An optimized prediction algorithm based on XGBoost

Abstract

The real estate market is closely related to people's life. It is very important to accurately predict the future real estate price. Traditional methods are difficult to describe the nonlinear characteristics of house price prediction. XGBoost algorithm can effectively represent the nonlinear relationship in house price prediction. However, the selection of parameters determines the learning and generalization ability of XGBoost, and it is very important to determine the parameters of XGBoost. Particle swarm optimization algorithm can select the training parameters of XGBoost more quickly and accurately. Therefore, this paper studies the house price prediction based on the hybrid model of particle swarm optimization XGBoost algorithm, namely PSO-XGBoost model. Using the collected sample data of houses in Ames, Iowa, five different machine learning algorithms including PSO-XGBoost are used to predict house prices. Finally, the results of five algorithms are compared and analyzed. The experimental results show that PSO-XGBoost model has the highest prediction accuracy and the best effect, and the prediction effect of integrated learning algorithm is better than that of linear regression model.