bakytius/freight-profitability-app

🚛 Freight Profitability Predictor

A Streamlit web application that predicts whether a freight load will be profitable, powered by an XGBoost machine learning model.

Features

• —Automatic Mileage Calculation — Enter pickup and delivery cities/states and the app automatically geocodes them using OpenStreetMap (via geopy) and estimates road miles.
• —Profitability Prediction — Uses a trained XGBoost model to classify a load as profitable or not, along with a probability score.
• —Load Details — Supports Dry Van, Reefer, and Flatbed load types.
• —Input Summary — Expandable section shows the exact feature values passed to the model.

Project Structure

freight-profitability-app/
├── app.py                  # Main Streamlit application
├── xgboost_model.pkl       # Trained XGBoost model (pipeline with preprocessing)
├── requirements.txt        # Python dependencies
├── inspect_model.py        # Utility to inspect model pipeline/features
├── check_imports.py        # Utility to verify installed packages
└── .venv/                  # Python virtual environment

Requirements

• —Python 3.9+
• —Dependencies listed in requirements.txt:
• —streamlit
• —pandas
• —scikit-learn==1.6.1
• —xgboost
• —joblib
• —geopy

Setup & Installation

1. 1.Clone or download the repository.

1. 1.Create and activate a virtual environment:

   python -m venv .venv
   source .venv/bin/activate   # On Windows: .venv\Scripts\activate

1. 1.Install dependencies:

   pip install -r requirements.txt

1. 1.Ensure the model file is present: Make sure xgboost_model.pkl is in the root directory.

Running the App

streamlit run app.py

Then open http://localhost:8501 in your browser.

How to Use

1. 1.Enter Pickup Location — Type a city/state (e.g., Dallas, TX). Mileage is auto-calculated when both locations are filled.
2. 2.Enter Delivery Location — Type a city/state (e.g., Chicago, IL).
3. 3.Select Load Type — Choose from Dry Van, Reefer, or Flatbed.
4. 4.Select Pickup Date — Used to extract month and day-of-week features.
5. 5.Enter Rate ($) — The total freight rate offered.
6. 6.Verify / Adjust Miles — Auto-filled from geocoding; can be manually overridden.
7. 7.Enter Weight (lbs) — Cargo weight.
8. 8.Click "🚀 Predict Profitability" — View the prediction and probability score.

Model Input Features

Notes

• —Geocoding uses the Nominatim service (OpenStreetMap). Road distance is estimated as geodesic distance × 1.17 to approximate road curvature.
• —The model file (xgboost_model.pkl) must be a scikit-learn Pipeline that handles categorical encoding internally.
• —State codes are extracted automatically from the City, State format (e.g., Houston, TX → TX).