Research on Social Risk Prediction Method Based on AHP-EWM-TOPSIS and XGBoost-BP Neural Network

With the rapid development of globalization and socio-economic factors, the prediction of social risks has become crucial for maintaining social stability and public safety. We propose a social risk prediction method based on the AHPEWM-TOPSIS comprehensive evaluation algorithm and the XGBoost-BP neural network model. Initially, data on population, defense technology, environment, education, economy, and agriculture for each country and region worldwide are collected and preprocessed for model preparation. Data preprocessing includes the deletion of missing values, outlier removal, and data dimensionality reduction. Subsequently, the AHP-EWM algorithm is utilized to determine the weights of indicators, and TOPSIS is employed to construct a social stability indicator evaluation system, followed by correlation and causality analysis. On this foundation, a social risk prediction model is built by integrating the XGBoost and BP neural network, with the XGBoost hyperparameters optimized using the GWO algorithm to enhance prediction accuracy. Additionally, the CEEMDAN algorithm is applied to mine the nonlinear information in the residuals of the XGBoost, and a BP neural network is used for residual analysis and prediction. Experimental results demonstrate that the proposed model has a high accuracy rate in predicting the social stability coefficient, providing a scientific basis for social risk assessment and early warning.