The study addresses the accounting and financial risk issues under the housing construction financing model, uses the decision tree algorithm to process the accounting data, and establishes the decision tree accounting data mining model, and adopts fuzzy correction processing to realize the data mining measurement. The XGBoost algorithm is introduced to establish a financial risk early warning model, and its hyperparameters are optimized with early warning accuracy and early warning efficiency as the optimization objectives, and multiple dimensional financial indicators are selected as the early warning features. Meanwhile, the prediction effects of different early warning models under different indicator sets are compared, and it is found that the XGBoost model performs optimally in each indicator, with accuracy, false positive rate, recall and precision of 0.9843, 0.1254, 0.9821 and 0.8921, respectively. Finally, by using the SHAP additive explanation algorithm, the key financial indicators that have an impact on the financial risk warning results were extracted. Indicators such as “asset return rate” and “investment return rate” have a positive impact on the financial situation of housing construction fund raising, while “total asset turnover rate” and “asset-liability ratio” have a negative impact on the financial situation of housing construction fund raising.
