Residential real estate price prediction based on adaptive loss function and feature embedding optimization

Abstract As a crucial sector underpinning national economic growth and social stability, accurately predicting real estate prices holds immense significance in directing regional development planning and refining resource allocation. To align with the dynamic nature of the real estate market, we propose a real estate price prediction model that leverages an adaptive loss function and optimizes feature embedding. Initially, we utilize diverse real estate factors to develop a representation method for real estate prices, rooted in feature embedding optimization, to thoroughly examine the interconnections among these factors. Subsequently, we introduce a reinforcement learning approach incorporating an adaptive loss function to emphasize the significance of each factor and facilitate accurate price predictions. Experimental results demonstrate that our method achieves the state-of-the-art performance, providing a robust data foundation for the real estate market, which enhances price forecasting accuracy, benefiting investors, developers, and policymakers by improving market analysis and investment decision-making This study advances the field of real estate price prediction by offering a novel approach to dynamic factor weighting. Our findings can support urban planners and government agencies in formulating more effective housing policies and resource allocation strategies.