Abstract

Valuation of real estate can be difficult as the values can veer greatly between regions with many variables to consider. What this article introduces is the summary of several studies that have been done between 2016 and 2026 where Machine Learning (ML) strategies have been used to project real estate pricing. In realtime, location and local variables didn’t receive much concentrate, and most studies have focussed on regression or tree-based models such as ensemble or neural networks. Trait selection wasn’t fully carried out and involved models are hard to explain, and so, this article will aim to create a 4 part framework. 6 mixed models will be compared (Linear Regression, Random Forest, XGBoost, LightGBM, CatBoost and Graph Neural Network (GNN)). SHAP will be applied to explain the models and a LLM will be applied to supply any summary of credit values. AutoML will be applied to do quality selection. From this view, one method has been put together for the assessment of this suggested framework applying MAE and RMSE values. As, this is a further review and design paper there are no findings offered.

Keywords

How to Cite This Article

Rights