Credit review model construction method, device and system
A technology for constructing methods and models, applied in the computer field, can solve the problems of difficult to guarantee the accuracy of evaluation results and difficult to guarantee timeliness.
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 2
[0127] see image 3 The second aspect of the embodiment of the present application shows a method for constructing a credit review model, the method comprising:
[0128] S101 acquires historical data, and divides the historical data into data sets, wherein the data sets include: a training set, and a test set;
[0129] Historical data consists of multiple modeling individuals, including credit review data, and evaluation results, where the evaluation results are converted into a computer-recognizable language, for example, "1" for passing, and "0" for failing;
[0130] The credit review data includes a large number of characteristics such as the user's basic information, credit review application information, credit report and user behavior information;
[0131] Select a part of the data from the historical data according to the purchase type and time, and split this part of the data into a training set and a test set. Two methods are used for data splitting. , and the other...
Embodiment 3
[0138] In the financial industry, as an important feature, the week usually has a greater impact on the evaluation results. For example, the contribution of Sunday to the credit review model is greater than that of Monday (working day) to the credit review model. Therefore, the week should be input into the computer as an important feature in the modeling process to participate in the construction of the model. However, in the process of obtaining historical data, the characteristics of the week are automatically generated in time format, for example: 12:31 on May 8, 2018. Obviously, the data in time format cannot reflect the contribution of the week to the credit review model, resulting in a low accuracy of the credit review model constructed.
[0139] In order to solve the above problems, the embodiment of this application shows a data conversion method. For details, please refer to Figure 4 The technical solution shown in embodiment 3 and embodiment 2 has similar steps. T...
Embodiment 4
[0146] Usually some features and useless information with poor predictive ability of credit review results. In the process of model construction, using these useless information as the characteristics of model construction will undoubtedly increase the amount of data processing by the computer, reduce the bandwidth of the system, and resources. utilization rate.
[0147] In order to solve the above problems, the embodiment of the application shows a method for determining invalid information. For details, please refer to Figure 5 :
[0148] The technical solution shown in embodiment 4 has similar steps with the technical solution shown in embodiment 3. The only difference is that in the technical solution shown in embodiment 3, the step of deleting useless features in the data set and obtaining the data to be processed include:
[0149] S1021111 Statistically predicting the response value of the prediction result in the data set, the non-response value of the prediction res...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More 


