Mining Fuzzy Association Rules In Medical Data

The fuzzy set theory proposed by Zadeh (L.A.Zadeh) in 1965. Fuzzy sets and fuzzy logic as a new math method has been widely studied and applied to various fields in data mining applications. Association rules mining is an important research area in data mining. The fuzzy sets and fuzzy logic introduces the concept of association rules mining, and proposed "fuzzy association rules." In the fuzzy association rules mining process, different data preprocessing methods and data mining algorithms will produce different results.The thesis is based on medical data mining. According to the characteristics of medical data, Norm-FD discretization method is proposed for fuzzy discretization of continuous data, and D-RFA algorithm is proposed for fuzzy association rule mining. The mining association rules is explained using the tone operator of the fuzzy degree words.Main tasks in the thesis are as follows:(1) Fuzzy discretization of continuous attributes:According to the characteristics of normal distribution, normal dispersion method as Norm-D algorithm is used to achieve the required results of the discrete intervals. Then the attribute values are converted to the three parameters for the membership, the partition number and the bias factor used F-Inter algorithm according to attribute value and its adjacent interval between the attribute values, which completed a fuzzy discretization of continuous attributes.(2) Redundant association rules:During the mining process of fuzzy association rules, removing the redundant association rules mining algorithm (D_RFA algorithm) is proposed according to the nature of strong association rules. In this algorithm the fuzzy implication operator and the operator were used to determine the implication degree to measure the association rules contains usefulness. This algorithm don't use the credibility of availability.(3) Fuzzy association interpretion:Subsequent Fuzzy association rules were interpreted to natural language by the tone operator of the fuzzy sets, which facilitate people's understanding of the association rules.(4) Mining the fuzzy association rules:The whole process of the fuzzy association rule mining is realized using the UCI's pima-indians-diabetes data set. The comparison between D-RFA algorithm and the traditional mining algorithms is mainly from the time spent in the case of different support and the time spent in the case of different attributes.