Performance Analysis and Architecture of a Clustering Hybrid Algorithm Called FA+GA-DBSCAN Using Artificial Datasets

Area of Science:

• Data Science
• Machine Learning
• Pattern Recognition

Background:

• Density-Based Spatial Clustering of Applications with Noise (DBSCAN) is a key unsupervised pattern recognition method.
• DBSCAN requires manual tuning of MinPts and Eps parameters, impacting clustering accuracy.
• High-dimensional datasets pose challenges for traditional clustering algorithms.

Purpose of the Study:

• To present a hybrid algorithm for automated DBSCAN parameter tuning.
• To improve the performance and reduce uncertainties in clustering applications.
• To evaluate the effectiveness of the proposed method on artificial datasets.

Main Methods:

• Developed a hybrid algorithm combining Factor Analysis (FA) for dimensionality reduction and a genetic algorithm (GA) with nearest neighbor search for DBSCAN parameter optimization.
• Implemented Factor Analysis (FA) for pre-processing high-dimensional datasets.
• Utilized a genetic algorithm (GA) to automate the selection of MinPts and Eps parameters for DBSCAN.

Main Results:

• The FA+GA-DBSCAN algorithm demonstrated effective automatic grouping of datasets in two-dimensional space.
• Evaluated performance using artificial datasets, measuring precision and entropy.
• The hybrid approach showed a reduced probability of error in clustering dense datasets.

Conclusions:

• The FA+GA-DBSCAN algorithm offers an effective solution for automated DBSCAN parameter tuning.
• The integration of FA and GA significantly enhances clustering performance, particularly for dense data.
• This automated approach reduces uncertainty and improves the reliability of clustering results.