📦 Number of Orders Prediction -- E-Commerce

📌 Overview

A machine learning project that predicts the number of customer orders an e-commerce store will receive on any given day using historical patterns, pricing, discounts and marketing campaign data.

🎯 Objective

• Predict daily order volumes accurately
• Help businesses optimize inventory and staffing
• Compare multiple ML models to find the best predictor
• Provide data-driven business recommendations

📂 Project Structure

orders_project/
│
├── data/
│   ├── orders_raw.csv
│   ├── orders_clean.csv
│   └── predictions.csv
│
├── outputs/
│   └── (10 charts saved here)
│
├── 01_generate_data.py
├── 02_data_cleaning.py
├── 03_eda.py
├── 04_model_training.py
└── 05_insights.py

📊 Dataset

• Type: Synthetically generated -- simulates real e-commerce patterns
• Size: 1095 rows -- 3 years of daily data (2021-2023)
• Target Column: num_orders -- total orders received per day

🛠️ Libraries Used

Library	Purpose
pandas	Data manipulation and feature engineering
numpy	Numerical computing and data generation
matplotlib	All charts and visualizations
seaborn	Correlation heatmaps
scikit-learn	Linear Regression, Random Forest, metrics
xgboost	XGBoost gradient boosting model
pickle	Save and load trained models

⚙️ Feature Engineering

Feature	Description
lag_1	Orders from previous day
lag_7	Orders from same day last week
rolling_avg_7	7-day moving average of orders
is_month_start	First 5 days of month flag
is_month_end	Last 5 days of month flag
quarter	Business quarter (Q1-Q4)
week_of_year	Week number (1-52)

🤖 Models Compared

Model	Description
Linear Regression	Baseline -- fits a straight line through data
Random Forest	Ensemble of 100 decision trees
XGBoost	Sequential gradient boosting -- best performer

📈 Key Findings

• XGBoost achieved the best accuracy with lowest RMSE
• Marketing campaigns boosted daily orders by 25-40%
• Weekend orders were 15-20% higher than weekdays
• October to December was the busiest quarter (festive season)
• lag_1 (yesterday's orders) was the single most important feature

💡 Business Recommendations

1. Run campaigns in February-May to lift slow months
2. Increase delivery staff on Fridays for weekend rush
3. Complete inventory buildup by September for festive season
4. Use targeted discounts only on Tuesday and Wednesday
5. Deploy XGBoost every Monday to forecast the coming week

🚀 How to Run

pip install pandas numpy matplotlib seaborn scikit-learn xgboost
 
python 01_generate_data.py
python 02_data_cleaning.py
python 03_eda.py
python 04_model_training.py
python 05_insights.py