HOUSE PRICES PREDICTION USING CONVOLUTIONAL NEURAL NETWORKS (CNNS)

Accurate house price prediction is essential for real estate planning, investment decisions, and market analysis. Traditional machine learning models such as linear regression, decision trees, and random forests have been widely applied but often struggle to capture complex nonlinear patterns inherent in housing datasets. This study proposes a novel approach using a 1D Convolutional Neural Network (CNN) tailored for structured tabular data to enhance prediction performance. The model was trained on a real-world dataset with ten numerical and categorical features, achieving a high R²-score of 0.96, surpassing traditional models including linear regression (0.78) and random forest (0.89). Key evaluation metrics such as RMSE (22,450.31) and MAE (16,308.19) further validate the model’s accuracy. Visualization results show strong alignment between predicted and actual values, with no signs of over fitting across 100 epochs. The proposed CNN architecture demonstrates significant potential in learning feature hierarchies and complex dependencies, offering a robust tool for datadriven real estate valuation. Future work will explore multimodal integration of image data, model explainability, and cross-regional generalization to enhance practical deployment.