Periodic Frequent Pattern Mining with Consideration of Periodic Improvements for Business Promotion

Periodic frequent pattern mining (PFPM) is an emerging topic in data mining. Periodicity of a pattern expresses its regularity in the transactional database. Existing PFPM approaches do not consider periodic improvements of the itemsets. PFPM approaches require to set up maximum periodicity threshold (maxPer) based on domain knowledge. In reality, databases are available without domain knowledge. The fact leads to improper specification of maxPer threshold, which in return produce spurious patterns. Keeping view of these problems, this paper presents an efficient technique to extract qualified periodic frequent patterns with consideration of their periodic improvements. The technique also does not require setting up maxPer threshold that reduces burden of the users. The proposed algorithm is named as Progressive Periodic Frequent Pattern Mining (ProPFPM). The proposed approach introduces a new interestingness measure called progressive periodic ratio (PPR) to measure the periodic interestingness of the patterns. To minimize the pattern search space, an efficient pruning technique is also introduced in this paper. A number of experiments are performed to evaluate the performance of the proposed approach in terms of pattern generation and runtime.