IPO: An Improved Parrot Optimizer for Global Optimization and Multilayer Perceptron Classification Problems

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Machine Learning

Background:

• The Parrot Optimizer (PO) is a novel algorithm inspired by Pyrrhura Molinae parrot behaviors.
• Existing optimization algorithms may require enhancements for complex global problems and neural network training.

Purpose of the Study:

• To introduce an improved Parrot Optimizer (IPO) for global optimization and multilayer perceptron (MLP) training.
• To enhance the exploration-exploitation balance of the basic PO algorithm.

Main Methods:

• IPO incorporates an aerial search strategy from Arctic Puffin Optimization.
• Modified staying and communicating behaviors using random movement and roulette fitness-distance balance selection.
• Evaluation using CEC2022 test functions, standard classification datasets, and MLP for oral English teaching quality evaluation.

Main Results:

• IPO exhibited superior performance compared to six other well-known optimization algorithms on complex global optimization problems.
• IPO-MLP achieved the highest classification accuracy of 88.33% on the oral English teaching quality evaluation dataset.

Conclusions:

• The proposed IPO algorithm is effective for solving complex global optimization problems.
• IPO demonstrates significant effectiveness in optimizing MLP models for classification tasks, proving its practical applicability.