Multi-population Black Hole Algorithm for the problem of data clustering.

Area of Science:

• Computer Science
• Data Mining
• Artificial Intelligence

Background:

• Data clustering (DC) is crucial for information retrieval, grouping similar data points.
• Traditional clustering methods face challenges, necessitating advanced optimization techniques.
• The Black Hole Algorithm (BHA) is a nature-inspired metaheuristic for optimization problems.

Purpose of the Study:

• To address the limitations of the original Black Hole Algorithm (BHA), specifically its exploration capability.
• To introduce a generalized, multi-population version of the BHA (MBHA) for improved performance.
• To evaluate the effectiveness of MBHA for data clustering (DC) tasks.

Main Methods:

• Developed a multi-population Black Hole Algorithm (MBHA), focusing on a set of best solutions rather than a single best-found solution.
• Tested the MBHA on nine benchmark test functions to assess its precision and robustness.
• Applied the MBHA to six real-world datasets from the UCL machine learning lab for data clustering evaluation.

Main Results:

• The MBHA demonstrated highly precise results and excellent robustness compared to the original BHA and other algorithms on benchmark functions.
• Achieved a high convergence rate on real-world datasets, indicating suitability for data clustering.
• Experimental outcomes confirmed the superiority of MBHA in resolving data clustering issues.

Conclusions:

• The proposed multi-population Black Hole Algorithm (MBHA) is a robust and effective optimization technique.
• MBHA significantly improves upon the original BHA, offering better exploration and convergence.
• The algorithm is well-suited for addressing complex data clustering challenges in machine learning.