📊 Invoice Intelligence System

An end-to-end Machine Learning project that predicts freight costs and identifies potentially risky invoices using supervised learning techniques.

The project includes:

• Freight Cost Prediction
• Invoice Risk Detection
• Batch Invoice Processing
• Interactive Streamlit Dashboard
• SQLite Database Integration
• Model Evaluation and Selection

---

🚀 Live Demo

🔗 Streamlit App: https://invoice-intelligence-system-ki3twtwru9ez2fnnehutch.streamlit.app/

---

🚀 Features

1. Freight Cost Prediction

Predict freight charges based on:

• Invoice Amount
• Invoice Quantity

Supported Regression Models:

• Linear Regression
• Decision Tree Regressor
• Random Forest Regressor

The best-performing model is automatically selected and saved.

---

2. Invoice Risk Flagging

Classify invoices as:

• ✅ Low Risk
• 🚨 High Risk

Features Used:

• Invoice Quantity
• Invoice Dollars
• Freight Cost
• Total Item Quantity
• Total Item Dollars

---

3. Batch Processing

Upload CSV or Excel files and:

• Predict Freight Costs
• Detect Risky Invoices
• Calculate Risk Probability
• Download Processed Results

---

4. Interactive Dashboard

Built with Streamlit and Plotly.

Includes:

• Real-time predictions
• Interactive charts
• Risk visualizations
• Model status monitoring

---

🏗️ Project Structure

Invoice-Intelligence/
│
├── app2.py
├──Freight_Prediction/
|   ├── data_preprocessing.py
│   ├── model_evaluation.py
│   └── train.py
|
├──invoice_flagging/
|   ├── data_preprocessing.py
│   ├── model_evaluation.py
│   └── train.py
|
├── inventory.db
│
├── models/
│   ├── predict_freight_model.pkl
│   ├── predict_flag_invoice.pkl
│   └── scaler.pkl
│
├── notebooks/
│   ├── model.ipynb
│   └── invoice_flag.ipynb
│
├── requirements.txt
└── README.md

---

📋 Required Input Columns

For Batch Processing:

Column Name	Description
invoice_quantity	Invoice quantity
invoice_dollars	Invoice amount
freight	Freight charge
total_item_quantity	Total item quantity
total_item_dollars	Total item amount

---

📊 Machine Learning Pipeline

Freight Prediction

Input Features:

• Dollars
• Quantity

Target:

• Freight

Models Evaluated:

• Linear Regression
• Decision Tree Regressor
• Random Forest Regressor

Evaluation Metrics:

• MAE
• RMSE
• R² Score

---

Invoice Flagging

Input Features:

• Invoice Quantity
• Invoice Dollars
• Freight
• Total Item Quantity
• Total Item Dollars

Output:

• Risk Flag (0/1)
• Risk Probability

---

🛠 Tech Stack

Backend

• Python

Machine Learning

• Scikit-Learn
• NumPy
• Pandas

Database

• SQLite

Visualization

• Plotly

Frontend

• Streamlit

---

🎯 Future Enhancements

• XGBoost Integration
• LightGBM Models
• SHAP Explainability
• Fraud Detection Module
• MLOps Pipeline
• Docker Deployment
• CI/CD Automation
• Cloud Deployment (AWS/Azure/GCP)

---

👨‍💻 Author

Ojas Shukla

⭐ If you found this project useful, don't forget to star the repository.