https://github.com/soroushesnaashari/customer-clustering

An Unsupervised Machine Learning project
https://github.com/soroushesnaashari/customer-clustering

dbscan gmm kmeans machine-learning

Last synced: 4 months ago
JSON representation

An Unsupervised Machine Learning project

• Host: GitHub
• URL: https://github.com/soroushesnaashari/customer-clustering
• Owner: soroushesnaashari
• Created: 2024-11-25T15:23:40.000Z (11 months ago)
• Default Branch: main
• Last Pushed: 2024-12-16T13:05:48.000Z (10 months ago)
• Last Synced: 2025-04-19T20:54:19.267Z (6 months ago)
• Topics: dbscan, gmm, kmeans, machine-learning
• Language: Jupyter Notebook
• Homepage:
• Size: 6.4 MB
• Stars: 5
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
## Customer Clustering Project

[![](Image.jpg)](https://unsplash.com/photos/group-of-people-standing-in-front-of-food-stall-counter-66RxrYlPShI)

### Overview

This project applies **Unsupervised Machine Learning** techniques to a customer dataset to identify meaningful clusters. The goal is to segment customers based on their features, using some popular clustering algorithms such as **K-Means**, **DBSCAN** and **Gaussian Mixture Models (GMM)**. The clustering results are evaluated and visualized to extract actionable insights.




### Process Workflow

The project follows a structured approach for clustering:

1. **Data Cleaning:** 


Imported the dataset and handled missing values using *dropna()*.

2. **Data Visualization:** 


Explored the dataset through detailed visualizations (scatter plots, pair plots, bar charts, etc.) to understand feature relationships.

3. **Data Scaling:** 


Applied *MinMaxScaler* to normalize the dataset to the range [0, 1] for better performance with clustering algorithms.

4. **Clustering Algorithms:** 


Performed clustering using the following methods:

  - *K-Means Clustering:* Identified optimal clusters using Silhouette Score Analysis.

  - *DBSCAN:* Explored density-based clustering with parameter tuning (eps and min_samples).

  - *Gaussian Mixture Models (GMM):* Modeled clusters using probability distributions.

5. **Evaluation:** 


Used *Silhouette Score* to evaluate the clustering performance for all algorithms.

6. **Parameter Optimization:** 


Conducted parameter tuning for *K-Means*, *DBSCAN* and *GMM* to find the optimal number of clusters and density thresholds.

7. **Visualization:** 


Created various plots to interpret and compare clustering results:

  - Scatter plots with PCA reduction.

  - Heatmaps of cluster feature averages.

  - Barplot to illustrate the cluster's size.




### Results

After evaluation and parameter tuning:

  - **K-Means** provided the best clustering results with:

    - Optimal Number of Clusters: 2

    - Silhouette Score: 0.3906

  - **DBSCAN** struggled with the dataset, forming a single cluster and noise.

  - **GMM** gave reasonable results but was outperformed by K-Means in this case.




### Repository Contents

- **Data:** Contains the [Original Dataset](https://www.kaggle.com/datasets/mahnazarjmand/customer-segmentation) and you can see the cleaned dataset in notebook.

- **Notebook:** Jupyter notebook detailing the entire process, including data cleaning, visualization, clustering, and evaluation.

- **README.md:** Project documentation.




### How to Contribute

Contributions are welcome! If you'd like to improve the project or add new features:

1. **Fork the repository.**

2. **Create a new branch.**

3. **Make your changes and submit a pull request.**