A Comparative Study of Machine Learning-Based Predictive Models for House Price Prediction

Abstract

Predicting house prices is vital in helping homebuyers, investors, real estate agencies, and policymakers make informed decisions. But it's hard to precisely value a house when there are a number of factors at play – from property details to location to market conditions. The present study accounts for the comparative analysis of four machine learning techniques Multiple Linear Regression (MLR), Support Vector Regression (SVR), Feed Forward Neural Network (FFNN) and Extreme Gradient Boosting (XGBoost)) on predicting house prices by applying the Boston Housing dataset. Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Mean Squared Error (MSE), and Coefficient of Determination (R 2 ) were used to assess the models. Experimental results showed that XGBoost had the lowest RMSE (3.9147), MAE (2.8454), MSE (15.3252) and the highest R 2 value (0.7910) among all the models. The second best result was obtained by SVR while FFNN gave an accurate prediction and MLR gave the least accurate result. The analysis results show that the XGBoost model can better capture the complex nonlinear relationship between house attributes and makes it significantly better than the traditional regression model and neural network model. In conclusion, XGBoost is a powerful and stable model that can be used to accurately predict house prices and value real estate.