Stability evaluation of mining -induced goaf site based on dimensionality reduction fuzzy C-means， clustering algorithm

Fuzzy cluster analysis is one of the main techniques of unsupervised machine learning,which can be used for data analysis and modeling.Fuzzy C-Mean s( FCM) clustering algorithm obtains membership degree of sample points to all class centers by optimizing objective function to achieve the purpose of automatically cl ustering sample data. However,in facing of large data samples with complex factors influencing the assessment of goaf site stability in colliery,it is easy to fall into local saddle points.In order to solve this problem,it is proposed an improved FCM algorithm based on Principal Components Analysis( PCA) dimensionality reduction in this p aper.Combining the theory of machine learning,the improved algorithm selects 7 colliery goaf area stability influence factors to construct assessment index system. Acc ording to dimensionality reduction after initial class center of FCM model, sample information and membership parameters for dynamic optimization,the robustness and generalization ability of FCM has been improved to suit for the stability evaluation of complex factors in colliery goaf area.In this experiment, 120 working face goaf cond itions of a colliery buried under Renxing Road section and five other coal mines of Jining expressway in Shandong Province were selected as samples for goaf area sta bility influence ,the experimental results showed that sample data after PCA dimensionality reduction,first four principal components of cumulative contribution rate was 81.86%,it has better ability to interpret original sample information. After fuzzy C-means clustering clustering,the proportion of the sample set was statitically analyzed, “Stable"sections accounting for 36.67%,"Basically Stable' accounting for 35%,“Understable and instbility' 'section of 28.33% ,compared to actual stability state of goaf-co llapse area,the fuzzy C-means algorithm can effectively improve clustering accuracy. In this paper,the proposed approach demonstrates the feaibility and effectiveness in field stability assessment of goaf.