Multilabel Classification With Group-Based Mapping: A Framework With Local Feature Selection and Local Label...

Area of Science:

• Machine Learning
• Data Mining
• Artificial Intelligence

Background:

• Multilabel learning assigns multiple labels to instances, a common scenario in real-world data.
• Existing methods often use global feature selection and label correlation, assuming uniformity across all instances.
• This global approach may not capture nuanced relationships present in diverse datasets.

Purpose of the Study:

• To propose a novel multilabel learning framework that incorporates local feature selection and local label correlation.
• To address the limitations of global methods by allowing instance-specific or group-specific feature weights and label correlations.
• To enhance the accuracy and adaptability of multilabel classification models.

Main Methods:

• Developed a framework assuming instances can be clustered into groups.
• Implemented group-specific feature selection by extracting instance-group correlations.
• Introduced a label-specific group selection process based on group-label correlations.
• Incorporated intergroup correlations for comprehensive modeling.

Main Results:

• The proposed framework demonstrated effectiveness in multilabel classification tasks.
• Experimental results on various datasets validated the superiority of the local approach over global methods.
• The framework successfully captured varying feature selection weights and label correlations across different instance groups.

Conclusions:

• The novel framework effectively utilizes local feature selection and label correlation for improved multilabel learning.
• Clustering instances and applying group-specific correlations enhances classification performance.
• The approach offers a more flexible and accurate solution for complex multilabel problems.