A Novel Function Mining Algorithm Based on Attribute Reduction and Improved Gene Expression Programming

Abstract

It is very interesting and important to determine the function model for remote-sensing data. The existing statistical and artificial intelligence models still have some defects. The statistical models rely heavily on prior knowledge and cannot objectively reflect the real function model contained in the remote-sensing data. In addition, the existing artificial intelligence models can very easily fall into the local optimum and have a low efficiency for high-dimensional remote-sensing data. In this paper, we first decrease the complexity of remote-sensing data by using rough sets and propose an attribute reduction algorithm based on rough sets for remote-sensing data (ARRS-RSD). On the basis of the algorithm, this paper presents a function mining algorithm for remote-sensing data by using gene expression programming and rough sets (FMRS-ARGEP). In FMRS-ARGEP, a dynamic population generation policy and a new mutation operation based on self-adaptive rate adjustment are introduced to improve the convergence of the algorithm. The experimental results show that the proposed algorithm outperforms traditional algorithms in terms of the average running time, the number of condition attributes after reduction, the attribute reduction ratio, the average convergence speed, the number of convergences, and the R2 value of the model.

Keywords

• Gene expression programming
• attribute reduction
• function model mining
• dynamic population generation
• self-adaptive mutation