煤矿瓦斯涌出时序预测的自组织数据挖掘方法

为分析煤矿瓦斯涌出复杂系统时间序列预测方法，提出自组织数据挖掘(SODM)与相空间重构(PSR)相结合的预测建模方法。首先应用C-C方法计算时间序列的最佳嵌入维数和延迟时间后进行PSR；然后以二元二次方程为传递函数，以嵌入维数变量为自变量，以延迟时间后的时间序列为因变量，通过内准则确定传递函数系数和外准则选择最优传递函数，并以最优传递函数的输出为下层迭代传递函数的输入，最后获得最优复杂度预测模型。算例结果表明:该方法对煤矿瓦斯涌出量预测的相对误差为-5.751 7% ~6.049 3%，平均相对误差2.145 7%，预测结果能满足煤矿安全生产实际工程应用要求。

Research on self-organizing data mining method for time series prediction of gas emission in coal mine

In order to analyze the time series prediction method for complex system of gas emission in coal mine, a prediction modeling method was proposed which combined the self-organizing data mining (SODM) with the phase space reconstruction (PSR). Firstly, the PSR was carried out after calculating the optimal embedding dimension and delay time of time series by using the C-C method. Secondly, the coefficients of transfer function were determined through the internal criteria, and the optimal transfer function was selected through the external criteria by taking the binary quadratic equations as the transfer function, taking the embedding dimension variable as the independent variable, and taking the time series after the delay time as the dependent variable, then the output of the optimal transfer function was taken as the input of the underlying iterative transfer function. Finally, the optimal complexity prediction model was established. The case results showed that the relative error of this method for the prediction of gas emission in coal mine was -5.7517% - 6.0493%, and the average relative error was 2.1457%. The prediction results can satisfy the application requirements of work safety in practical engineering of coal mine.