基于SPSS多元回归分析的回采工作面瓦斯涌出量预测

Forecast of the gas emission quantity of the working face based on the SPSS multiple regression analysis

采用多元回归分析原理及模型,结合回采工作面瓦斯涌出量的实测数据,利用SPSS(Statistical Product and Service Solutions)软件对回采工作面瓦斯涌出量的影响因素进行多元回归分析,建立了回归方程来预测回采工作面瓦斯涌出量。结果表明,利用SPSS软件直接对影响回采工作面瓦斯涌出量的因素进行回归分析,避免了复杂的推导与计算,预测精度较高。

In order to forecast the gas emission quantity of the work- ing face in a coal mine, this paper, in combination with the measured data of the gas emits from the working face, has established a multi- variate regression equation by means of the statistical products and service solutions （SPSS） software. When doing the SPSS regression analysis, we have first of all adopted the method of all variables into the regression compulsively. Although the regression effect of the re- sult is nice enough, there are still many insignificant partial regression coefficients, which suggest that no significant effects were made on their corresponding variables. Besides, the large variance inflation factor （VIF） proves that multiple collinear exists among the indepen- dent variables, which implies little practical significance for introduc- ing such variables into the model. To overcome the problem of collincar and simplify the model, we have to adopt the method of stepwise regression analysis. All of these have made us realize that the model we have to choose should be the one of mining intensity, pushing speed and interlayer lithology as the predictor variables in- stead of other variables, though the regression and the regression co- efficients are perfect. Thus, using the established model, we have to predict the amount of gas emission from the working face of 15, 16, 17, 18 with the maximum relative error being 7.24% and the small- est relative error^only 1.24% and the average error^4.587%. Compared with the principal component analysis model, the gray cor- relation analysis model, the fuzzy comprehensive evaluation model and the neural network model, it can be concluded that the accuracy of this paper is strongly beneficial, and, therefore, this method can help to forecast the amount of gas effectively by avoiding complicated derivation and calculation. Thus, our method can lead to an intuitive and precise result, which can significantly reduce the calculation time and convenient in use.