Trust-Score Calculator B2B

A small Flask application for B2B trust scoring that loads a trained classification model and exposes a prediction UI and API.

Repository structure

• app.py — Flask web service and prediction API
• randfor.py — Random Forest training pipeline and feature selection script
• main.py — XGBoost training script with feature selection
• templates/ui.html — front-end dashboard for uploading CSV files and viewing predictions
• data/ — dataset files used for training and testing
• models/ — serialized model artifacts and encoder metadata
• docs/ — supporting documentation and project notes

Setup

1. Create and activate a virtual environment:

python -m venv .venv
.\.venv\Scripts\Activate.ps1

2. Install dependencies:

pip install -r requirements.txt

3. Ensure the dataset is available at data/b2b_credit_scoring_dataset.csv.

Run the app

Start the Flask server:

python app.py

Open http://localhost:5000 to use the prediction UI.

API usage

• Endpoint: POST /api/predict
• Supported input:
  • multipart form file upload named file
  • JSON body with data containing a list of row objects

Example JSON payload:

{
  "data": [
    {
      "customer_id": "123",
      "company_name": "Example Inc.",
      "external_credit_rating": 5,
      "company_size": "Small"
    }
  ]
}

Training

• randfor.py trains a Random Forest model and saves the artifacts under models/.
• main.py trains an XGBoost model and saves separate artifacts under models/.

Update the dataset path in data/b2b_credit_scoring_dataset.csv if necessary.

Notes

• Do not commit local environment files or generated binaries.
• models/ and data/ are included in the project structure, but model artifacts and raw CSV files should be managed outside version control when possible.