🛒 Customer Purchase Behavior Analysis in Retail

This project presents a full-cycle data science solution for analyzing and deriving insights from retail transactional data using Python and Power BI. It involves structured data cleaning, EDA, customer segmentation using RFM + KMeans, Market Basket Analysis using the Apriori algorithm, and business dashboard creation. Outputs are stored in CSV, visualized with Matplotlib/Seaborn, and optionally integrated with a MySQL database or BI dashboards (Power BI/Tableau).

---

🎯 Business Objective

This project aims to derive strategic insights from customer purchase data in an e-commerce/retail environment by:

• Identify customers purchasing patterns and trends
• Segmenting customers based on behavioral metrics (Recency, Frequency, Monetary)
• Generate association rules for product bundling
• Recommend strategies for targeted marketing and inventory optimixation
• Visualize insights through an interactive Power BI dashboard

---

📦 Dataset Overview

• Source: Kaggle – Online Retail II Dataset
• Period: Dec 2009 – Dec 2011
• Size: 779,000+ records
• Columns:

Feature	Description
InvoiceNo	Unique transaction ID
StockCode	Unique product ID
Description	Product description
Quantity	Quantity purchased
InvoiceDate	Date and time of transaction
UnitPrice	Price per item
CustomerID	Unique customer ID
Country	Country of purchase

⚠️ Note: Full dataset not included due to GitHub file size limits. Sample CSVs are used for demonstration.

---

🧰 Tools and Technologies

Layer	Technology
Language	Python 3.10+
Data Handling	Pandas, NumPy
Visualization	Matplotlib, Seaborn
ML Algorithms	KMeans (Scikit-learn), Apriori (mlxtend)
Database	MySQL (via mysql-connector-python)
Notebook	Jupyter Notebook
Dashboard (opt.)	Power BI (4-page executive dashboard)

---

🧱 Project Architecture

customer_purchase_analysis/
├── data/ # Raw dataset
│ └── online_retail.csv
│
├── scripts/ # Modular ETL/ML scripts
│ ├── _init_.py
│ ├── utils.py
│ ├── data_cleaning.py
│ ├── mysql_pipeline.py
│ ├── eda_analysis.py
│ ├── rfm_segmentation.py
│ └── market_basket.py
│
├── notebooks/ # Main pipeline orchestrator
│ └── purchase_analysis.ipynb
│
├── outputs/
│ ├── data/
│ │ ├── clean_online_retail.csv
│ │ ├── rfm_segments.csv
│ │ └── association_rules.csv
│ └── figures/
│ ├── eda_fig/
│ ├── rfm_fig/
│ └── mba_fig/
│
├── logs/
│ └── process_log.log
├── Reports/
│ ├── Customer_Purchase_Analysis.pbix
│ ├── Customer_Purchase_Analysis.pdf
│ ├── Customer_Purchase_Analysis.pptx
│ ├── BI_Executive_Summary.png
│ ├── BI_Sales_Trend.png
│ ├── BI_RFM_Segments.png
│ └── BI_Market_Basket.png
├── requirements.txt
└── README.md

---

🔍 Project Workflow

📌 Step 1: Data Cleaning

• Drop duplicates
• Handle missing values (esp. CustomerID)
• Removes missing or invalid values
• Creates new columns like TotalPrice
• Logs all steps and saves cleaned file

📁 Outputs:

📄 Cleaned dataset: outputs/data/clean_online_retail.csv

📌 Step 2: MySQL Pipeline

• Insert & retrieve cleaned data into/from MySQL
• Optional for production deployment and data integration
• Handles deduplication and backup

📁 Outputs:

📌 Step 3: Exploratory Data Analysis (EDA)

• Top 10 selling products
• Monthly & daily revenue trends
• Hourly purchase patterns (peak times)
• Country-wise revenue distribution`

📁 Outputs: outputs/figures/eda_fig/

📌 Step 4: Key Performance Indicator (KPI) Calculation

• Total Revenue
• Unique Customers
• Quantity Sold
• Average Order Value
• Core KPIs (Revenue, Quantity, AOV, etc.)

📁 Outputs

📌 Step 5: RFM Segmentation

• Calculates Recency, Frequency, and Monetary values
• Removes outliers using IQR
• Scales features and applies KMeans clustering → 4 customer segments
• Uses silhouette + elbow methods to determine optimal k
• Segments labeled for business use:
  • Loyal Valuable Customers
  • Recent High-Spenders
  • Occasional Low-Spenders
  • Inactive Spenders

📁 Outputs: outputs/figures/rfm_fig/

📄 RFM: outputs/data/rfm_segments.csv

📌 Step 6: Market Basket Analysis

• Applies Apriori algorithm to find frequent itemsets
• Generates association rules (support, confidence, lift)
• Visualizes top rules (bubble chart, lift bar chart)
• Great for cross-selling & bundling strategies

📁 Outputs: outputs/figures/mba_fig/

📄 Rules: outputs/data/association_rules.csv

📊 Step 7: Power BI Dashboard Visualization

• 4 Pages:
  • Executive Summary
  • Sales Analysis
  • Customer Segments
  • Association Rules

---

📈 Example Business Insights

Insight	Value
📌 70% of revenue	Comes from top 20% of customers
🎯 Peak time	10 AM – 2 PM on weekdays
💰 Best countries	UK (80%), Germany, Netherlands
🛍️ Bundling	"Gift box set" + "Teacups" has 62% confidence
📊 Segment	4 clusters with tailored marketing strategies

---

💼 Use Cases

• 📦 Inventory planning based on top co-purchases
• 🎯 Loyalty programs for high-value customers
• 📢 Targeted email offers during peak purchase times
• 📊 Executive dashboards via Power BI (optional)

---

🤖 Results (Summary)

• Customer Segments:

  • Loyal Valuable Customers
  • Inactive Spenders
  • Occasional Low Spenders
  • Recent High Spenders

• Example Association Rule:

  • If user buys "Set of Teacups" → 62% likely to buy "Gift Wrap"

---

📊 Power BI Dashboard Overview

4 Page Executive BI Dashboard (Reports/):

Page 1: Executive Summary

• Total Revenue, Orders, Customers
• Revenue by Country
• Segment distribution (from RFM)

Page 2: Sales Trends

• Monthly/Weekly Revenue Trend
• Top Products Sold
• Peak Hour Purchases

Page 3: RFM Customer Segments

• RFM Cluster Scatter Plots
• Segment-specific KPIs

Page 4: Market Basket Rules

• Rules Table (A ➡ B)
• Top Rules by Lift
• Scatter: Confidence vs Support

---

🚀 Getting Started

Prerequisites

• Python 3.10+
• MySQL Server (optional)
• Jupyter Notebook

Installation

git clone https://github.com/Ayesha24banu/Customer-Purchase-Behaviour-Analysis-in-Retail.git
cd Customer-Purchase-Behaviour-Analysis-in-Retail
pip install -r requirements.txt

Run Notebook

jupyter notebook notebooks/purchase_analysis.ipynb

⚠️ Update your MySQL credentials inside purchase_analysis.ipynb and mysql_pipeline.py.

---

output videos:

Jupyter Notebook: https://drive.google.com/file/d/1Jip6S2ppr5XhR7zQi5dREoGPfj3NtdXh/view?usp=drive_link

Power BI: https://drive.google.com/file/d/1NLckyX9VrAv5E3ddQYDTrqJXJIr0i5D8/view?usp=drive_link

---

📝 Conclusion

• RFM segmentation helps personalize marketing and optimize offers.
• Market Basket Analysis guides product placement, bundling, and inventory management.
• Visual outputs can be used by business teams with minimal technical effort.

---

🔄 Future Enhancements

• Live segmentation using streaming data
• Recommendation engine using collaborative filtering
• Customer lifetime value prediction
• Streamlit app for business teams
• AutoML for dynamic segmentation
• NLP analysis on customer reviews
• Real-time customer segmentation pipeline
• API-based deployment via FastAPI or Flask

---

📎 Deliverables

• purchase_analysis.ipynb: Master notebook
• rfm_segments.csv: RFM clustering results
• association_rules.csv: Market basket rules results
• Visual charts in /outputs/figures
• MySQL-ready table insertions (optional)
• Power BI dashboard images in Reports/

---

🙏 Acknowledgment

Thanks to the UCI & Kaggle community for the retail dataset.

---

👤 Author

Ayesha Banu

• 🎓 M.Sc. Computer Science | 🏅 Gold Medalist
• 💼 Data Scientist | Data Analyst | Full-Stack Python Developer | GenAI Enthusiast
• 📫 LinkedIn
• Project: Customer Purchase Behavior Analysis in Retail -- 2025

---

📄 License

Distributed under the MIT License. See LICENSE file for details.