Résumé
Fuzzy temporal gradual patterns (FTGPs) enable the identification of correlations between multiple attributes in time series data and data streams. For example, in a data stream with the attributes {date, temperature, rainfall}, FTGPs can uncover patterns like "an increase in temperature is followed by a decrease in rainfall approximately two days later." In this paper, we introduce novel techniques for mining FTGPs that incorporate mutual information, clustering, metaheuristics and fuzzy inference. We evaluate our approach using real-world datasets and benchmark it against existing methods. The results show that our technique surpasses current methods in both accuracy and efficiency. Additionally, we show that FTGPs are effective in analyzing multiple data streams or time series data.