Data-driven decisions on the future of the housing market with explainable artificial intelligence

The advancement of artificial intelligence (AI) and machine learning (ML) techniques presents new opportunities to enhance data-driven decision-making processes in complex markets such as the real estate sector. This study explores how AI and explainable AI (XAI) techniques can be integrated to improve price predictions and understand market dynamics in the Istanbul housing market. Using a comprehensive dataset obtained from the Kaggle platform (25,154 rows and 37 columns), variables influencing housing prices in Istanbul were analyzed. In the initial phase, preprocessing steps, including data cleaning and outlier management, were performed, followed by the application of various machine learning algorithms to predict housing prices. The performance of these algorithms was evaluated using metrics such as R², MAE, and RMSE, with the Random Forest model demonstrating the highest performance. The XAI techniques employed in the study, particularly the SHAP method, enhanced the transparency of model predictions and identified the relative importance of variables affecting price predictions. The findings revealed that square meter size is the most significant variable influencing housing prices, while regional factors (e.g., infrastructure investments and transportation projects in socially advantageous areas like Kadıköy and Sarıyer) and the number of rooms also provided meaningful contributions. Additionally, the impact of factors such as government investments, urban transformation projects, and social dynamics (e.g., migrant settlements, transportation infrastructure, and social amenities) on housing prices was discussed. The contributions of infrastructure investments and migrant settlements in areas like Kadıköy and Sarıyer to price fluctuations were found to align with model predictions. Furthermore, in light of economic forecasts in the literature, it is anticipated that large-scale projects such as Kanal Istanbul could lead to long-term price fluctuations. In conclusion, this study demonstrates that integrating AI and XAI techniques enables more reliable and transparent housing price predictions, underscoring the significant potential these approaches offer for the real estate sector. Future studies could benefit from incorporating larger datasets, detailed integration of social and environmental factors, and more advanced XAI techniques to further improve housing market predictions.