If you have any questions, please contact the author at anyaxiao0604@outlook.com.
Nowadays, credit risk modelling is widely used in financial institutions. However, no software applications have been developed for loan managers to enhance their understanding and help them utilise the model without statistical analysis background. This project aims to bridge the gap by implementing an application to help loan managers analyse the risk of default and predict potential losses, specifically focusing on default rates and the age-period-cohort (APC) model efficiently predicts default rates. Firstly, age-period-cohort models are implemented to analyse separate effects of default. Then, the tool visualises the default rate and lexis diagram to enhance understanding of the output of dynamic survival model for credit risk management. This project uses two simulation data set and one Freddie Mac mortgage data set to validate the model.
The application should be installed and run locally. The following requirement should be satisfied to ensure a smooth experience with the application. Because of size limitations, the default dataset is not provided along with the source code, which may result in some problems when executing the application. Please do not hesitate to ask the author for the database if you want to try the application.
No specific hardware is required for this project. Any computer support Python and Java could run the application.
Python version 3.10.8 or above is recommended. To install external dependencies, navigate to the root directory and enter the command:
pip install -r requirements.txtThe dependencies listed below are used in this project:
| Dependencies | Version |
|---|---|
| Django | 4.1.3 |
| djangorestframework | 3.14.0 |
| djangorestframework_simplejwt | 5.2.2 |
| drf_yasg | 1.21.4 |
| jsonschema | 4.17.3 |
| psycopg2 | 2.9.5 |
| requests | 2.28.1 |
| rest_framework_simplejwt | 0.0.2 |
| matplotlib | 3.6.2 |
| numpy | 1.23.5 |
| pandas | 1.5.2 |
| plotly | 5.11.0 |
| python_dateutil | 2.8.2 |
| scikit_learn | 1.2.2 |
Java JDK version 19.0.1 is required. To install external dependencies, navigate to the root directory and enter the command:
mvn installAlternatively, use Java IDEs such as IntelliJ IDEA and open the project root directory, following online instructions to install dependencies.
After installing the application and required packages, navigate to demo/src/backend/mysite and enter the command python manage.py runserver. The server is run successfully when the command line gives the output below:
\src\backend\mysite> python manage.py runserver
Watching for file changes with StatReloader
Performing system checks...
Register successfully
System check identified no issues (0 silenced).
April 09, 2023 - 11:32:14
Django version 4.1.3, using settings 'mysite.settings'
Starting development server at http://127.0.0.1:8000/
Quit the server with CTRL-BREAK.To start the JavaFX application, the simplest way is to use Java IDEs like IntelliJ IDEA, navigate to src/interface/java/com/example/demo/App.java, and run the App class. For example, click the green "Play" icon.
As shown in the figure below, enter your email and password and click log in to access full functions. You can use the test account here:
- username: test@gmail.com
- password: test123456
Default dataset simDTS1 will be used to initialise the graphs.
The default rate, lexis diagram and APC analysis are provided.
By default, the visualisation of the default rate before the current age and calendar time is provided. Real data calculate the default rate, while the expected default rate is the default rate predicted using discrete survival models. The predicted default rate will also be visualised by clicking predicted, as presented in the figures below.
For the lexis diagram, 9 themes are provided. You could simply choose the theme to switch from each other.
For APC analysis, the age effect, period effect and cohort effect are analysed and visualised.
To upload personalised data, click the upload button in the dashboard, and the database information page will pop up as shown below. The default information is provided as shown in the graphs below. Once the related information is finished, click Choose File to find the file in the chooser. After choosing the file, the application will process and visualise the data.
To filter the data according to a specific condition, enter filter, condition and value and click go. The data that satisfy the provided condition will be used for visualisation.
Slope adjustment is to adjust the slope of APC analysis. Entering the slope and clicking adjust, the age, period and cohort effect diagram will be updated accordingly. The slope of every segment will be adjusted according to the provided information.
Scale, including sinh and exp scales, is provided to enhance the understanding of unusual data distribution. The default rate and APC effect will be scaled.
Language changing is provided at the bottom right of the log-in and dashboard page. To access more accessibility functions, click the settings button. Colour vision deficiency, changing font size and dark mode is supported in this stage.
Documentation is accessible following the guidance below.
- Open
src/docs/API documentation.mdwith a markdown renderer such as Visual Studio Code. - Click API documentation. You can ask the author for accessibility.
A screenshot of API documentation is presented in Figure 1.
Figure 1: Screenshot of API documentation
- Open
src/docs/frontenddoc/index.htmlwith any browser. - Click Java documentation.
Figure 2 presents a screenshot of Java documentation.
Figure 2: Screenshot of Java documentation
- Open
src/docs/visualizationdocs/index.htmlfor visualization documentation andsrc/docs/backenddoc/index.htmlfor backend documentation with any browser. - Click visualization documentation for visualization documentation and backend documentation for backend documentation.
Figure 3 presents a screenshot of Python documentation.
Figure 3: Screenshot of Python documentation