Prediction of coke mechanical strength based on support vector machines

This paper reasonably analyses, selects the indexes of the blend coal and establishes the predictive models of the coke mechanical strength by the blend coal indexes. The parameters of ε-Support Vector Regression Machines and the Kernel Function are optimized by using the Mind Evolutionary Computation algorithm, and then the ε-SVR models are built to predict the mechanical strength of the coke. The paper provides a new and effective ways to set up the predictive model of the coke mechanical strength. The method is applied in some Shanxi coking enterprise, and the results show that the error of the predictive models based on EMC and ε-SVR is small, the correlation is good, and the forecast-hit probability is above 95% (the relative error within 5%).