From Sequence Mining to Multidimensional Sequence Mining

Sequential pattern mining has been broadly studied and many algorithms have been proposed. The first part of this chapter proposes a new algorithm for mining frequent sequences. This algorithm processes only one scan of the database thanks to an indexed structure associated to a bit map representation. Thus, it allows a fast data access and a compact storage in main memory. Experiments have been conducted using real and synthetic datasets. The experimental results show the efficiency of our method compared to existing algorithms. Beyond mining plain sequences, taking into account multidimensional information associated to sequential data is for a great interest for many applications. In the second part, we propose a characterization based multidimensional sequential patterns mining. This method first groups sequences by similarity; then characterizes each cluster using multidimensional properties describing the sequences. The clusters are built around the frequent sequential patterns. Thus, the whole process results in rules characterizing sequential patterns using multidimensional information. This method has been experimented towards a survey on population daily activity and mobility in order to analyze the profile of the population having typical activity sequences. The extracted rules show our method effectiveness.