Harmonious Genetic Clustering.

Area of Science:

• Computational intelligence
• Data mining
• Bio-inspired algorithms

Background:

• Automated clustering requires determining the number of clusters and ensuring cluster quality.
• Existing genetic clustering methods primarily rely on fitness, potentially limiting their effectiveness.
• Eugenic theory principles offer a framework for optimizing population dynamics and selection.

Purpose of the Study:

• To propose a novel harmonious genetic clustering algorithm (HGCA) for automated cluster number determination and enhanced cluster quality.
• To integrate harmonious mating principles from eugenic theory into a genetic clustering framework.
• To improve upon existing genetic clustering methods by incorporating multiple chromosome attributes.

Main Methods:

• Developed HGCA, a genetic clustering algorithm inspired by harmonious mating in eugenic theory.
• Incorporated chromosome gender, age, and fitness into mating attractiveness calculations.
• Designed three mating prohibition schemes (none, lineal, collateral) and three mating strategies (greedy, weighted bipartite matching, unweighted bipartite matching).
• Introduced a variable-length-and-gender-balance crossover operator for population dynamics control.

Main Results:

• HGCA demonstrated superior performance compared to existing genetic clustering methods.
• The algorithm showed improvements in robustness, efficiency, and effectiveness across real-life and artificial datasets.
• The novel crossover operator successfully balanced population gender ratio and chromosome length dynamics.

Conclusions:

• HGCA offers a robust and effective approach for automated data clustering.
• The integration of eugenic principles, particularly harmonious mating, enhances genetic clustering algorithms.
• The proposed methods provide a significant advancement in the field of genetic clustering.