Fuzzy Cognitive Map-Driven Comprehensive Time-Series Classification.

Area of Science:

• Artificial Intelligence
• Machine Learning
• Data Science

Background:

• Time-series classification is crucial in many fields.
• Existing methods often lack interpretability or optimal performance.
• Fuzzy Cognitive Maps (FCMs) offer interpretability but require enhancement for complex classification tasks.

Purpose of the Study:

• To propose a novel, comprehensive approach for time-series classification.
• To enhance the performance and interpretability of Fuzzy Cognitive Maps (FCMs) for classification.
• To develop a model that effectively captures temporal dynamics in data.

Main Methods:

• Utilized a Fuzzy Cognitive Map (FCM) as the core classification engine.
• Employed a moving-window technique for time-series data staging to capture temporal flow.
• Implemented backward error propagation for hyperparameter tuning, including FCM size and learning rate.
• Separated the classification engine from specialized pre- and post-processing stages.

Main Results:

• The proposed model demonstrated strong performance against various state-of-the-art time-series classification algorithms.
• The integration of pre- and post-processing significantly improved classification accuracy.
• The model successfully combined FCM interpretability with high classification performance.

Conclusions:

• The developed FCM-based approach offers a robust and interpretable solution for time-series classification.
• The model's architecture, separating processing stages, is key to its effectiveness.
• This method advances the field by providing a high-performing, interpretable classifier for temporal data.