Predictive modeling of human operator cognitive state via sparse and robust support vector machines

Area of Science:

• Computational intelligence
• Human-machine systems
• Cognitive state monitoring

Background:

• Accurate prediction of human operator cognitive state (HCS) is critical in safety-critical systems.
• The complex relationship between HCS and electrophysiological responses necessitates data-driven modeling approaches.
• Existing models often lack sparseness and robustness, especially with noisy data.

Purpose of the Study:

• To develop a data-driven computational intelligence model for accurate HCS estimation.
• To utilize psychophysiological and performance measures for predictive modeling.
• To evaluate advanced least squares support vector machine (LS-SVM) algorithms for HCS prediction.

Main Methods:

• Employed data-driven modeling using multiple psychophysiological and performance measures.
• Utilized advanced least squares support vector machines (LS-SVM) with parameter optimization via grid search and cross-validation.
• Applied sparse LS-SVM and weighted LS-SVM (WLS-SVM) for enhanced modeling.

Main Results:

• Sparse LS-SVM achieved high modeling accuracy with improved sparseness.
• WLS-SVM demonstrated robustness against noisy training data.
• Both methods exhibited superior generalization performance, capturing HCS temporal fluctuations.

Conclusions:

• Sparse and weighted LS-SVM are effective for modeling HCS temporal variations.
• These intelligent system modeling approaches offer robust and accurate HCS estimation.
• The findings are applicable to safety-critical human-machine systems requiring cognitive state monitoring.