An Intelligent Ensemble Architecture to Accurately Predict Housing Price for Smart Cities

Accurate housing price prediction is vital for various real estate applications. This paper presents a comprehensive study on housing price prediction of smart cities, evaluating the performance of individual models and introducing an ensemble approach. The study investigates Linear Regression, Random Forest, Gradient Boost, and XGBoost, while introducing a novel ensemble model combining Gradient Boosting and a Feedforward Neural Network. Using a diverse dataset of housing attributes, we preprocess and engineer features for improved predictive performance. To access the accuracy and explanatory capacity of the model, metrics such as mean squared error (MSE), Mean absolute error (MAE) and R-Squared are utilized. Results indicate that the ensemble model achieves predictive accuracy compared to XGBoost, demonstrating competitive MSE, high R-squared, and low MAE values. Our findings underscore the value of ensemble methods in housing price prediction. The ensemble model’s success, alongside the performance of individual models, contributes to informed decision-making for real estate professionals and policymakers. This study advances the discourse on predictive modelling within housing economics, emphasizing the efficacy of ensemble techniques in capturing complex price trends. This research paves the way for further exploration of advanced ensemble methods and feature engineering strategies, offering a foundation for accurate housing price prediction and its implications for real-world applications. The problem of accurately predicting housing prices in smart cities has persisted, stemming from the limitations of existing models to capture complex price trends and provide reliable decision-making insights. This study aims to address these limitations by rigorously evaluating the performance of individual models, proposing an innovative ensemble approach (GBR + NN), and ultimately contributing to the advancement of predictive modelling within the dynamic landscape of housing economics and urban planning. The suggested model outperforms the conventional model in terms of performance indicators. In this work, the Mean Absolute Error is 105.44, R-squared is 0.99 and Mean Squared Error is 26.63.