🧠 Customer Lifetime Value (CLV) Prediction Dashboard

A full-stack interactive Streamlit application that predicts the future value of a customer to a business using historical transaction data.

This project was developed as part of the Celebal Internship Program and uses machine learning and data-driven segmentation to classify customer types, predict repeat business in 3 and 6 months, and visualize their purchasing patterns.

---

🚀 Live Demo

🌐 Streamlit App: Click Here to Launch

---

🚀 Features

• 🧾 Customer-Level CLV Prediction

  • Predicts how much a customer is likely to spend in the next 3 and 6 months
  • Built using XGBoost Regression

• 🔍 Customer Type Detection

  • Classifies customers as Retail or Wholesaler based on behavioral patterns

• 📊 Interactive Dashboard

  • Select a customer ID and instantly get:
    • CLV predictions
    • Country, customer type
    • Transaction history
    • Visualizations (monthly trends, product purchases, etc.)

• 📈 Visual Analytics

  • Product frequency bar charts
  • Monthly invoice trends
  • Customer type segmentation (Retail vs Wholesaler)

---

🛠️ Tech Stack

Layer	Tools Used
Frontend UI	Streamlit
ML Model	XGBoost, scikit-learn
Data Wrangling	Pandas, NumPy
Visualization	Matplotlib, Seaborn
Deployment	Streamlit Cloud / Local

---

📂 Project Structure

customer-lifetime-value-prediction/
├── app.py
├── CLV_Prediction.ipynb
├── EDA.ipynb
├── README.md
├── requirements.txt
├──data/
│   ├── online_retail_II.xlsx
│   ├── retail_cleaned_data.csv
│   └── features.csv
└── models/
   ├── clv_model_3m.pkl
   └── clv_model_6m.pkl

---

🧪 How It Works

1. Data Cleaning:

   • Canceled transactions, missing CustomerIDs, and invalid records are removed

2. Feature Engineering:

   • RFM metrics: Recency, Frequency, Monetary
   • Avg Unit Price, Avg Quantity
   • Segmentation into Retail/Wholesaler using heuristics

3. Model Training:

   • CLV prediction model trained using past customer features
   • Labels created using a split between historic and recent months

4. Dashboard & Inference:

   • User selects a Customer ID
   • Dashboard displays predictions, insights, and visuals

---

📊 Example Output

• Customer 17850
  • Country: United Kingdom
  • Type: 🏢 Wholesaler
  • Predicted Spend:
    • In 3 Months: £3450
    • In 6 Months: £5862

---