An efficient rotational direction heap-based optimization with orthogonal structure for medical diagnosis

Area of Science:

• Computational intelligence
• Optimization algorithms
• Medical informatics

Background:

• The heap-based optimizer (HBO) faces challenges with local stagnation and slow convergence.
• Existing optimization methods require enhancement for effective medical diagnosis applications.

Purpose of the Study:

• To develop an improved optimization algorithm, MGOHBO, to overcome HBO's limitations.
• To apply the enhanced algorithm to medical diagnosis problems by integrating it with Kernel Extreme Learning Machines (KELM).

Main Methods:

• Introduced the Modified Grey Wolf Optimizer based on Heap-based Optimizer (MGOHBO) by incorporating Modified Rosenbrock's Rotational Direction Method (MRM), Grey Wolf Optimizer (GWM) operator, and Orthogonal Learning (OL).
• Evaluated MGOHBO against eleven state-of-the-art optimizers on IEEE CEC 2017 benchmark functions.
• Developed the MGOHBO-KELM model by applying MGOHBO to optimize KELM parameters.
• Validated MGOHBO-KELM on seven disease diagnostic datasets.

Main Results:

• MGOHBO demonstrated significant improvements in convergence accuracy and speed compared to other optimizers on benchmark functions.
• Detailed analysis confirmed MGOHBO's enhanced diversity and balance.
• MGOHBO-KELM achieved optimal results in disease diagnosis tasks, outperforming HBO-KELM and BP models.

Conclusions:

• The proposed MGOHBO algorithm effectively addresses the limitations of HBO, offering superior optimization performance.
• The MGOHBO-KELM model demonstrates practical significance and high efficacy for medical diagnosis applications.