Research On Beijing Second-hand Housing Evaluation Model Based On LightGBM Algorithm

As the Beijing real estate industry enters the era of stock housing,second-hand housing is gradually becoming the main body of transactions.The rise of the second-hand housing market in Beijing has increased the frequency of second-hand housing transactions.Second-hand housing valuation has attracted attention as a key part of the transaction.Accurately estimating the price of second-hand housing in Beijing is of great significance to reduce transaction risks,protect the interests of buyers and sellers,and promote the healthy and stable development of real estate.Current second-hand housing valuation models include linear regression,neural networks,etc.,but there are generally problems with low efficiency and low accuracy.Therefore,scientific research is needed on the valuation model of second-hand housing in Beijing.In this study,the characteristic price theory is used as a theoretical support to construct a Beijing second-hand house valuation model based on the Light GBM algorithm,and a grid search algorithm is used to improve the model.This article reasonably collects Beijing second-hand housing data based on web crawler technology,using data cleaning technology to delete duplicate and irrelevant data,deal with outliers and missing values,and using data conversion technology to complete operations such as field addition and category conversion,finally,120,772 complete second-hand housing data were obtained.Based on the feature price theory,the time feature is incorporated,and the candidate index set is constructed from four aspects: location,building,neighborhood and time.Based on the packaging algorithm,feature selection is performed on the input data,the irrelevant and redundant variables are deleted through multiple iterations,and finally 39 features are retained to obtain the optimal feature set.Based on the Light GBM algorithm,a second-hand housing valuation model in Beijing was constructed,and a grid search algorithm was used to improve the model,and an improved Light GBM valuation model was established.In order to verify the effectiveness of the valuation model,construct a linear regression,BP neural network,random forest and XGBoost valuation model for comparative analysis.Finally,based on the data of second-hand houses in Beijing,randomly divide the data set according to 4: 1,and the model effect is comprehensively evaluated from the aspects of determination coefficient,average absolute error,and relative error percentage through ten-fold cross-validation.Through comparative analysis,it can be found that the model effects of the Light GBM valuation model and the improved Light GBM valuation model are better than the four comparative valuation models.Moreover,the prediction accuracy of the model performance of the Light GBM valuation model based on the improved grid search algorithm is higher,its determination coefficient is 0.961,the average absolute error is 0.32,and the average relative error percentage is 4.99%,and is an excellent model that can be applied to solve the problem of second-hand housing valuation in Beijing.