Abstract: In order to research the spatial distribution of forest fire and predict the occurrence of forest fires,we selected recorded forest fire data from 1986 to 2020 in Anning City.Using this data,we constructed a generalized linear model and ARIMA algorithm model to study the spatial distribution of forest fire fires and time series analysis of forecast and climate factors,and a comparative analysis of the two forecasting methods.The research results show that the generalized linear model is more suitable for studying fires caused by human activities,and requires a higher sample size.The ARIMA algorithm model is more suitable for studying fires caused by nature and pays more attention to changes in climate and environmental factors.The predictions of the two prediction models have reached the medium accuracy level,and there is still room for further optimization.