Customer Segmentation with RFM Analysis

This project demonstrates a complete end-to-end workflow for customer segmentation using RFM (Recency, Frequency, Monetary) analysis and unsupervised machine learning techniques.

Dataset Description

• File: orders.parquet
• Columns:
  • id: Unique order ID
  • created_at: Timestamp of the order
  • sales_amount: Order value in MYR
  • customer_id: Unique customer identifier

Objectives

• Perform data cleaning and quality checks
• Conduct exploratory data analysis (EDA)
• Engineer RFM features at customer level
• Cluster customers with unsupervised learning (MiniBatch KMeans)
• Generate business insights and strategies per segment

Tools & Libraries Used

• pandas, numpy
• matplotlib, seaborn, plotly
• scikit-learn (MiniBatchKMeans, silhouette_score)
• datetime, warnings, resample

Project Workflow

1. Data Loading & Cleaning

• Loaded data from .parquet format
• Checked for missing values and data types
• Parsed created_at to datetime and sorted records

2. Exploratory Data Analysis (EDA)

• Visualized order volume by:
  • Quarter (Q1–Q4)
  • Month (line plot)
  • Week
  • Day of Week (Mon–Sun)
  • Hour of Day (peak hours)

3. RFM Feature Engineering

• Grouped transactions by customer_id:
  • Recency: Days since last purchase
  • Frequency: Number of orders
  • Monetary: Total spending
• Normalized features for clustering

4. Customer Segmentation (Clustering)

• Used MiniBatchKMeans for efficient clustering
• Determined best number of clusters using:
  • Elbow Method (SSE)
  • Silhouette Score
• Labeled clusters as:
  • Champions
  • Loyal Customers
  • Normal Customers

5. Cluster Interpretation

• Analyzed RFM distribution per cluster
• Created segment-level visualizations and summaries
• Proposed business strategies per segment

How to Run

# Install necessary packages
pip install pandas matplotlib seaborn scikit-learn plotly

# Launch Jupyter
jupyter notebook test.ipynb