IMCKDE algorithm: an improvement in a clustering technique based on kernel density estimation

Area of Science:

• Data Science
• Machine Learning
• Artificial Intelligence

Background:

• The increasing volume of raw, unclassified data necessitates advanced pattern-finding algorithms.
• Clustering algorithms group similar data points, aiding in data analysis and interpretation.
• MulticlusterKDE, a kernel density estimation-based clustering method, identifies centroids at data density peaks.

Purpose of the Study:

• To introduce IMCKDE, an improved clustering algorithm based on MulticlusterKDE.
• To enhance both the quality of clustering results and the computational efficiency of the algorithm.
• To address the significant computational time limitations of MulticlusterKDE with large datasets.

Main Methods:

• Development of the Improved MulticlusterKDE (IMCKDE) algorithm.
• Modification of the centroid searching mechanism within the kernel density estimation framework.
• Comparative analysis of IMCKDE against MulticlusterKDE on various datasets, focusing on performance metrics and execution time.

Main Results:

• IMCKDE demonstrates superior clustering performance compared to the original MulticlusterKDE.
• Significant reductions in computational time were achieved with IMCKDE, especially for large-scale datasets.
• The algorithm effectively identifies patterns in raw, unclassified data.

Conclusions:

• IMCKDE offers a more efficient and effective solution for clustering large, unclassified datasets.
• The improvements in IMCKDE make it a valuable tool for data mining and pattern recognition.
• This enhanced algorithm addresses the scalability issues of previous kernel density estimation-based methods.