House Price Prediction and Analysis Based on Random Forest and XGBoost Models

Abstract

Accurate prediction of house price is important in housing market. It’s difficult to forecast housing price because it’s influenced by many factors. There has been many discussions on housing  price prediction by kinds of machine learning algorithm. This paper attempt to predict housing price by Random Forest and XGBoost models, and compares the performance between them. In this paper, missing values processing, correlation analysis and standardization of samples are carried on the initial data at first, then two machine learning models are constructed, trained and test on the same dataset. The Kaggle house price dataset (“House Prices-Advanced Regression Techniques”) is used in the paper. The dataset contains 1420 samples with 79 features that represent almost all characteristics of house. The results show that XGBoost algorithm achieve higher R square score of 89% , indicating that XGBoost model can be more  efficient and accurate on measurement and prediction of housing sales prices.