Multipopulation Whale Optimization-Based Feature Selection Algorithm and Its Application in Human Fall Detection...

Area of Science:

• Computer Science
• Artificial Intelligence
• Machine Learning

Background:

• Feature selection (FS) is crucial for dimensionality reduction in pattern recognition.
• Traditional FS methods struggle with complex, high-dimensional datasets.
• Identifying optimal feature subsets remains a significant challenge.

Purpose of the Study:

• To introduce a novel meta-heuristic algorithm for effective feature selection.
• To address limitations of existing methods in handling high-dimensional data.
• To improve the performance and efficiency of feature selection processes.

Main Methods:

• Development of a Multispiral Whale Optimization Algorithm (MSWOA).
• Incorporation of an Adaptive Multipopulation Merging Strategy (AMS) to prevent premature convergence.
• Implementation of a Double Spiral Updating Strategy (DSS) to escape local optima.
• Integration of a Baleen Neighborhood Exploitation Strategy (BES) to enhance convergence speed.

Main Results:

• MSWOA demonstrated superior performance compared to six state-of-the-art meta-heuristic algorithms and six WOA-based algorithms.
• The proposed method achieved better results on 20 UCI datasets.
• Successful application in human fall-detection tasks, validating its practical utility.

Conclusions:

• The MSWOA offers a robust and efficient solution for feature selection in complex, high-dimensional scenarios.
• The novel strategies within MSWOA effectively improve search capabilities and convergence.
• MSWOA shows significant potential for real-world applications, including human fall detection.