Application of fixed-structure genetic programming for classification

  • Authors:

  • Xiaojun Wu;Yue Ma

  • Affiliations:

  • Key Laboratory of Modern Teaching Technology, Shaanxi Normal University, Ministry of Education, Xi'an, Shaanxi, China, School of Computer Science, Shaanxi Normal University, Xi'an, Shaanxi, China;School of Computer Science, Shaanxi Normal University, Xi'an, Shaanxi, China

  • Venue:
  • ICIRA'12 Proceedings of the 5th international conference on Intelligent Robotics and Applications - Volume Part I
  • Year:
  • 2012

Quantified Score

Hi-index 0.00

Visualization

Abstract

There are three improvements based on GP algorithm in this paper and a fixed-structure GP algorithm for classification was proposed. Traditional GP algorithm relies on non-fixed-length tree structure to describe the classification problems. This algorithm uses a fixed-length linear structure instead of the traditional structure and optimizes the leaf nodes' coefficients based on the hill-climbing algorithm. Meanwhile, aiming at the samples on the classification boundaries, an optimization method of classification boundaries is proposed which makes the classification boundaries continuously tend to the optimal solutions in the program evolution process. At the end, an experiment is made by using this improved algorithm and a two- categories sample set with classification boundary is correctly classified (This sample set is an accurate data set from UCI database) Then it shows the analysis of classification results and the classification model produced by this algorithm. The experimental results indicates that the GP classification algorithm with fixed structure could improve the classification accuracy rate and accelerate the solutions' convergence speed, which is of great significance in the practical application of classification systems based on GP algorithm.