Abstract: In view of problem of differences in tendency of coal spontaneous combustion at different regions in goaf, data of measured points were classified by using division method of spontaneous combustion "three zones" in goaf based on fuzzy clustering analysis and differences of each classification under different λ-intercept matrix were analyzed. Analysis results showed that the data can be divided into different categories by different values of λ; in a certain range, original categories will be divided into smaller subclasses with increasing of value of λ; when value of λ increases, data of measured points in areas close to working face can be divided into smaller subclasses more easily and data of measured points in areas far from working face are stable, so environmental properties in goaf far from working face are relatively stable and forepart changes greatly; the categories to optimal λ calculated by F-statistics uniform with division of "three zones" in goaf.