Streamlit Credit Score Prediction App

This project is a Streamlit application designed to predict credit scores based on user-uploaded CSV data and to scrape and analyze tweets. The application processes the data through cleaning and explainability steps, providing model predictions and SHAP values for interpretability.

Project Structure

streamlit-app
├── src
│   ├── app.py              # Main entry point for the Streamlit application
│   ├── data_clean.py       # Functions for cleaning input CSV data
│   ├── explainability.py   # Logic for model training, prediction, and SHAP values
│   ├── llm_explain.py      # Functionality for generating recommendations using a language model
│   ├── tweet.py            # Functions for scraping and analyzing tweets
├── requirements.txt        # List of dependencies required for the project
└── README.md               # Documentation for the project

Setup Instructions

1. Clone the repository:

   git clone <repository-url>
   cd streamlit-app
   

2. Install the required packages:

   pip install -r requirements.txt
   

3. Run the Streamlit application:

   streamlit run src/app.py
   

Usage Guidelines

• Credit Score Prediction:

  • Upload a CSV file containing the necessary data for credit score prediction.
  • The application will clean the data, process it through the model, and display the predicted credit scores along with SHAP values for interpretability.

• Scrape and Analyze Tweets:

  • Enter a Twitter username and the number of tweets to fetch.
  • The application will scrape the tweets and provide recommendations and evaluations based on the input tweet.

Application Functionality

• Data Cleaning: The application handles duplicates, imputes missing values, and formats the data for analysis.
• Model Training and Prediction: It utilizes ensemble models to predict credit scores based on the cleaned data.
• SHAP Values: The application generates SHAP values to explain the model's predictions, providing insights into feature importance.
• Tweet Scraping and Analysis: The application scrapes tweets from a specified user and provides recommendations and evaluations for a given tweet.

License

This project is licensed under the MIT License.