dataco-supply-chain-delay-prediction

Supply Chain Delay Prediction (DataCo Dataset)

This repository contains the trained classification model, feature engineering pipelines, exploratory notebooks, and database integration schemas for predicting supply chain shipment delays.

📊 Model Performance

The predictive model was trained on the DataCo Supply Chain dataset (excluding post-shipment features to avoid data leakage):

• XGBoost Classifier (Best Model - Tuned):
  • Accuracy: 77.44%
  • ROC-AUC: 89.22%
  • Late Delivery Class Precision: 87% (high precision ensures low false-alarm rates for operations)
  • Late Delivery Class Recall: 69%
• Random Forest Classifier (Baseline):
  • Accuracy: 72.19%
  • ROC-AUC: 83.41%

Top Feature Importances (XGBoost)

1. days_for_shipping_scheduled (36.6%) - The scheduled shipping window constraint.
2. shipping_mode (33.6%) - Shipping level (First Class, Second Class, Same Day, Standard).
3. type (2.8%) - Payment type (Debit, Transfer, Cash, etc.).
4. order_hour (2.5%) - Time of day the order was placed.

---

📁 Repository Structure

├── Models/
│   ├── best_xgb_model.json           (XGBoost model structure)
│   ├── best_xgb_model.pkl            (XGBoost model pickle)
│   ├── label_encoders.pkl            (Categorical feature encoders)
│   └── scaler.pkl                    (Feature scaling parameters)
├── Scripts/
│   ├── data_cleaning.py              (Preprocessing logic)
│   ├── database_loader.py            (MySQL database bulk uploader)
│   └── model_training.py             (Machine learning pipeline)
├── Notebooks/
│   ├── 1_Data_Cleaning_EDA.ipynb     (EDA & SQL queries)
│   └── 2_Delay_Prediction_ML.ipynb   (Machine learning experiments)
├── SQL/
│   ├── schema.sql                    (MySQL 3NF schema definitions)
│   └── analysis_queries.sql          (Analytical queries)
├── requirements.txt                  (Python dependencies)
└── PowerBI_Design_Guide.md           (Power BI visualization layout design blueprint)

---

🛠️ How to Use

1. Requirements

Ensure you have the required libraries installed:

pip install -r requirements.txt

2. Loading the Model and Predicting in Python

import pickle
import pandas as pd

# Load encoders, scaler, and model
with open("Models/label_encoders.pkl", "rb") as f:
    encoders = pickle.load(f)
with open("Models/scaler.pkl", "rb") as f:
    scaler = pickle.load(f)
with open("Models/best_xgb_model.pkl", "rb") as f:
    model = pickle.load(f)

# Example: Make a prediction (ensure feature engineering matches model_training.py)
# predictions = model.predict(X_scaled)