Title: An improved multi-expression programming algorithm applied in function discovery and data prediction

Authors: Qingke Zhang; Bo Yang; Lin Wang; Jianzhang Jiang

Addresses: Shandong Provincial Key Laboratory of Network based Intelligent Computing, University of Jinan, Jinan, 250022, China ' Shandong Provincial Key Laboratory of Network based Intelligent Computing, University of Jinan, Jinan, 250022, China ' Shandong Provincial Key Laboratory of Network based Intelligent Computing, University of Jinan, Jinan, 250022, China ' Department of Information Tech, China United Telecommunications Corporation SD Branch, Jinan, 250001, China

Abstract: This paper presents an improved multi-expression programming (MEP). In the algorithm, each individual is encoded as a double-layer structure, and two-dimension space operators are introduced through two-dimension crossover and mutation. The problems of symbolic expression are defined and used as benchmarks to compare the effectiveness of proposal method against the baseline single-layer MEP. Experiments showed that our method using two-dimensional super chromosome can find the optimal solution in a short time with small population. Then the improved algorithm is applied to the prediction of 28-day cement compressive strength. Comparison with other three soft computing models, namely MEP model, neural networks (NN) model and fuzzy logic (FL) model on cement strength prediction revealed that the improved MEP model has a lower rate in RMSE and MAE. Test results demonstrate the proposed method is efficient and performed better in function discovery and data prediction.

Keywords: multi-expression programming; MEP; double-layer chromosome; prediction modelling; function discovery; data prediction; soft computing; cement strength prediction.

DOI: 10.1504/IJICT.2013.054952

International Journal of Information and Communication Technology, 2013 Vol.5 No.3/4, pp.218 - 233

Received: 07 Jan 2013
Accepted: 06 Feb 2013
Published online: 19 Dec 2013 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article