Mining Temporal Patterns in an Automotive Environment

Steffen Kempe, Rudolf Kruse

Mining frequent temporal patterns from interval-based data proved to be a valuable tool for generating knowledge in the automotive business. Many problems in our domain contain a temporal component and thus can be formulated by using interval sequences. In this paper we present three substantially different applications which can all be addressed by the same mining task: mining of frequent temporal patterns. We show that contemporary approaches for temporal pattern mining are not addressing this task sufficiently and present our algorithmic solution FSMTree. Further, we discuss the assessment of temporal rules which can be derived from the set of frequent patterns.

PDF full paper