Dynamic optimisation based fuzzy association rule mining method

摘要: Techniques of performance analysis, comprising various metrics such as accuracy, efficiency and consuming time, have been conducted to evaluate the measures properties interestingness for association rule mining method. Therefore, these combined with different parameters (partitioning points, fuzzy sets) should be analysed thoroughly balanced simultaneously enhance entire (effectiveness, accuracy efficiency) an algorithm. As a result, Most current algorithms face pressure from tradeoff parameters, which becomes even rougher when we employ it in resources data (discrete data, categorical continuous data). Specifically, serial (i.e., sequences or transactions floating point numbers), analysis sensor streaming financial medical sentimental are discrete variables, boolean (e.g., sentiment: negative positive represented ‘0’ ‘1’ separately) ‘young age’, ‘middle ‘old age’). The main difference is that sharp boundary’s problem. That is, hard decide boundary values single points partition into value groups), few solved methods. This paper aims resolve problem boundaries balance multiple performances our algorithm by developing novel dynamic optimisation (parameters metrics) based (DOFARM) proposed method can applied wide range classifying problems, classification sentiment strength (negative positive). In DOFARM method, instead partitioning use smoothly separate two consecutive partitions develop corresponding membership function generate sets original physical emotional diseases. Mainly, design dual compromise scheme: first balances better out-putting rules more widely applicable while second reduces time parameter well enhances feasibility certified theoretically experimentally. Besides, demonstrate effectiveness experiments according both synthesis real datasets.