Prediction of residential property prices using machine learning algorithms

Abstract

Residential property prices prediction is essential for evaluating market value and identifying over-pricing or under-pricing. This study investigates the performance of various machine learning algorithms, including Decision Tree (DT), Random Forest (RF), and Multilayer Perceptron (MLP) in predicting residential property prices. The study performs exploratory data analysis and principal components analysis (PCA) to reduce the dimensionality of the variables and extract the most useful variables affecting terrace house prices in Kuala Lumpur, Malaysia. A publicly available dataset is used for training and testing the algorithms, with a 70:30 proportion after pre-processing procedures. Performance indicators such as Kappa statistics, r-squared, Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and Root Mean Squared Error (RMSE) are used to evaluate the algorithms. The results show that RF outperforms DT and MLP, achieving the highest accuracy score of 85.82%, and highest Kappa statistics of 0.8307. The study also finds that the predicted data by RF algorithm are reliable from the train set. After performing exploratory data analysis and PCA, RF-PCA demonstrated the best performance in residential property price prediction, with an r-squared value of 0.7497, the lowest values of MAE (0.6091), MAPE (19.23%), and RMSE (1.066) compared to DT-PCA and MLP-PCA.