Extracting Frequent (Closed) Seasonal Gradual Patterns Using Closed Itemset Mining
Résumé
In this paper, we address the issue of mining seasonal gradual patterns, which consists in identifying attribute co-variations in the form “when X increases/decreases, Y in- creases/decreases”, that seasonally appear in data (”X and Y co- increase from March to June in more than 75% of the observed years”). Such kind of patterns has recently emerged for analyzing the sequences of temporal data and some algorithms have been proposed to automatically extract these patterns from numerical data. However, mining seasonal gradual patterns remains very challenging as the task is more complex than extracting gradual patterns in a simple sequence. In fact, one can note that extracting gradual patterns frequently appearing over a precise period of time (season) amounts to extracting, for each season, the gradual patterns common to multiple sequences. From this observation, we propose an algorithm to extract seasonal gradual patterns from temporal data sequences which exploits closed frequent itemsets mining algorithms to perform a search season by season. Experimental results obtained on a real world dataset show that compared to state-of-the-art algorithms, our proposed algorithm is efficient and can extract the complete set of frequent (closed) seasonal gradual patterns from the sequences of temporal numerical data.