AWLSSVM Gas Prediction Research Based on Chaotic Particle Swarm Optimization

Abstract

In order to improve the accuracy of mine gas concentration prediction, an improved multi-variate adaptive weighted least squares support vector machine(AWLSSVM) gas prediction model for chaotic particle swarm optimization is proposed, and multi-step prediction is realized. Firstly, the particle swarm optimization algorithm is analyzed. An improved chaotic particle swarm optimization algorithm with faster convergence speed and stronger global search ability is proposed. Aiming at the shortcomings of the weighted least squares support vector machine(WLSSVM)weight distribution, AWLSSVM is proposed according to the distribution characteristics of discrete points. Secondly, the chaotic theory is used to construct a sample set. Finally, the performance of the prediction model was evaluated by mine monitoring data. The results indicate that in the univariate prediction, the prediction of AWLSSVM is 5.3% and 6.7% higher than the least squares support vector machine and WLSSVM. The multivariate accuracy is increased by 39.3%, 49.6%, 55.9%, 59.7% and 62.5%, relative to the univariate five-step prediction.

Keywords

• gas prediction
• chaotic particle swarm optimization
• multi-variate phase space reconstruction
• least squares support vector machine
• predictive model