Research On House Price Forecasting Based On Swarm Optimization GBDT Composite Model

Housing is the foundation of human survival and a key issue of concern for people’s livelihoods.Therefore,housing price prediction has become a popular topic of research for experts and scholars at home and abroad.The application of traditional machine learning methods in housing price prediction is becoming increasingly widespread.However,their poor parameter finding performance and low operational efficiency have led to low prediction accuracy and large regression errors.In recent years,algorithms for optimizing machine learning methods have emerged one after another,especially with the emergence of various swarm optimization algorithms,which have made traditional machine learning methods exhibit better performance in predicting housing prices.Based on the above theoretical background,in the research process,the article first uses a typical machine learning based housing price prediction method for regression prediction.Through comparison of results,it is found that the most ideal method for predicting results is the Gradient Boosting Decision Tree(GBDT).Subsequently,a series of optimizations are carried out based on the GBDT method.The main research content of this article includes:(1)Four GBDT composite housing price prediction models based on swarm optimization have been proposed.Firstly,based on the GBDT method,four classic swarm optimization algorithms,Ant Colony Optimization(ACO),Grey Wolf Optimizer(GWO),Whale Optimization Algorithm(WOA)and Sparrow Search Algorithm(SSA),are combined with this method to obtain a composite housing price prediction model.Next,experiments are conducted using the Boston dataset,the results of which show that the four composite models proposed in the paper have better prediction performance than traditional GBDT,with SSA-GBDT having the highest prediction accuracy.(2)An optimized prediction model ISSA-GBDT based on sparrow search is proposed.In response to the problems existing in traditional SSA,the article adds the Tent Order in population initialization to randomize the discovery position update.During the following process,the multi-directional learning strategy and adaptive t-distribution mutation are added to obtain ISSA-GBDT(Improved Sparrow Search Algorithm,GBDT)with fusion strategy.Subsequently,empirical analysis and model comparison are conducted using the real dataset of second-hand housing prices in Taiyuan City.The experimental results show that the ISSA-GBDT proposes in the paper has better predictive performance compared to other models.