Credit risk analysis for credit card applicants
MIT License
Banner source
Badge source
This app predicts if an applicant will be approved for a credit card or not. Each time there is a hard enquiry your credit score is affected negatively. This app predict the probability of being approved without affecting your credit score. This app can be used by applicant who wants to find out if they will be approved for a credit card without affecting their credit score.
Correlation between the features.
Confusion matrix of gradrient boosting classifier.
ROC curve of gradrient boosting classifier.
Top 3 models (with default parameters)
Model | Recall score |
---|---|
Support vector machine | 88% |
Gradient boosting | 90% |
Adaboost | 79% |
The final model used is: Gradient boosting
Metrics used: Recall
Why choose precision as metrics: Since the objective of this problem is to minimize the risk of credit default for the financial institution, the metrics to use depends on the current economical situation:
During the time of a bull market (when the economy is expending), people feel wealthy and usually are employed. Money is usually cheap and the risk of default is low. The financial institution is able to handle the risk of default therefore is not very strict on giving out credit. The financial institution can handle a number of bad clients as long as the vast majority of applicants are good clients (aka those who payback their credit).In this case, having a good recall (sensitivity) is ideal.
During a bear market (when the economy is contracting), people loose their jobs and their money through the stock market. Many people struggle to meet their financial obligations. The financial institution therefore tend to be more conservative on giving out credit or loans. The financial institution can't afford to give out credit to clients who won't be able to pay back their credit. The financial institution would rather have a smaller number of good clients even if it means that some good clients where denied credit, and ideally not have any bad client. In this case, having a good precision (specificity) is desirable.
Note: There is always a trade-off between precision and recall. Choosing the right metrics depends on the problem you are solving.
Conclusion: In our case, since we are in the longest bull market (not including the March 2020 flash crash), we will use recall as our metric.
Initialize git
git init
Clone the project
git clone https://github.com/semasuka/Credit-card-approval-prediction-classification.git
enter the project directory
cd Credit-card-approval-prediction-classification
Create a conda virtual environment and install all the packages from the environment.yml (recommended)
conda env create --prefix <env_name> --file assets/environment.yml
Activate the conda environment
conda activate <env_name>
List all the packages installed
conda list
Start the streamlit server locally
streamlit run cc_approval_pred.py
If you are having issue with streamlit, please follow this tutorial on how to set up streamlit
To explore the notebook file here
To deploy this project on streamlit share, follow these steps:
├── datasets
│ ├── GDP.csv <- the data used to feature engineering/enriched the original data.
│ ├── test.csv <- the test data.
│ ├── train.csv <- the train data.
│
│
├── assets
│ ├── confusion_matrix.png <- confusion matrix image used in the README.
│ ├── gif_streamlit.gif <- gif file used in the README.
│ ├── heatmap.png <- heatmap image used in the README.
│ ├── Income_classification.png <- banner image used in the README.
│ ├── environment.yml <- list of all the dependencies with their versions(for conda environment).
│ ├── roc.png <- ROC image used in the README.
│
├── pandas_profile_file
│ ├── income_class_profile.html <- exported panda profile html file.
│
│
├── .gitignore <- used to ignore certain folder and files that won't be commit to git.
│
│
├── Income_Classification.ipynb <- main python notebook where all the analysis and modeling are done.
│
│
├── LICENSE <- license file.
│
│
├── income_class_st.py <- file with the best model and best hyperparameter with streamlit component for rendering the interface.
│
│
├── README.md <- this readme file.
│
│
├── requirements.txt <- list of all the dependencies with their versions(used for Streamlit ).
Pull requests are welcome! For major changes, please open an issue first to discuss what you would like to change or contribute.
MIT License
Copyright (c) 2022 Stern Semasuka
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
Learn more about MIT license