Research on the House Price Forecast Based on machine learning algorithm

Abstract

House price experiences some fluctuations every year, due to some potential factors such as location, area, facilities and so on. Housing price prediction is a significant topic of real estate, and it is beneficial for buyers to make strategy decisions about house dealing. There are many research on house price forecast, yet the current research cannot comprehensively compare and analyze the popular house price prediction approach. Constructing a model begins with pre-processing data to fill null values or remove data outliers and the categorical attribute can be shifted into required attributes by using one hot encoder methodology. This paper used the following five algorithms decision tree, random forest regression, Adaptive Boosting (AdaBoost), Gradient Boosting Decision Tree (GBDT), and extreme gradient boosting (XGBoost) this paper utilized to predict house prices and compared according to the root mean squared error. This paper found GBDT and XGBoost have more accurate prediction results compared with other algorithms. Besides, this paper found which features most affect the price of a house. In real-world applications, machine learning based housing price prediction models are utilized by banks and financial institutions to obtain better house price assessment, risk analysis and lending decisions.