Abstract: In view of the problems of difficulty in data integration, weak structural modeling and low processing efficiency in the data assessment of existing methods, this paper proposes a comprehensive business environment assessment model based on deep learning (DBEAM). First, descriptive statistics such as skew coefficients and kurtosis were used, combined with the seasonal smoothing adjustment algorithm SSA and distance correlation analysis DCor, and the structural characteristics in business data were extracted. Secondly, the multi-p value evaluation protection MEP algorithm is designed to screen significant variables and achieve feature normalization through the scale unified algorithm DSH. Finally, a deep evaluation model DBEAM, which introduces self-attention mechanism and Dropout strategy, is constructed to evaluate the overall score of the business environment. The experimental results show that this model is significantly better than existing methods such as CEDML, DDS and DEA-T in multiple indicators such as accuracy, F1-score, and G-mean, and has good adaptability and generalization capabilities.