一种组合K近邻聚类在煤与瓦斯突出预测中的应用

Application of K nearest neighbor clustering based on the combination technology on coal and gas outburst forecasting

针对煤与瓦斯突出影响因素复杂,即不仅具有随机性,又具有模糊性。为了保证预测的准确性,采用组合聚类策略。通过建立多个k近邻聚类器,可以产生多个簇集。来自不同簇集的子簇之间必然存在交集,最后利用子簇的加权连通图合并子簇。以平顶山八煤矿煤与瓦斯突出的相关因素指标为基础,对历年的煤与瓦斯突出的数据进行聚类分析,预测结果表明,该方法具有较好的预测效果,为煤与瓦斯突出预测提供了一种新的解决方案。

Coal and gas outburst is affected by many complicated factors,which are uncertain due to their randomness and fuzziness.This paper adopts a strategy of combination clustering in order to enhance the forecast accuracy.By constructing various K nearest clustering models,a collection of clusters can be obtained.The different sub-clusters from the cluster collection must contain intersections.Therefore,the sub-clusters can be merged by a weighted connection graph.A case study at No.8 Mine at Pingdingshan was conducted to apply the forecast method developed in this paper.The clustering analysis was based on the relevant factors with coal and gas outburst.The results show that the method generates a better forecast result,and can provide a new solution in forecasting coal and gas outburst.