a statistical criterion, that aims at measuring the quality of obtained partitions. Carried out experiments on real-life data sets highlights a very encouraging results in terms of exact determination of optimal number of clusters.

Folksonomy systems and social bookmarking tools are rapidly spreading on the Web. Their steady increase reveals a more dynamic and interactive space, in which users can individually and freely share online resources and assign terms to them. In this paper, we address the problem of how to exploit these rich systems to enrich existing ontologies. Our approach taps into external tools such as the sense inventory WordNet to fix mistakes resulting from the free tagging and takes advantage from semantic relationships between tags to recognize resources' context. We carried out evaluations that mainly paid attention on the particularly challenging problem of resources' ambiguity to prove that not all resources shared in folksonomies are relevant resources for Dbpedia ontology. In this respect, it is worth of mention that our pre-processing step increases precision values to achieve an average precision of 80%.

In this paper, we address the problem of how toexploit folksonomy systems and social bookmarking tools toenrich DBpedia ontology while unifying the formal knowledgerepresented by DBpedia ontology with the informal knowledge emerging from tagging. Our approach taps into external tools such as the sense inventory WordNet to correct mistakes resulted from the free tagging and takes advantage from semantic relationships between tags to recognize resources'context. We performed evaluations with focus on the particularly challenging problem of resources' ambiguity to prove that not all resources shared in folksonomies are relevant resources for DBpedia ontology and to show that our pre-processing step increases precision values.

Tackling, within a classification task, to the problem of inaccuracy explains the development of new theories that offer a formal treatment of imprecise information, especially the theory of fuzzy sets who suggested a new approach taking advantage of the concept of membership function. Nevertheless, clustering algorithms still show limits, particularly for the estimation of the number of clusters. In this paper, through a state of the art of the main fuzzy classification algorithms, we introduce a new algorithm, called Fuzzy-MSOM. The latter aims at palliating to drawback of the determination of the suitable number of clusters in a given data set. Thus, the clustering process is carried out through a multi-level approach. Through the use of fuzzy clustering validity indices, Fuzzy-MSOM overcomes the problem of the estimation of clusters number. The experimental result shows that the proposed clustering technique provides better results compared to the previous algorithms.

The huge number of association rules represent the main hamper that a decision maker faces. In order to bypass this hamper, an efficient selection of rules has to be performed. Since selection is necessarily based on evaluation, many interestingness measures have been proposed. However, the abundance of these measures gave rise to a new problem, namely the heterogeneity of the evaluation results and this created confusion to the decision. In this respect, we propose a novel approach to discover interesting association rules without favoring or excluding any measure by adopting the notion of dominance between association rules. Our approach bypasses the problem of measure heterogeneity and unveils a compromise between their evaluations. Interestingly enough, the proposed approach also avoids another non-trivial problem which is the threshold value specification.

Improving trust in social networks appears as the first step toward addressing the existing confidence and privacy concerns related to online social networks. Direct trust is used to develop different trust-based methods such as transitivity and access control, however how to compute direct trust levels is rarely discussed in the literature. To address some of the current limitations, we introduce a novel approach for generating trusted social networks and we compute trust levels between users having direct relationships. Experimental results with data extracted from FOAF files show that our work presents high accuracy.

The extremely large number of association rules that can be drawn from – even reasonably sized datasets, bootstrapped the development of more acute techniques or methods to reduce the size of the reported rule sets. In this context, the battery of re

Applying classical association rule extraction framework on fuzzy data sets leads to an unmanageably highly sized association rule sets -compounded with an information loss due to the discretization operation -that often constitutes a hamper towards an efficient exploitation of the mined knowledge. To overcome such drawback, we advocate the extraction and the exploitation of compact and informative generic basis of fuzzy association rules. This generic basis constitutes a compact nucleus of fuzzy association rules. In addition, we introduce an axiomatic system to ensure the derivation mechanism of all the remaining rules. Obtained preliminary results are very encouraging and they highlight a very important reduction of the number of the extracted fuzzy association rules without information loss.

Providing efficient and easy-to-use graphical tools to users is a promising challenge of data mining (DM). These tools must be able to generate explicit knowledge and to restitute it. Visualization techniques have shown to be an efficient solution to achieve such goal. Even though considered as a key step in the mining process, the visualization step of association rules received much less attention than that paid to the extraction one. Nevertheless, some graphical tools have been developed to extract and visualize association rules. In those tools, various approaches are proposed to filter the huge number of association rules before the visualization step. However both DM steps (association rule extraction and visualization) are treated separately in a one way process. Our approach differs, and uses meta-knowledge to guide the user during the mining process. Standing at the crossroads of DM and Human-Computer Interaction (HCI), we present an integrated framework covering both steps of the DM process. Furthermore, our approach can easily integrate previous techniques of association rule visualization.

Traditional framework for mining association rules has pointed out the derivation of many redundant rules, in order to be reliable in a decision making process, such discovered rules have to be concise and easily understandable for users or as well as an input to visualization tools. We present a 3D histograms-based visualization prototype for handling generic bases of association rules. An interesting feature of the prototype is that it provides a "contextual" exploration of such rule set. Such additional displayed knowledge, based on the construction of fuzzy meta-rules, enhances man-machine interaction by emulating a cooperative behavior.

Traditional framework for mining association rules has pointed out the derivation of many redundant rules. In order to be reliable in a decision making process, such discovered rules have to be both concise and easily understandable for users, and/or as an input to visualization tools [P. Adriaans et al. (1997)]. We present a graphical visualization prototype for handling generic bases of association rules. We discuss also the most adequate graphical visualization technique depending on the intrinsic structure of the generic bases of association rules. An interesting feature of the prototype is that it provides a "contextual" exploration of such rule set. Such exploration, based on the discovery of fuzzy meta-rules, enhances man-machine interaction by emulating a cooperative behavior.

The need for sophisticated analysis of textual data is becoming very apparent. In the general context of knowledge discovery, text mining techniques aim to discover additional information from hidden patterns in unstructured large textual collections. Hence, we are interested especially in the extraction of the associations from unstructured databases. The objective is two fold. First, to propose a conceptual approach, based on the formal concept analysis (Ganter and Wille, 1999) and a semantic pruning, in order to discover implicit association rules, from large textual corpus. Second, to introduce an algorithm to derive additional and implicit association rules, using an associated taxonomy, from the already discovered association rules.