Partial Data Clustering - PHCL-LR (Partial Data Hierarchical Clustering based on DBSCAN and Logistic Regression)

PHCL-LR (Partial Data Hierarchical Clustering based on DBSCAN and Logistic Regression) is a novel algorithm developed to address the challenge of classifying partial or incomplete data with high accuracy. This ongoing project combines three powerful techniques—DBSCAN (Density-Based Spatial Clustering of Applications with Noise), hierarchical clustering, and logistic regression—to create an improved method for data classification.

Algorithm Overview

PHCL-LR integrates the strengths of each technique:

• DBSCAN: Used for identifying dense regions in data, DBSCAN helps to find core clusters while handling noise effectively.
• Hierarchical Clustering: This technique is employed to form a hierarchical structure of clusters, providing a flexible and interpretable framework for data analysis.
• Logistic Regression: Logistic regression is then applied to classify data points based on the hierarchical structure and the clusters formed.

Key Features:

• Classification of Partial Data: The algorithm is specifically designed to handle incomplete datasets, which is a common challenge in real-world applications.
• Hybrid Approach: Combines clustering and classification techniques for better performance.
• Ongoing Research: This project is part of a Doctorate research and is being developed in collaboration with researchers. The algorithm is still evolving and subject to changes as the research progresses.

Project Structure

The project is organized as follows:

• hierarchical_clustering.ipynb: Contains the Jupyter notebook that implements hierarchical clustering on the dataset. It demonstrates how the algorithm forms clusters from partial data.
• main_werk.ipynb: A notebook focused on integrating DBSCAN and logistic regression with hierarchical clustering to demonstrate the full functionality of the PHCL-LR algorithm.

Installation

To use this project locally, follow the steps below:

1. Clone the repository:

   git clone https://github.com/your-username/Partial-Data-Clustering.git

2. Navigate to the project directory:

   cd Partial-Data-Clustering

3. Install the required dependencies:

   pip install -r requirements.txt

4. Open the Jupyter notebooks to explore the algorithm and its implementation:

   jupyter notebook hierarchical_clustering.ipynb
   jupyter notebook main_werk.ipynb

Usage

Once the project is set up, you can experiment with the provided notebooks to explore how the PHCL-LR algorithm works. The steps are outlined in the notebooks to show:

• Clustering Data: Using DBSCAN and hierarchical clustering to create clusters.
• Classifying Data: Applying logistic regression to classify data points within the clusters.
• Handling Partial Data: Demonstrating how the algorithm handles incomplete datasets.

Example

In the main_werk.ipynb, you will find examples of applying PHCL-LR to real-world datasets, including:

• Loading a dataset with missing or incomplete values.
• Applying the algorithm to form clusters and classify data.
• Visualizing the clusters and classification results.

Research and Future Work

This project is an ongoing effort as part of a Doctorate research in collaboration with fellow researchers. The algorithm is still evolving, and improvements are being made to enhance its accuracy and scalability. Future work includes:

• Improving the robustness of the algorithm on more complex datasets.
• Expanding the range of datasets that can be classified.
• Fine-tuning the model for better performance on real-world partial data.

Contributions

Contributions to this project are welcome! Please feel free to open an issue or submit a pull request if you have any suggestions for improvement or new features.

License

This project is licensed under the MIT License. See the LICENSE file for more information.