Gradual Trends in Fuzzy Sequential Patterns

Céline Fiot, Florent Masseglia, Anne Laurent, Maguelonne Teisseire

Fuzzy sequential pattern mining is a relevant approach when dealing with temporally annotated numerical data since it allows discovering frequent sequences embedded in the records. However, such patterns, in their current form, do not allow extracting another kind of knowledge that is typical of sequential data: temporal tendencies. Thanks to a relevant use of fuzzy sequential patterns, we propose the GraSP algorithm that discovers gradual trends in sequences. Our proposal is validated through experiments on web access logs.

PDF full paper