Real Estate Price Prediction

Abstract

Real estate price predictions are critical for stakeholders, including investors and developers, because they have a considerable impact on investment decisions and market stability. In order to fill in the shortcomings in earlier approaches, this work presents a novel methodology by utilizing Deep Learning(DL) and Machine Learning (ML) techniques to improve real estate price forecast accuracy. We used the "House Prices 2023 Dataset" from Kaggle, which contains 168,000 entries of Pakistani property data. Our methodology included extensive data preparation, feature engineering, and the use of various algorithms, including Linear Regression, Gradient Boosting, Random Forest, Convolutional Neural Networks (CNN), and K-Nearest Neighbors (KNN). The models were tested using MSE, RMSE, R-squared, and accuracy. KNN outperformed the other models, with a lower RMSE of 13.79 and a higher R-squared value of 0.85, indicating improved predictive accuracy. RF also produced impressive results, with an accuracy of 80%. Handling complicated feature interactions, guaranteeing model scalability, and controlling hardware resources were all challenges that suggested possibilities for future improvement. As a result, our research offers a solid foundation for raising forecasting accuracy in fluctuations in the market and emphasizes the possibility of utilizing ML approaches for better real estate price prediction.