Prediction Of Water Inrush From Coal Seam Floor Based On Data Mining Classification Technology

Water inrush from coal seam floor has restricted safety production for a long time. Many accidents of casualty and property loss are resulted from water inrush from coal seam floor. The prediction of water inrush from coal seam floor became a key question that need to be solved urgently. Water inrush from coal seam floor caused by many influence factors, to predict water inrush from coal seam floor with all selected influence factors, and get rules of coal floor water inrush, prediction of water inrush from coal seam floor was concluded as classification problem of data mining, and then the data mining classification technology was imported.A new kind of techonology of prediction of water inrush from coal seam floor was put forward.Based on data mining classification technology, pieces of water inrush information were selected, predictive modles were set up, rules of water inrush were found and a classification system was developed. First, based on the analysis of information of water inrush from coal seam floor, binary logistic regression method was used to establish a probabilistic model of water inrush from coal seam floor. In the regression model, pieces of water inrush information were set as independent variable, water inrush result was set as dependent variable, and then pieces of water inrush information were selected for the first time.Then, classification models were set up based on BP neural network and probabilistic neural network, the classification accuracy of BP neural network and probabilistic neural network models with different character combination was contrasted. Pieces of water inrush information were selected for the second time based on the classification tests of neural netwoks. Third, decision trees were built to predict water inrush from coal seam floor by C4.5 decision tree and classification and regression tree, and rules of water inrush from coal seam floor were found. A classification test showed that rules from calssification and regression tree had more classification capability than rules from C4.5 decision tree. Finally, a data mining classification system which contained probabilistic neural network and classification and regression tree method was developed to predict water inrush from coal seam floor, and the system was used for a working face of lower coal seam in northern China coal field. The result showed that the data mining classification techonology can predict water inrush from coal seam floor effectively, the data mining classification system has practical application in predicting water inrush from coal seam floor.