To find out the key financial indicators of risks in the business process, a financial early warning model based on improved decision tree is proposed. The scheme improves the traditional ID3 algorithm and establishes a simplified financial index system. A decision tree model is constructed on multiple training subsets as a basic learner. Then the formal gain criterion is used to rank and reduce the dimensions of the financial indicators of the enterprise, thus avoiding the impact of redundant information. Finally, data mining and analysis technology is applied to establish the financial budget and early warning model. The experimental analysis results show that the financial early warning model based on the decision tree established in this paper has a high accuracy, and the overall accuracy of early warning reaches 75%. It can judge financial failures and financial risks, so that enterprises can do a good job in prevention as early as possible to avoid financial crisis.