A Fuzzy Association Rules Mining Algorithm with Fuzzy Partitioning Optimization for Intelligent Decision Systems

Fuzzy association rule mining is one of the important problems in fuzzy machine learning. The fuzzy association rule is intended to handle quantitative data that was indecisive in the classical association rule. However, the fuzzy set of quantitative attributes and their membership functions depends on expert opinion, subjective or available. In this paper, we propose a Node-list Pre-order Size Fuzzy Frequent (NPSFF) algorithm in fuzzy association rule mining based on Node-list data structure combined with the Pre-order Size Code structure suitable for the organization of fuzzy data and the manipulation of this structure. This organization speeds up tree building and finds frequent fuzzy item sets. To increase the efficiency of the NPSFF algorithm, we performed the data preprocessing by applying affinity propagation clustering (AP) technique to specify the suitable number of clusters. Next, we convert the quantitative values of the clusters to the fuzzy values using fuzzy partitioning method. Experiment results show that NPSFF algorithm efficiency is always better than FFP Growth and MFFP algorithms in execution time and memory usage.