Usually, it is a value between 0 and 1, indicating the possibility of default by the borrower. Generally speaking, the higher the probability of default, the weaker the borrower's ability to take risks.
Therefore, banks and financial institutions can determine the loan amount, interest rate and repayment period of borrowers according to the probability of default, and strengthen risk management measures for borrowers who may default. At the same time, the probability of default can also help banks to evaluate their own risk level and business risk tolerance more accurately, and effectively prevent systemic risks.
The borrower's default probability is an important risk assessment index, and internal model is one of the commonly used methods at present. The following is a classic internal model evaluation process:
1. Data collection:
First of all, we need to collect a certain amount of historical data, including the borrower's personal and financial situation, repayment records and other information. These data need to be preprocessed, cleaned and standardized to ensure data quality.
2. Function selection:
In the feature selection stage, it is necessary to determine which data have important parameters for the default probability model. This can be achieved by exploratory data analysis, statistical methods and algorithms based on machine learning. Finally, the selected features should be able to make a certain contribution to the default probability model, while excluding those potential neural interference factors.
After data processing, feature selection is needed to screen out the feature dimension that has the greatest influence on the default prediction. Statistical methods, machine learning algorithms and domain knowledge are usually used to select features.
3. Model construction:
After feature selection is completed, a default prediction model needs to be established. The model can adopt various machine learning algorithms, statistical methods or expert systems, and be trained through training data sets.
4. Model verification:
After the training is completed, the model needs to be verified. The common method is to divide the data set into training set and test set, and test the model with the test set to evaluate the prediction accuracy and stability of the model.
5. Model application:
When the model is verified, it can be applied to new and unknown borrower data to predict its default probability. It should be pointed out that it is a complex system engineering to evaluate the borrower's default probability by using the internal model, which needs careful design and management in data processing, feature selection, model establishment and verification to maximize the accuracy and effectiveness of the model.
In practical application, using the internal model to evaluate the borrower's default possibility needs to provide security, and at the same time, it needs to strictly deal with filing, supervision and risk management. In addition, the process of model establishment and verification involves machine learning and big data technology, and there are many bottlenecks such as data privacy, security, interpretation and fairness that need to be fully considered.