Based on the data of geological hazards and rainfall in Wuhan from 2016 to 2021, the characteristics of rainfall before geological hazards in Wuhan were classified by using machine learning methods, and the impact of rainfall on different types of geological hazards was analyzed. The results show that the geological hazards in Wuhan can be basically divided into two types: slope type and ground collapse. There were 253 geological hazards from 2016 to 2021, mainly small-scale hazards. 52.9% of 134 geological disasters had heavy rainfall within 15 days. And 57 Geological hazards occurred after the 65 rainstorm days. The method of calculating effective rainfall has sound effects on the meteorological risk analysis of slope geological hazards. Four heavy rainfall models inducing geological disasters are established. Machine learning method is used to extract features respectively, and their influence is analyzed.