Abstract: To predict the height of the water flowing fractured zone of the mine roof accurately, the RFR algorithm was applied to determine main factors of the height of water flowing fractured zone in mining face. With the help of QGA, the most superior combination of two key parameters-splitting attribute value and the decision tree number-in RFR was ensured, and a prediction model based on QGA-RFR was built. The 124 groups of water flowing fractured zone measured data were brought into the QGA-RFR model for training and prediction, and a prediction result comparison was made between the model and other models like GA-RFR, RFR, BP and SVM, etc. The experimental results show that the most superior combination of two parameters for the QGA-RFR model is (5,350), the prediction error under the parameters is only 0.113 8, that compared with GA-RFR, RFR, BP and SVM model, the QGA-RFR model have a smaller value of mean absolute percentage error (0.037 63), mean square error (2.129) and the maximum relative error value (0.055 06), which reveal the QGA-RFR model have a better fitting effect.

Key words: height of water flowing fractured zone, random forest regression algorithm(RFR), quantum genetic algorithm(QGA), QGA-RFR model, support vector machine(SVM), mining face