Fuzzy Association Rules Mining based on Type-2 Fuzzy Sets over Data Stream

Abstract

Fuzzy set theory aims to solve the problem of mining more valuable human reasoning knowledge from quantitative databases, and to narrow the gap between computer and human understanding. Fuzzy Frequent Itemset (FFI) mining, which is very important in fuzzy association rule mining, converts quantitative values into language, and finds fuzzy frequency itemset according to the defined membership function. In the traditional fuzzy frequent pattern mining method, only the language item with the largest scalar base is considered, and the uncertainty of the membership function is not considered. At the same time, it is mainly considered for the fixed data set. This paper proposes a Fuzzy Frequent Pattern mining algorithm based on the Type-2 Fuzzy Set (T2FS) theory of the data stream. Under the T2FS theory, the data stream is dynamically divided based on the sliding window method, and the ambiguity is quickly found from the numerical data stream. In order to reduce the search space and improve over-mining, this paper designs a dynamic compressed lists structure and trees method. Under different minimum support degrees and sliding window sizes, many experiments have been carried out, and the experiments have verified the efficiency and effectiveness of the designed method in terms of running time and memory usage. The results can also contribute to exchange the information and preferences among the people who cooperate in decision-making.