Graph Embedded Intuitionistic Fuzzy Random Vector Functional Link Neural Network for Class Imbalance Learning.

Area of Science:

• Machine Learning
• Data Science
• Artificial Intelligence

Background:

• Class imbalance (CI) poses significant challenges in machine learning, leading to biased models that underrepresent minority classes.
• Random Vector Functional Link (RVFL) networks, while effective, struggle with imbalanced datasets.

Purpose of the Study:

• To propose a novel model, GE-IFRVFL-CIL, to overcome the limitations of RVFL networks in CI learning.
• To enhance classification accuracy on imbalanced datasets by incorporating graph embedding and intuitionistic fuzzy theory.

Main Methods:

• Developed a Graph-Embedded Intuitionistic Fuzzy RVFL for CI Learning (GE-IFRVFL-CIL) model.
• Integrated a weighting mechanism to manage imbalanced data.
• Utilized graph embedding (GE) to preserve topological data structures.
• Employed intuitionistic fuzzy (IF) theory to handle data uncertainty and imprecision.

Main Results:

• The GE-IFRVFL-CIL model demonstrated superior performance on KEEL benchmark imbalanced datasets, even with Gaussian noise.
• Achieved promising results on the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset, showcasing real-world applicability.
• The model effectively mitigated noise and outliers while preserving inherent dataset geometrical structures.

Conclusions:

• The proposed GE-IFRVFL-CIL model offers a robust solution for class imbalance learning.
• It effectively handles noisy and imprecise data, enhancing model generalization.
• The approach successfully preserves critical data structures, leading to improved classification outcomes.