An enhanced version of Harris Hawks Optimization by dimension learning-based hunting for Breast Cancer Detection

Area of Science:

• Computational intelligence
• Medical data mining
• Swarm intelligence

Background:

• Swarm intelligence techniques are applied in medical data mining for disease classification and prediction.
• Breast cancer, a leading cause of death, requires effective screening and diagnostic tools like mammography.
• Existing optimization algorithms like Harris Hawks Optimization (HHO) face challenges such as lack of diversity and premature convergence.

Purpose of the Study:

• To develop an enhanced Harris Hawks Optimization (HHO) algorithm, termed DLHO, for biomedical databases.
• To address limitations of HHO, including reduced crowd diversity, premature convergence, and exploration-exploitation imbalance.
• To improve the accuracy and robustness of disease classification models in medical data mining.

Main Methods:

• An enhanced Harris Hawks Optimization (HHO) algorithm, DLHO, was developed by integrating the dimension learning-based hunting (DLH) search strategy.
• The DLH strategy facilitates neighborhood paradigm and information sharing among search agents to maintain diversity and balance global/local search.
• The DLHO algorithm was evaluated using 29-CEC-2017 test suites and applied to biomedical databases (MIAS, UCI Machine Learning Repository for breast cancer, Balloon, and Heart).

Main Results:

• The proposed DLHO algorithm demonstrated superior performance compared to other optimizers on the CEC-2017 test suites.
• Experiments on biomedical databases, including breast cancer datasets (MIAS, UCI), showed the effectiveness of DLHO.
• The robustness of DLHO was further confirmed through testing on additional UCI datasets (Balloon, Heart).

Conclusions:

• The developed DLHO algorithm effectively overcomes the limitations of the standard HHO, enhancing diversity and balancing exploration-exploitation.
• DLHO shows significant promise for improving the classification and prediction of diseases, particularly breast cancer, in medical data mining.
• The proposed method offers a robust and effective approach for analyzing biomedical data, contributing to advancements in computational intelligence for healthcare.