Predicting House Prices with Machine Learning: A Comparative Study of Linear, Lasso, and Ridge Regression

Abstract

Machine Learning has seen rapid advancements this decade, marked by daily innovations in applications and algorithms. One significant application is in the real estate sector, particularly in predicting house prices—a crucial task given the continual rise in property values. This project employs Linear Regression, Lasso Regression, and Ridge Regression, utilizing the Scikit-Learn tool to explore the potential of these models. By considering various factors such as the number of bedrooms, the age of the house, proximity to transportation hubs, local schools, and shopping centers, these models aim to assist potential buyers in identifying homes that best meet their requirements. Among the tested methods, the Lasso Regression emerged as the most effective, boasting the lowest root-mean-square error of 190,124.751, which notably outperformed both Ridge and Linear Regression. This outcome not only underscores the efficacy of Lasso Regression but also highlights the overall utility of regression models in accurately predicting house prices.