Based on analysis of the monthly water consumption data from the national water resource management information system, the ARIMA model, the BP neural network model and the BP neural network model optimized by the genetic algorithm (the GA-BP neural network model) were used for monthly water consumption simulation. In the process of constructing the BP neural network model, the combined method of mean impact value algorithm (MIV) and Pearson correlation coefficients was used to screen the key influencing factors of monthly water consumption through the integration and analysis of multi-source socio-economic data. The research results show that all three models exhibit relatively high accuracy in the monthly water consumption prediction in Shaanxi Province, among which the GA-BP neural network model has the highest prediction accuracy. To further verify the impact of influencing factors on the simulation results, different methods were used to screen the influencing factors as the input of the GA-BP neural network model. The simulation results indicate that the combined method of MIV and Pearson correlation coefficients improves the selection accuracy of influencing factors and can effectively enhance the simulation performance of the GA-BP model.