Prediction Model for Gas Emission Quantity in Mining Face Based on Partial Least Squares Regression

Aiming at the problem that the independent variables appear multi-collinear in regression modeling process of mining face gas emission, an idea using the partial least-squares (PLS) regression technology to find prediction model of the gas emission quantity in mining face is put forward. The partial least squares regression model which regards geological and mining two respects in all 12 index as regression factor to predict gas emission quantity mining face is established by using 15 gas emission examples for modeling sample. There is a good fitting result to training sample by established regression models, that the maximum error is 6.09%, the average error is only 2.06%. The PLS model is better than principal component regression analysis method and BP neural network in the prediction of the remaining cases, and which is consistent with the least squares-support vector machines method. It is an effective and feasible method by using partial least squares regression to predict gas emission quantity of mining face.