Track: Financial Engineering

Abstract

House rent is a crucial factor in any country’s socio-economic scenario. This can act as an indicator of the financial and developmental situations of the stakeholders of the real estate business. So far, there have been approaches to predict house rent prices in several regions of Bangladesh, but without any clear explanation of the association of different factors to the prediction and how they are affecting the prices. This study touches on this lack of understanding and proposes an Explainable AI based framework. This framework produces accurate predictions on house rents with small margin of error with regression algorithm and can accurately depict the connection of various demographic features of the dataset by visualizing SHAP values. Our study finds that tree-based algorithms such as Decision Tree, Random Forest, XGBoost, Gradient Boost and Light Gradient Boost performed better at regression analysis on the nonlinear dataset of ours. The final model was a voting ensemble of all the tree-based algorithms, which encompasses the strengths of all the base models. We achieved an MAPE of 11% and R2 Score of 86%. The RMSE and MAE of the ensemble were 4398.23 and 2536.77 respectively. Finally, the SHAP Explainable AI determined how the features were correlated to the prediction and overall rent prices. The research introduces a novel framework for predicting house rents in Bangladesh, offering valuable insights through data preparation, model selection, performance assessments, and interpretability analyses, benefiting both scholars and stakeholders.