Enhanced Polar Lights Optimization with Cryptobiosis and Differential Evolution for Global Optimization and Feature...

Area of Science:

• Computational Intelligence
• Optimization Algorithms

Background:

• Optimization algorithms are vital for complex problem-solving.
• Feature selection (FS) is a critical preprocessing step in machine learning.

Purpose of the Study:

• To introduce an improved optimization algorithm, CPLODE, enhancing global optimization and feature selection.
• To integrate cryptobiosis and differential evolution (DE) into the Polar Lights Optimization (PLO) algorithm.

Main Methods:

• Replaced PLO's particle collision with DE's mutation and crossover operators.
• Incorporated a dynamic crossover rate for improved convergence.
• Implemented a cryptobiosis mechanism to reuse successful solutions.

Main Results:

• CPLODE demonstrated superior performance on 29 CEC 2017 benchmark functions compared to eight other algorithms.
• Achieved higher average ranks and faster convergence rates.
• Showcased competitive results in feature selection on ten real-world datasets, outperforming existing algorithms.

Conclusions:

• CPLODE is an effective algorithm for both global optimization and feature selection.
• The integration of cryptobiosis and DE significantly enhances optimization capabilities.
• CPLODE offers improved classification accuracy and feature reduction.