Some data examples of the data set.

While the traditional genetic algorithms are capable of forecasting house prices, they often suffer from premature convergence, which adversely affects the reliability of the forecasts. To address this issue, the research employs a genetic-particle swarm optimization (GA-PSO) algorithm and develops a GA-PSO-BP neural network model through the integration of the BP neural network. Building upon this foundation, the study considers several pivotal factors affecting housing prices and employs a dataset comprising 1,824 transactions of second-hand homes from 2023 to 2024, gathered from Lianjia.com, to forecast housing prices in China. This work shows that the GA-PSO-BP neural network model demonstrates exceptional forecasting performance when dealing with complex and high-dimensional data, significantly minimizing forecasting errors. The test set achieved an RMSE of 0.786 and a MAPE of 8.9%. Its effectiveness in forecasting prices of second-hand houses notably surpasses that of a BP neural network model optimized by a single algorithm. This research provides more accurate forecasts of second-hand house prices in rapidly growing urban areas such as Guangzhou, thus providing essential insights for investors contemplating real estate investment.