Concept learning based on improved FCM- BiLSTM for fuzzy data classification and fusion.

Area of Science:

• Artificial Intelligence
• Machine Learning
• Cognitive Science

Background:

• Concept-Cognitive Learning (CCL) simulates human cognition for knowledge discovery.
• Existing CCL methods struggle with inaccurate labels and object dependencies, limiting adaptability.

Purpose of the Study:

• To propose a novel Fuzzy Concept-Cognitive Learning Model (FCLSCL) addressing limitations in current CCL approaches.
• To enhance concept learning by integrating fuzzy clustering and deep learning for improved accuracy and adaptability.

Main Methods:

• Developed FCLSCL integrating an improved Fuzzy C-means (FCM) and Bidirectional Long Short-Term Memory Network (BiLSTM).
• Introduced weighted fuzzy concepts to manage data uncertainty by considering membership degrees and pseudo-concepts.
• Utilized concatenated fuzzy concepts as input for BiLSTM to enable bidirectional concept learning.

Main Results:

• The proposed FCLSCL model demonstrated superior effectiveness compared to popular machine learning models, existing CCL methods, and LSTM.
• The model successfully addressed challenges related to inaccurate labels and object dependencies in concept learning.
• Experimental results validated the model's ability to handle diverse datasets and complex relational patterns.

Conclusions:

• FCLSCL offers a significant advancement in concept learning by effectively handling data uncertainty and complex relationships.
• The integration of fuzzy clustering and BiLSTM provides a robust framework for knowledge discovery.
• The FCLSCL model shows promise for applications requiring accurate and adaptable concept classification.