Constraining the Search Space in Temporal Pattern Mining

Agents in dynamic environments have to deal with complex situations including various tem-poral interrelations of actions and events. Dis-covering frequent patterns in such scenes can be useful in order to create prediction rules which can be used to predict future activities or situa-tions. We present the algorithm MiTemP which learns frequent patterns based on a time interval-based relational representation. Additionally the problem has also been transfered to a pure re-lational association rule mining task which can be handled by WARMR. The two approaches are compared in a number of experiments. The ex-periments show the advantage of avoiding the creation of impossible or redundant patterns with MiTemP. While less patterns have to be explored on average with MiTemP more frequent patterns are found at an earlier refinement level.