Classification Algorithm Based On Multi Expression Programming

Data mining has become one popular of the computer applications. As an importantbranch of data mining classification algorithm has caused many scholars from different fieldsof attention in the past ten years. With the rapid expansion of the size of the database, themining complex of classification rule, the traditional classification algorithms have beenunable to satisfy the needs of the society. Evolutionary algorithm which is represented byGenetic Algorithms and Genetic Programming has become a new direction for its uniquecombination of intelligence, parallelism, uncertainty, and many other features.Gene Expression Programming is a branch of Genetic Programming. Multi ExpressionProgramming is another new outcome of Gene Expression Programming in recent years. Themain feature of Multi Expression Programming is a chromosome contains many genes, whichof a gene may be selected to represent the chromosomes. In recent years, multi-expressionprogramming has been successfully applied to many fields, but in data mining is still in itsearly stages. It is a try which Multi Expression Programming is applied to the classificationproblem.The main object of this paper is the classification algorithm based on multi-expressionprogramming. First, introduces the basic concepts of classification and the process ofmulti-expression programming development. Second, by analyzing the shortcoming of thebasic multi-expression programming in the classification, we propose an improved algorithm.The improved algorithm improves in selection strategy, fitness function and genetic operators.Last, we prepare the program and conduct the experiment according to algorithm. The resultsshow that, the improved multi expression programming classification algorithm in theclassification accuracy and convergence rate is better than the basic multi-expressionprogramming classification algorithm.Main work and innovation of this article are as follow:(1) Multi expression programming is applied to classification problems, and get betterclassification accuracy and convergence speed.(2) Analysis of the basic multi-expression programming deficiencies in the classificationproblem and proposed an improved multi-expression programming classification algorithm.