Implementation of the Linear Regression Method for Predicting House Prices

Abstract

House prices are an important indicator in the property sector and are influenced by 
various factors, including physical characteristics of the building and location. This study aims 
to analyze and predict house prices using a linear regression method by utilizing several 
variables, namely land area, building area, number of bedrooms, number of bathrooms, parking 
availability, and distance to the city center. The data used in this study are secondary data 
collected through a web scraping process and are focused on houses with a price range of 300
700 million rupiah to represent the middle-market segment. 
The research stages include data preprocessing, Pearson correlation analysis, 
multicollinearity testing, multiple linear regression modeling, and model performance 
evaluation using the coefficient of determination (R²) and Root Mean Square Error (RMSE). 
The dataset is divided into 80% training data and 20% testing data. The results show that the 
constructed linear regression model achieves an R² value of 0.3078, indicating that the 
independent variables are able to explain 30.78% of the variation in house prices. The RMSE 
value of 117,482,242 indicates that prediction errors remain relatively high due to the wide 
variation in house prices. 
The correlation analysis results reveal that the number of bathrooms and the distance 
to the city center have a relatively stronger relationship with house prices compared to other 
variables. This study demonstrates that linear regression can be used as an initial approach for 
house price prediction; however, it still has limitations in explaining overall price variations. 
Therefore, future research is expected to improve prediction performance by incorporating 
additional variables or applying more advanced modeling methods.