The Prediction Of Gas Emission Based On Relevance Vector Machine And Improved Shuffled Frog Leaping Algorithm

Gas emission in working face is a key index of the mine ventilation design, and it will result in a significant impact on the safety of coal production and life safety of the miners, therefore to achieve accurate prediction gas emission volume will has important realistic significance. The relationship between the various influence factors of gas emission is dynamic and nonlinear, so the forecasting model built by the traditional linear prediction method can not meet the practical needs.On the basis of analyzing the existing gas emission prediction method, based on machine learning problems, this paper proposes the use of relevance vector machine to create the predictive model, at the same time, the improved parameters on prediction model of the leapfrog algorithm, a control system of the relevance vector machine and prediction based on modified shuffled frog leaping algorithm. Firstly, paper analyzes the basic theory of relevance vector machine, aiming at the limitation of single kernel function, proposed the use of a combination kernel function to build relevance vector machine; Leapfrog proposed algorithm using improved its parameter optimization parameters for its selection. Then analyzes the basic theory leapfrog algorithm and genetic algorithm is proposed improvements to the original leapfrog algorithm evolutionary grouping and sub-populations, namely in the traditional grouping randomly join other groups in addition to its own outside the group to one individual sub-populations of scale, improve the diversity of sub-populations, and the genetic algorithm key step in adding the child population, so that the evolution of the sub-populations to take genetic evolution process.