House Price Prediction Based on Machine Learning Models

Abstract

This study aims to provide accurate house price predictions using machine learning algorithms. These predictions can assist decision-makers in making informed property investments and planning. Multiple linear regression and random forest were employed to achieve this goal. First, the acquired data underwent thorough analysis, including preprocessing and visualization. Subsequently, the study employed multiple linear regression and random forest models for house price prediction and evaluated their performance. The multiple linear regression model yielded promising results with an R² score of 0.73, explaining 73% of the target variable's variance. However, it exhibited prediction errors in specific cases, suggesting potential areas for improvement. In contrast, the Random Forest model achieved a slightly lower R² score of 0.69. Nonetheless, it excelled at capturing complex nonlinear relationships. Additionally, it identified the top five key features influencing house prices: house size, number of bathrooms, number of floors, parking spaces, and air conditioning. This study highlights the potential of machine learning models for house price prediction. Future research can further enhance these models and consider other influential factors to explain house price fluctuations comprehensively. The results offer valuable applications for investors, brokers, and government planners in the real estate market.