|
|
--- |
|
|
title: Credit Worthiness Risk Classification |
|
|
emoji: 🏆 |
|
|
colorFrom: blue |
|
|
colorTo: pink |
|
|
sdk: gradio |
|
|
sdk_version: 3.27.0 |
|
|
app_file: app.py |
|
|
pinned: false |
|
|
license: apache-2.0 |
|
|
--- |
|
|
|
|
|
This is a case study about creditworthiness classification where I did the whole process of building a creditworthiness classification model and app, from the data cleansing part until the deployment of the model in an application simulated for the banks managers. To achieve this goal, I analyzed and prepared the dataset for machine learning models. The applied models are: Logistic Regression, Decision Tree Classifier, and Random Forest Classifier, which are available in Python's sklearn library. To optimize the workflow and model results, I applied a personalized pipeline for model application and GridSearchCV for parameter optimization. The app development was made using gradio app. |
|
|
|
|
|
The data is uploaded in path: german_credit_risk/data/raw, but it was first obtained from Kaggle and can be obtained [HERE](https://www.kaggle.com/datasets/mpwolke/cusersmarildownloadsgermancsv). |
|
|
|
|
|
I used a previous work from Pennsylvania State University as a reference in many parts of the code, you can find it [HERE](https://online.stat.psu.edu/stat508/resource/analysis/gcd). Also, as this is a case study, the code steps are commented in english. |
|
|
|
|
|
It is worth mentioning that the feature selection process was carefully performed according to the regulations of the Central Bank of Brazil, as I'm Brazilian. |
|
|
|
|
|
--- |
|
|
|
|
|
Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference |
