Prediction of House Price Index Based on Machine Learning Methods

Abstract

A house price index (HPI) is significant for people to receive accurate information such as banks, financial departments, real estate industry investors, and home owners. Data from Kaggle website by using neural network and regression models, such as linear, ridge Lasso regression. There are 99326 samples. We explore the relationships between factors of frequency, HPI flavor, HPI type, HPI index, level, period, place id, place name, and the year of houses sold or rent. In terms of the accuracy of the prediction, the accuracy of BP neural network is slightly better since the value is smaller than other two regression prediction both on the training set and testing set. However, better model like XGBoost could be chosen to improve the prediction result. Since an international concern about house prices raises recently, the precise calculation of house prices is important as well.