Enhancing House Price Prediction Using Hybrid Model Approach

Abstract—Accurate prediction of real estate house prices is a difficult task due to the non-linear interactions among different property attributes and market factors. This paper proposes a hybrid model integrating Extreme Gradient Boosting (XG- Boost), Artificial Neural Networks (ANN), and Deep & Cross Networks (DCN) to predict house prices with higher accuracy. All the components contribute differently: XGBoost performs well on structured tabular data, ANN identifies complicated nonlinear patterns, and DCN captures feature interactions well. The ensemble approach combines the best features of these models to reach a prediction accuracy of 96%,which is much better than conventional models. In addition, the model uses SHAP (SHapley Additive exPlanations) to offer interpretable explanations of feature contributions, improving the transparency and credibility of the model. The hybrid model is tested using R-squared, RMSE, MSE, MAPE, and visualizations of actual vs. predicted prices. SHAP summary and dependence plots provide detailed explainability by measuring the effect of each feature on predictions. The improved model can be used as a solid decision- making tool for buyers, investors, and real estate experts. Keywords— House Price Prediction, Machine Learning, XG- Boost, Artificial Neural Network, Deep Cross Network, SHAP, Hybrid model.