is a Post-doctoral researcher at LIS (UMR 7020) and LPL (UMR 7309) laboratories, Aix-Marseille Université, France. He received his Ph.D. in 2018 from Aix-Marseille Université. He does research in time series prediction, causality detection, and machine learning.
is Full Professor at Aix-Marseille Université, France. He received his Ph.D. (1986) and the Habilitation for Directing Research (1992) from the University of Nice-Sophia Antipolis, France. He joined the Department of Mathematics and Computer Science of Blaise Pascal University, Clermont-Ferrand, France, and became a full professor in 1993. In 2000, he joined the Mediterranean University, IUT d’Aix-en-Provence, France. His research, at LIS laboratory UMR 7020, interests include statistical databases, database design, formal concept analysis, data mining, and data warehouses.
obtained the Ph.D. degree in computer science from the Aix Marseilles University (France) in 2005. He is an assistant professor at the Aix Marseilles University - IUT of Aix en Provence and is a member of the LIS laboratory. He studies lattice algorithmic and temporal database mining.
Knowledge discovery systems are nowadays supposed to store and process very large data. When working with big time series, multivariate prediction becomes more and more complicated because the use of all the variables does not allow to have the most accurate predictions and poses certain problems for classical prediction models. In this article, we present a scalable prediction process for large time series prediction, including a new algorithm for identifying time series predictors, which analyses the dependencies between time series using the mutual reinforcement principle between Hubs and Authorities of the Hits (Hyperlink-Induced Topic Search) algorithm. The proposed framework is evaluated on 3 real datasets. The results show that the best predictions are obtained using a very small number of predictors compared to the initial number of variables. The proposed feature selection algorithm shows promising results compared to widely known algorithms, such as the classic and the kernel principle component analysis, factor analysis, and the fast correlation-based filter method, and improves the prediction accuracy of many time series of the used datasets.