| In recent years,the national economy has been in a steady rise,but the process of urbanisation has also accelerated.As one of the three horse-drawn carriages of the domestic economy,the property market is enjoying the dividends brought by the urbanisation rate and rapid development.However,with the outbreak of the epidemic and the decline in consumer confidence,the real estate industry is also highlighted behind the rapid development of the debt crisis.Due to the characteristics of the real estate business model,it has the characteristics of "high debt,slow turnover,high cost",and the trend of diversified expansion in recent years,its business operation is more complex than other industries.According to statistics,only in 2021 there are 16 real estate companies declared bankrupt,and in 2022 there are 9 real estate companies declared bankrupt.During the three years affected by the epidemic,the chain reaction caused by the real estate thunder seriously damaged the interests of buyers and investors.In recent years,our country has issued a series of regulatory measures one after another,but some real estate companies still have higher risks.And some real estate enterprises in the financial risk warning mechanism failed to play enough attention,it is difficult to start the corresponding plan at the first time.Therefore,it is of great importance for listed real estate enterprises to study the causes of financial risks,the establishment of early warning system,and the prevention and control of financial risks.This paper collects the relevant data of Shanghai and Shenzhen listed real estate companies from 2000 to 2020,examines the relevant characteristic variables from the perspective of macro environment,financial indicators and operating conditions,forms the preliminary financial risk warning indicators,and generates the indicators containing time increment information from the vertical time latitude.At the same time,the industry comparison indicators are derived from the horizontal comparison.Through the SMOTE oversampling method,this paper solves the sample imbalance problem when applying machine learning to financial risk warning research.Secondly,based on the XGBoost algorithm,the model is synthesized under different parameters by using the stacking and voting integration method for the first time to improve the model performance.In the overall attribution analysis,it is found that the industry gap of operating profit ratio and the industry gap of ROE growth rate rank high;In addition,through the local attribution analysis of a single sample of Hengda Thunderstorm that has been confirmed,it is found that the model pays more attention to the profitability index and solvency index of China Hengda Thunderstorm,and further analysis shows that the main reason for Hengda Thunderstorm is that the profit quality drops significantly,and there are a lot of interest expenses in the unreasonable capital structure,which ultimately leads to Hengda’s difficulty in cashing,and then leads to Thunderstorm.At the same time,the core judgement standard of the model to other samples is also the index of profit.The significance of this paper is that it combines XGBoost and other methods in the field of financial risk early warning of listed property companies,which is beneficial to the application of machine learning in the field of finance. |
