Significant Features for House Price Prediction Using Machine Learning

Predicting house prices is a creative way to comprehend actual home costs. Therefore, changes in house prices affect both buyers and sellers. It also shows the state of the housing economy at the moment. The purpose of this study is to identify the features that have the largest impact on house prices. The methods used to produce the most relevant feature combinations are Chi-Square and Mutual Information. To determine which combination of features has the greatest impact, we will feed the feature selection results into several machine learning algorithms—including Linear Regression, Logistic Regression, and Random Forest Regression—and evaluate them using Mean Absolute Error (MAE). The Linear Regression model shows that the combination of features based on the Mutual Information method with $K=20$ produces the smallest MAE, namely $23,251.43$. The Logistic Regression model shows that the combination of features based on the Mutual Information method with $K=10$ produces the smallest MAE, namely $\mathbf{3 3, 6 5 4 . 8 5}$. The Random Forest Regression model shows that the combination of features based on the Mutual Information method with $\text{K}=15$ produces the smallest MAE, namely $18,648.54$. This research shows that overall, the feature combination based on the Random Forest Regression model produces the smallest MAE with a value $18,648.54$. The 5 best features of the combination are OverallQual (Overall material and finish quality), Neighborhood (Physical locations within Ames city limits), GrLivArea (Above grade (ground) living area square feet), GarageCars (Size of garage in car capacity), and GarageArea (Size of garage in square feet).