The Study On Particle Swarm Optimization For Discovering The Small-disjunct Rules

This paper addresses the well-known classification problem of data mining, where the objective is to predict the class which an example belongs to. The task of classification problem is to mine a rule-set for forecasting the class of a new example by studying the data sets with class label. In essence, a set of classification rules can be regarded as logical disjunction of rules, so that each rule can be regarded as a disjunct. A small disjunct is a rule covering a small number of examples. Due to their nature, small disjuncts are error prone and unreliability. However, although each small disjunct covers just a few examples, the set of all small disjuncts can cover a large number of examples. This will reduce the accuracy of the entire classification rules set seriously, so it is important to develop new effective approaches to cope with the problem of small disjuncts.At present, the results of the research on small disjuncts are relatively few. Because the methods to define and estimate the small disjuncts are simplification and don't get a consentaneous criterion, this paper advances a new dynamic method estimating the small disjunct, which takes both example space and error rate into account, and which imports the illegibility theory. Secondly, we find that maximum generality bias is fit for the large disjunct and maximum particularity bias is fit for small disjunct. We know that Decision Tree (DT) algorithm make the Information Rate as the attribute selection criterion which belongs to maximum generality bias and Particle Swarm Optimization (PSO) make the Sensitivity and Specialty as the fitness function which belongs to maximum particularity bias. So we introduce a hybrid decision tree/particle swarm optimization (DT/PSO) algorithm for coping with the problem of small disjuncts.At last, We implement and analyze both DT/PSO and DT/GA to cope with the small disjuncts problem. The result displays the good efficiency of the hybrid DT/PSO coping the problem of small disjuncts.