Research on Prediction of Gas Emission based on Self-organizing Data Mining in Coal Mines

In order to accurately predicting gas emission in coal mines, the complicated nonlinear characteristics of gas emission was analysis, the prediction method was put forward for gas emission based on self-organizing data mining. It was used the ternary quadratic polynomial for the local function and the original variable was used in each generation, and the minimum deviation principle was used for criteria of model selected. And then, the high-order equation of prediction was established for gas emission by self-organizing data mining method. The fitness relative error of this prediction model was ±0.03% and predictive relative error was ±1.45% to gas emission in coal mine. The results show that: self-organizing data mining method can automatically analyze non-linear relation between the gas emission and the factors, and can be establish the explicit high order equation to descript the gas emission laws, and the prediction model has enough prediction accuracy for application of actual engineering in coal mines.