Based on the mathematical theory of discriminant analysis method, the Fisher's discriminant analysis model is applied to select support pattern for mining roadway in fully mechanized caving face, eight indexes reflecting the selection of support pattern, that is, surrounding rock intensity, coal seam intensity, mining depth, joint and crack of surrounding rock, mining influence coefficient, top coal thickness, coal pillar width, and section area of roadway are taken as the discriminant factors of Fisher's discriminant analysis model. Discriminant functions are obtained by training a large set of expansive samples. The results show that the forecasting model of Fisher's discriminant analysis has the excellent classification performance, high prediction accuracy, and is an effective method for selecting support pattern for the mining roadway in fully mechanized caving face.