Forecasting Housing Price Using GRU, LSTM and Bi-LSTM for California

This study aims to address the challenge of predicting housing prices in California using various neural network models, including LSTM, Bi-LSTM, GRU, and a hybrid GRU-LSTM. To enhance predictive accuracy, an innovative approach involving spatial sequences based on Haversine is introduced. By considering the spatial correlation of housing prices, the data is spatialized, and the results demonstrate that the collaborative integration of spatial sequences and neural networks significantly improves model prediction accuracy. This not only provides a novel approach for the precise forecasting of individual housing prices but also holds potential to offer additional support for government oversight and regulation, benefiting homebuyers in understanding the complexity of the home-buying process. The findings emphasize the importance of considering spatial factors in housing price predictions, offering a promising direction for future research and practical applications in the real estate domain.