Biomass is a key indicator for evaluating the growth status of rubber forest Accurate estimation of biomass is not only crucial for maintaining the ecological balance and environmental protection of rubber forest but also directly affects the estimation of carbon storage and health status of rubber forest The rubber forest in Danzhou City Hainan Province was taken as the research object three phases of Sentinel-2 images were selected by using the Google Earth Engine GEE cloud platform the band reflectivity was extracted and the vegetation index and texture features were calculated as the basic data for the inversion Pearson correlation coefficient method random forest-based recursive feature elimination RF-RFE and Boruta feature selection methods were used to screen out the optimal feature subset Four machine learning models support vector machine random forest back propagation neural network and Bayesian neural network were used to estimate the rubber forest biomass The results showed that 1 The Boruta-BNN meth od showed the highest prediction accuracy of R²=0 73 RMSE=17 42 t hm² 2 Multi-temporal imagery as a data source can obtain better estimation results compared to single moment imagery In general the Boruta-BNN method can well reveal the relationship between remote sensing features and biomass and multi-temporal image information can help improve the estimation accuracy of rubber forest biomass These research results provide scientific basis and technical support for the protection and field management of rubber forest in Hainan Province