关键词: 差分自回归滑动平均(ARIMA), 模糊信息粒化(FIG), 支持向量机(SVM), 矿业安全生产态势, 区间预测

Abstract: In order to improve prediction accuracy of mining work safety situation, aiming at problems of low prediction accuracy and difficult model selection for non-stationary nonlinear time series by a single prediction model, a mining work safety situation interval prediction model based on FIG was proposed. Firstly, time series of mining work safety situation was mapped into fuzzy information granules containing three parameters: L, R and U. Then, ARIMA model was used to predict linear part of fuzzy particle sequence to obtain a nonlinear residual sequence. Finally, nonlinear residual sequence was used as an input variable to establish a SVM model, and prediction result of ARIMA model was superimposed with residual sequence prediction value of SVM model to obtain interval of mining work safety situation time series predictive value. The results show that accuracy of interval prediction model based on FIG is verified by 21 test sets of samples, average relative errors of L, R and U are 10.834 57%, 20.207 90% and 0.651 97%, respectively, fitting effect of interval prediction model of mining work safety situation based on fuzzy information granulation is better than ARIMA and SVM, with higher accuracy and reasonable interval range.

Key words: autoregressive integrated moving average(ARIMA), fuzzy information granulation(FIG), support vector machine(SVM), mining work safety situation, interval prediction