FDEPCA: A Novel Adaptive Nonlinear Feature Extraction Method via Fruit Fly Olfactory Neural Network for IoMT Anomaly...

Area of Science:

• Computer Science
• Biomedical Engineering
• Data Science

Background:

• The Internet of Medical Things (IoMT) generates complex, high-dimensional, and nonlinear data, challenging anomaly detection.
• Existing nonlinear feature extraction methods struggle with high-dimensional outliers and computational expense, distorting data structures.

Purpose of the Study:

• To propose a novel adaptive nonlinear feature extraction method for IoMT data.
• To enhance the efficiency and accuracy of anomaly detection in IoMT systems.
• To address challenges in extracting features from high-dimensional, nonlinear IoMT data.

Main Methods:

• A new method, Fruit Fly dimension expansion projection and remain main components by PCA (FDEPCA), is introduced.
• FDEPCA involves mean-centering data, dimension expansion projection using a binary sparse random projection matrix, and principal component analysis (PCA).
• The method is evaluated by applying extracted features to anomaly detection models using ROC curves and AUC metrics.

Main Results:

• The FDEPCA algorithm demonstrated superior classification performance compared to existing nonlinear feature extraction techniques.
• FDEPCA offers a significant advantage in projection time, making it computationally efficient.
• The method showed strong applicability across various anomaly detection models, including proximity-based, probability-based, and ensemble-based classifiers.

Conclusions:

• FDEPCA effectively extracts nonlinear features from high-dimensional IoMT data with minimal information distortion.
• The proposed method improves anomaly detection performance and computational efficiency in IoMT applications.
• FDEPCA is a versatile and robust technique applicable to diverse anomaly detection tasks within the IoMT.