Abstract: Rice is one of the important food crops in China and is grown in different agricultural production regions in China, and it is important to achieve accurate prediction of rice yield to stabilize food security in China. In order to improve the traditional yield prediction method and achieve efficient rice yield prediction, this study firstly extracts principal components based on principal component analysis, and then uses the principal components as the input of BP neural network model to predict and analyze the data from 2011-2020 in Heilongjiang Province, Jiangsu Province, Hunan Province and Hubei Province, which have large rice cultivation areas. The results showed that rice yield was highly significantly correlated with monthly maximum soil temperature, monthly minimum soil temperature, monthly average soil temperature, monthly maximum atmospheric temperature, monthly average atmospheric temperature, monthly average atmospheric humidity, significantly correlated with monthly rainfall, and weakly correlated with monthly minimum atmospheric temperature. The prediction accuracy of rice yield under the combined model of principal component analysis and BP neural network was significantly higher than that of the traditional BP neural network model, with R~2 reaching 0.86, MAPE only 0.95%, and RMSE 1.13. The degree of fit between the predicted and experimental values was high, and the model validation results showed that the model prediction results were accurate and stable. The model validation results showed that the model prediction results were accurate and stable. The research results have important guiding significance for more scientific and reasonable rice yield prediction.