Employee retention is a major issue for companies. In our previous work [1], we proposed an approach that extracts frequent patterns from employee trajectories to identify those associated with improved retention thanks to survival analysis. Those are called survival-changing patterns. However, the large volume and complexity of these patterns limited their interpretability and practical use by Human Resources (HR) teams. In this paper, we introduce an interactive dashboard that combines sequential pattern mining with survival analysis to support HR decision-making. Our approach refines and visualizes survival-changing patterns, enabling a clearer understanding of attrition dynamics. The dashboard provides a global view of retention trends, an identification of at-risk employees, and a simulation of targeted interventions. Our contributions include the definition of HR personas and use cases, a method for selecting representative pattern of a sequence, the design of tailored visualizations for employee trajectories, and a real-world illustration of how the tool works in practice.

This paper presents a survey on automatic or semi-automatic recommendation systems which help users to create dashboards. It starts by showing the important role that dashboards play in data science, and give an informal definition of dashboards, i.e., a set of visualizations possibly with linkage, a screen layout and user feedback. We are mainly interested in systems that use a fully or partially automatic mechanism to recommend dashboards to users. This automation includes the suggestion of data and visualizations, the optimization of the layout and the use of user feedback. We position our work with respect to existing surveys. Starting from a set of over 1000 papers, we have selected and analyzed 19 papers/systems along several dimensions. The main dimensions were the set of considered visualizations, the suggestion method, the utility/objective functions, the layout, and the user interface. We conclude by highlighting the main achievements in this domain and by proposing perspectives.

Employee attrition is a pervasive problem for many organizations, and reducing it has become a key goal in the business world. Although there is a substantial body of literature on predicting customer attrition, the literature on employee attrition is comparatively limited. Moreover, even studies that do address employee attrition often fail to consider the impact of time and duration on attrition rates. In this context, the present paper aims to fill this gap in the literature by combining frequent pattern mining in sequences of events and survival analysis with Kaplan-Meier to examine how event sequences affect employee attrition. We introduce the notion of survival-changing sequential patterns that highlight events that significantly impact the survival estimator. Our findings suggest that certain patterns are associated with a higher rate of employee retention, while the addition of specific events can have a positive or negative impact on employee survival. This research highlights the importance of analyzing event sequences and duration when attempting to reduce employee attrition rates. The practical implications of this research are significant, as it provides a framework for organizations seeking to retain their employees and enhance their overall performance.

In this paper, we present a method for the automatic generation of dashboards (DBo) using a genetic algorithm (GA). A DBo is a set of visualizations, with possible linkage, intended to help users explore and analyse a dataset. Our Automatic Dashboard Generation System (ADGS) considers several input models for data, user and visualizations. We propose to represent a solution (i.e. a DBo) as a variable size matrix in which rows are visualizations and columns are data attributes. This representation can be evolved with a GA. For this purpose, we define genetic operators for DBo such as random generation, crossover, and mutation. We propose a fitness function to evaluate the quality of a DBo, as well as selection and replacement schemes. Finally, we present results with a benchmark dataset and a given user scenario. We show that the GA can find DBo that maximizes the evaluation function and that can be interesting for novice users. In perspectives, we will improve further the GA and we will study how to automatically propose a layout of the optimized DBo.

Evolutionary Algorithms have been regularly used for solving multi and many objectives optimization problems. The effectiveness of such methods is determined generally by their ability to generate a well-distributed front (diversity) that is as close as possible to the optimal Pareto front (proximity). Analysis of current multi-objective evolutionary frameworks shows that they are still sub-optimal and present poor versatility on different geometries and dimensionalities. For that, in this paper, we present AGE-MOEA++, a new Multi and Many Objective Evolutionary Algorithm that: (1) incorporates the principle of Pareto Front (PF) shape fitting to enhance the convergence in different shaped high dimensional objective spaces, and (2) adapts K-means ++ fundamentals in order to best manage the diversity in non-uniform distributed PF. The empirical study shows that our proposal has better results than the state-of-the-art approaches in terms of IGD and is competitive in terms of GD.

The adoption of deep learning models has brought significant performance improvements across several research fields, such as computer vision and natural language processing. However, their “black-box” nature yields the downside of poor explainability: in particular, several real-world applications require - to varying extents - reliable confidence scores associated to a model's prediction. The relation between a model's accuracy and confidence is typically referred to as calibration. In this work, we propose a novel calibration method based on gradient accumulation in conjunction with existing loss regularization techniques. Our experiments on the Named Entity Recognition task show an improvement of the performance/calibration ratio compared to the current methods.

The study of temporal sequences is a main topic in different domains, especially for human mobility mining. This article defines the Fuzzy Temporal Hamming (FTH) distance between temporal sequences. This new measure generalises the Hamming distance and improves it by introducing a fuzzy time-window. This fuzzy approach tolerates temporal distortions as shifting and permutations. Moreover, the time computation of FTH is competitive with other Optimal Matching methods used for temporal sequences comparison. To validate this approach, we cluster data from a real Time-Use Survey and we compare the results obtained with other methods.

Both minimal set cover and maximum 1-coverage are known to be NP-Hard when using homogeneous and heterogeneous wireless sensor networks. The proposed solutions in the literature guaranteeing k-coverage in WSN affect either the energy, the network lifetime, or the quality of service. These solutions still incur time complexity and energy overhead that increase with WSN large-scale, connectivity and size of coverage holes. To avoid these issues, this paper proposes a minimal semi-deterministic deployment model in a mobile wireless sensor network. This model is based on Pick's theorem to guarantee a maximum 1-coverage. This requires the subdivision of the area of interest into square sub-areas according to a pre-established grid, and then an initial deployment of sensor nodes randomly around each sub-area center. The main strengthen of the proposed deployment approach is the minimization of the nodes' movement from their initial positions. Where, each node moves once and remains active until it is exhausted. The results show a high efficiency in terms of increasing the network lifetime and the coverage compared to some well-known approaches.

Cooperative Intelligent Transport Systems(C-ITS) focus on improving safety, comfort, traffic and energy efficiency. Vehicle speed and other speed based indicators are commonly used parameters in traffic research for generation of driving profiles. The main goal for studying speed variation is to gain a better understanding on why drivers respond in certain ways to road/traffic conditions and to discover factors which affect their actions. The aim of this paper is to use a real data-set of Cooperative Awareness Messages generated in a naturalistic driving C-ITS environment to generate speed signatures. We apply a segmentation technique and statistical analysis in generation and evaluation of road speed signatures. Based on our approach, interesting characteristics on the evolution of driving behavior are revealed.

This paper deals with the Campaign Allocation Problem of commercial Ads in TV breaks that we formalize as a multi-stakeholders multiobjective problem with highly competing objectives for different brands and numerous constraints. The problem is NP-hard with a high dimensional objective space and scalability issues in terms of the number of breaks. Moreover, the expected solution should be able to focus on a sub-part of the Pareto front according to decision maker’s (DM) knowledge. To tackle these challenges, we propose to use R-NSGA-II, a Many-Objective Evolutionary Algorithm (MaOEA), combined with a novel gene encoding/decoding process. Experiments show that this approach obtains better results than usual MaOEA (NSGA-II, NSGA-III) according to industrial performance criteria, scales to large instances, and incorporates decision maker’s preferences during the optimization process.

In this era of big data, as relational databases are inefficient, NoSQL databases are a workable solution for data storage. In this context, one of the key issues is the veracity and therefore the data quality. Indeed, as with classic data, geospatial big data are generally fuzzy even though they are stored as crisp data (perfect data). Hence, if data are geospatial and fuzzy, additional complexities appear because of the complex syntax and semantic features of such data. The NoSQL databases do not offer strict data consistency. Therefore, new challenges are needed to be overcome to develop efficient methods that simultaneously ensure the performance and the consistency in storing fuzzy geospatial big data. This paper presents a new methodology that tackles the storage issues and validates the fuzzy spatial entities' consistency in a document-based NoSQL system. Consequently, first, to better express the structure of fuzzy geospatial data in such a system, we present a logical model called Fuzzy GeoJSON schema. Second, for consistent storage, we implement a schema-driven pipeline based on the Fuzzy GeoJSON schema and semantic constraints.

Vehicles in an Intelligent Transport Network exchange a lot of messages. Every message sent is generated with an identifier of the transmitting vehicle. To respect the user privacy, an identifier is kept only over a specified time interval. The need that arises is, given that multiple identifiers are assigned to a vehicle, are we able to group the identifiers and detect those which belong to the same vehicle? We solved this Trajectory-User Linking problem by chaining anonymous trajectories to potential vehicles by considering similarity in movement patterns. Our method managed to link trajectory segments to their common vehicles which we validated through map matching of the trajectories using QGIS.

Cooperative Intelligent Systems development is growing and the data they produce is increasing exponentially. This amount of data will soon be large enough to fall in big data paradigm. We propose to exploit these data as data stream. We aim to detect anomaly on the road using concept drift detection methods over data stream. To achieve this purpose, we create a data generation tool to obtain large data-sets of vehicles taking an avoiding behavior and detect obstacles through crowdsensing. We use two scenarios that we aim to detect: a stopped car and a growing pothole. We focus our study on the vehicle orientation information on which we apply Page-Hinkley and ADWIN methods. We obtain interesting detection results with ADWIN on the stopped car scenario. The Page-Hinkley algorithm is obtaining good results but with a latency that makes it unexploitable in real context. But for the pothole detection, both approaches are not providing significant results.

Connected vehicles is a growing field of research that will produce great amount of data in near future. These data can be mined to generate traffic prediction, detect driver profile, find alternative route, etc.. This will help car manufacturers, road operators, telecom operators and other actors in the sector to improve road safety and drivers comfort. But nowadays few data are collected to create these tools.In this paper we compare different completion approach on data extracted from real experimentation on road to perform efficient driving profile detection. We analyze the deviations of the driver headings along a defined trajectory on specific Points of Interest (POI) to extract the driving profiles.

The area coverage is a non-trivial problem in Wireless Sensors Network (WSN) due to the lack of knowledge in the minimum sensor nodes set that can cover an area of interest and limitation of energy reserve, monitoring and communication ranges. Coverage protocols address two fundamental questions: how can the coverage performance of deployed sensor nodes in a monitoring region be evaluated; and how can the coverage performance be improved when wireless sensor network cannot effectively satisfy quality of service requirements? Constructing sets of minimal nodes to optimize the coverage problem push researchers to use different techniques from different domains, such as Voronoï Diagram, Connected Dominating Set, Clustering, and others. Each technique deals with the problem by its own philosophy. Our approach aims to use a hybrid model combining Diagram of Voronoï, Clustering, and Connected Dominating Set in order to benefit from the advantages of these three models to optimize area coverage, keep connectivity, and minimize energy consumption. The simulation showed the ability of our approach to ensure optimal coverage with minimal power consumption for a longtime.

Sensor nodes suffer from a lot of constraints, such as the limited energy source, the short radius of communication, the low processing and storage power. These fuzzy constraints deplete the energy of the implemented sensor network, especially, in coverage and routing data in the large-scale zones. The main constraint in WSNs is the energy. To outcome this constraint, we have proposed a Data Routing-Scheduling via Efficient-Energy Area Coverage in wireless Sensor Network protocol (EE_AC_DR) based on our proposed Dijkstra Front-Back algorithm. In fact, our protocol uses Dijkstra algorithm with steps to pass the nodes in the superior Cluster, or return it to the lower Cluster to choose the designed node as Cluster-Heads for playing a multiple role; a role of Cluster-Head, a role of coverage in the entire cluster and a role of communication routing the information in the best and shortest path towards the base station. This protocol is divided into the following five (5) steps: 1) the construction of the Clusters; 2) the choice of the Clusters-Heads; 3) the choice of the best neighbors; 4) the construction of Head-Clusters-Head, and 5) the choice of best Cluster-Head neighbors. We demonstrate, using some large scale simulations, that this protocol enhances the performances of the network and allows taking the best choices regarding the use of only one node in each cluster to monitor and communicate data using the shortest path with minimal cost.

C-ITS (Cooperative Intelligent Transport Systems) are growing very quickly in many parts over the world. Their benefits are of importance for fuel consumption, traffic management and road safety. Their deployments are in advanced steps in many countries. Their impacts on human life are not clearly known. For this reason, we propose to analyze a large set of data collected during real tests on open roads with many connected vehicles. This analysis allows us to focus on relevant information like driver profiles, abnormal driving behaviours, etc. In this paper, we present a methodology to analyze data provided by a real experimentation of C-ITS mobile stations. We mainly analyze the headings of each driver when approaching some Points of Interest (POI). We use unsupervised machine learning approaches to detect driver profiles. The interesting features about driver profiles obtained need to be enhanced and confirmed for larger data-sets.

This paper aims to qualify the behaviour of contributors to OpenStreetMap (OSM), a volunteered geographic information (VGI) project, through a multigraph approach. The main purpose is to reproduce contributor’s interactions in a more comprehensive way. First, we define a multigraph that combines existing spatial collaboration networks from the literature with new graphs that illustrate collaboration based on specific aspects of the VGI modes of contribution through semantics, geometry and topology. Indeed, the ways that contributors interact with one another through editing, completion, or even consumption may provide additional information on each user’s operation mode and therefore, on the quality of the contributed data. Social collaborations drawn from indirect criteria – for example, comparisons between contributors’ activity areas – can also be contemplated under another network. Second, the resulting multigraph is analysed using data mining approaches to characterise individuals and identify behavioural groups. The implementation of a multiplex network based on an OSM data sample and an initial analysis make it possible to identify useful behaviours for data qualification. The initial results characterise some contributors as pioneers, moderators and truthful contributors, according to their special roles in the graphs. Mapping elements that include these contributors’ participation are likely to be reliable data

Internet of Things (IoT) and Blockchain (BC) is an innovative paradigm that is gaining ground in smart environments. In intelligent classrooms, IoT makes our exchange easier with the prominent advent of smart devices, connected objects and sensors. However, the important research direction in this kind of IoT-Based Ubiquitous Learning Environment (ULE) is security and privacy that remaining essential challenges. Previously, we exposed the initial architecture of ULE based on BC technology and the educational services that can be delivered via this platform. In this paper, we investigate deeper and we highlight the main component of our ULE known as integrated IoT-ubiquitous platform using BC. The collection of data exchanged across devices is determined by the miner that preserves security using transactions that trace communications. Finally, in this study we demonstrate our preliminary experimental results that show the effectiveness of the proposed decentralized platform which is more secure by analyzing confidentiality, integrity, and availability.

In order to inform about sensors veracity and handle the data imprecision, an interactive visualization tool for industrial needs has been developed and presented in this paper. The tool allows user to get deep understandings in a multi-sensor context, especially when considering harvested web data. In order to deal with data imperfection, our methodology is based on quantiles and on the specific modeling for missing values. We present diverse dashboards and visual indicators serve to validate common flow data and help to discover hidden knowledge. According to a use case, we show how our visualization approaches can assist to review data quality about possible critical situations.

Ubiquitous learning environments have an increasing trend considering the huge number of connected smart devices dedicated to educational services. Ubiquitous learning (U-learning) provides to students the possibility to learn at anyplace and anytime within the collaborative environment using interactive multimedia system that enables effective communication among teacher and learners. The architecture of ubiquitous learning environment (ULE) still suffers from the problems of vulnerability. Blockchain (BC) technology recently explored to provide much more privacy and security using essentially peer-to-peer (P2P) networks has a significant role in the development of decentralized topologies. This paper expounds a novel BC-based architecture for ULE that preserves the benefits of security and privacy. The architecture offers new opportunities to design secured collaborative learning system. It provides data exchange with BC using trust methods within the decentralized topology. The evaluation of this implemented system demonstrates its efficiency while it delivers security and privacy for ULE.

This paper presents a new approach for visually mining multivariate datasets and especially large ones. This unsupervised approach proposes to mix a SOM approach and a pixel-oriented visualization. The map is considered as a set of connected pixels, the space filling is driven by the SOM algorithm, and the color of each pixel is computed directly from data using an approach proposed by Blanchard et al. The method visually summarizes the data and helps in understanding its inner structure.