Data Segmentation with Improved K-Means Clustering Algorithm

Overview

This repository contains an implementation of the Mean-kMeans clustering algorithm, an enhanced version of the traditional kMeans algorithm. Our Mean-kMeans algorithm improves clustering performance by optimizing the iteration process and utilizing in-place mean calculations for cluster assignments, thus resulting in faster convergence.

Mean-kMeans Algorithm Details

Mean-kMeans introduces key enhancements:

• Dynamic Mean Calculation: Instead of fixed centroids, the algorithm calculates the mean of points currently assigned to each cluster during each iteration.
• Optimized Convergence: By refining the centroid selection and leveraging in-place calculations, the algorithm significantly reduces the number of iterations needed for convergence.

Installation

To run the code, ensure you have Python installed. You can create a virtual environment for better package management. Use the following commands to set up your environment:

python3 -m venv venv
source venv/bin/activate

And install numpy, matplotlib, scikit-learn, and pandas following:

pip install numpy matplotlib scikit-learn pandas

Usage

1. Clone the repository:

   git clone https://github.com/abdurrahman4127/Mean-kMeans_ICCIT23.git

   Then navigate to the repository directory

2. Run the scripts:

   • To execute the KMeans clustering algorithm:

     python normal_kmeans.py

   • To execute the Mean-kMeans clustering algorithm:

     python mean_kmeans.py

   • To visualize the Elbow method and clustering results, run:

     python elbow_method.py

Notebook is provided at 📙Mean-kMeans

Citation

This study was accepted in 2023 at the 26th International Conference on Computer and Information Technology (ICCIT). If you find our work useful, consider citing it as:

@INPROCEEDINGS{10441078,
  author={Hasan, Emam and Rahman, Md. Abdur and Shojib Talukder, MD. and Utsho, Md Farnas and Shakhan, Md. and Farid, Dewan Md.},
  booktitle={2023 26th International Conference on Computer and Information Technology (ICCIT)}, 
  title={Data Segmentation with Improved K-Means Clustering Algorithm}, 
  year={2023},
  volume={},
  number={},
  pages={1-5},
  keywords={Technological innovation;Clustering methods;Clustering algorithms;Machine learning;Partitioning algorithms;Unsupervised learning;Convergence;Leaning by Observation;Partition-based Clustering;Unsupervised Learning},
  doi={10.1109/ICCIT60459.2023.10441078}}