Customer Segmentation and Lookalike Modeling Project

Overview

This project focuses on analyzing customer behavior using data from an e-commerce or retail dataset. The primary objectives are:

1. Exploratory Data Analysis (EDA): Understand customer profiles, transaction patterns, and product details.
2. Customer Segmentation: Group customers into distinct clusters based on spending, transaction frequency, recency, and signup year.
3. Lookalike Modeling: Identify customers similar to high-value customers based on behavioral features.

The project uses Python with libraries like Pandas, NumPy, scikit-learn, Matplotlib, and Seaborn, executed in Jupyter Notebooks.

Dataset

The dataset consists of three CSV files:

• Customers.csv: Contains customer profile information (e.g., CustomerID, SignupDate).
• Transactions.csv: Includes transaction details (e.g., TransactionID, CustomerID, ProductID, TotalValue, TransactionDate).
• Products.csv: Provides product information (e.g., ProductID).

These files are used across EDA, clustering, and lookalike modeling tasks.

Project Structure

The repository contains the following files:

• 1.Abutalha_Shaikh_EDA.pdf: Summary of exploratory data analysis findings.
• 2.Abutalha_Shaikh_EDA.ipynb: Jupyter Notebook with EDA code and visualizations.
• 3.Abutalha_Shaikh_Lookalike.csv: Output file with lookalike customer mappings for the first 20 customers.
• 4.Abutalha_Shaikh_Lookalike.ipynb: Notebook implementing the lookalike model using cosine similarity.
• 5.Abutalha_Shaikh_Clustering.pdf: Report summarizing customer segmentation results.
• 6.Abutalha_Shaikh_Clustering.ipynb: Notebook implementing KMeans clustering for customer segmentation.
• README.md: This file, providing an overview and instructions.

Methods

1. Exploratory Data Analysis (EDA)

• Objective: Uncover insights into customer demographics, transaction behaviors, and product trends.
• Key Findings:
  • High-value customers drive significant revenue.
  • Certain products are frequently purchased, indicating popularity.
  • Transaction frequency and recency vary across customers, suggesting diverse engagement levels.
• Notebook: 2.Abutalha_Shaikh_EDA.ipynb
• Report: 1.Abutalha_Shaikh_EDA.pdf

2. Customer Segmentation

• Objective: Segment customers into meaningful groups based on behavioral and demographic features.
• Features:
  • TotalSpending: Total amount spent by each customer.
  • Frequency: Number of transactions.
  • Recency: Days since the last transaction.
  • SignupYear: Year of customer signup.
• Method: KMeans clustering with 4 clusters, selected based on the Davies-Bouldin Index (DBI = 1.2528).
• Results:
  • Cluster 0: Higher spending ($4596.93 mean), moderate frequency (6.61), signed up in 2022.
  • Cluster 1: Lowest spending ($1742.44), low frequency (2.80), older purchases (141.30 days), signed up ~2022.58.
  • Cluster 2: Highest spending ($5875.09), highest frequency (7.55), recent purchases (45.97 days), signed up ~2023.45.
  • Cluster 3: Moderate spending ($3043.83), moderate frequency (4.66), signed up ~2023.71.
• Notebook: 6.Abutalha_Shaikh_Clustering.ipynb
• Report: 5.Abutalha_Shaikh_Clustering.pdf

3. Lookalike Modeling

• Objective: Identify customers similar to high-value customers for targeted marketing.
• Features:
  • total_spending: Sum of transaction values.
  • transaction_count: Number of unique transactions.
  • avg_spending: Average transaction value.
• Method: Cosine similarity to compute customer similarity scores.
• Output: Top 3 lookalike customers for the first 20 customers, with similarity scores (e.g., C0001 has lookalikes C0137, C0152, C0121 with scores ~0.999).
• Notebook: 4.Abutalha_Shaikh_Lookalike.ipynb
• Output File: 3.Abutalha_Shaikh_Lookalike.csv

Installation

To run the notebooks, ensure you have Python 3.8+ installed. Follow these steps:

1. Clone the repository:

    git clone https://github.com/38832/CustomerSegmentation.git
    cd CustomerSegmentation

2. Install dependencies:

   pip install -r requirements.txt

   Create a requirements.txt file with:

   pandas
   numpy
   scikit-learn
   matplotlib
   seaborn
   jupyter
   

3. Ensure the dataset files (Customers.csv, Transactions.csv, Products.csv) are in the same directory as the notebooks.

4. Launch Jupyter Notebook:

   jupyter notebook

Usage

1. EDA:

   • Open 2.Abutalha_Shaikh_EDA.ipynb to explore the data.
   • Run all cells to generate visualizations and statistics.
   • Refer to 1.Abutalha_Shaikh_EDA.pdf for a summary of insights.

2. Customer Segmentation:

   • Open 6.Abutalha_Shaikh_Clustering.ipynb.
   • Run the notebook to perform KMeans clustering and visualize clusters.
   • Check the cluster summary and DBI (1.2528) in the output.
   • See 5.Abutalha_Shaikh_Clustering.pdf for a detailed report.

3. Lookalike Modeling:

   • Open 4.Abutalha_Shaikh_Lookalike.ipynb.
   • Run the notebook to compute cosine similarity and generate lookalike mappings.
   • The output is saved as Lookalike.csv (see 3.Abutalha_Shaikh_Lookalike.csv for results).

Results

• EDA: Identified high-value customers and popular products, guiding segmentation and lookalike tasks.
• Segmentation: Formed 4 clusters with distinct spending and engagement patterns, enabling targeted marketing strategies.
• Lookalike Modeling: Produced high-similarity scores (often >0.99), indicating strong behavioral matches for the first 20 customers.

Future Improvements

• EDA: Incorporate additional features like product categories or customer demographics.
• Segmentation: Test other clustering algorithms (e.g., DBSCAN, hierarchical clustering) or optimize the number of clusters using silhouette scores.
• Lookalike Modeling: Include more features (e.g., product preferences) or expand to all customers.

Contact

For questions or contributions, please contact [Your Name/Email] or open an issue on the repository.