Robust adaptive fault-tolerant learning control for human height-weight prediction based on DNN

Area of Science:

• Biometrics
• Control Theory
• Deep Learning

Background:

• Traditional physiological characteristic prediction methods struggle with data noise, environmental changes, and outliers.
• Accurate prediction of individual physiological characteristics is crucial for public security, including crime prevention and behavior analysis.

Purpose of the Study:

• To propose a distributed robust adaptive confined fault-tolerant optimal control method based on deep neural networks for human height and weight prediction.
• To address complexity and uncertainty issues in prediction models, enhancing robustness and adaptability.

Main Methods:

• A novel framework combining deep learning and fault-tolerant control theory.
• Introduction of a limited fault-tolerant mechanism to maintain prediction accuracy and stability under perturbations and incomplete data.
• Evaluation using statistical similarity, overall predictive performance, and minority-class detection ability.

Main Results:

• Achieved high prediction accuracies for human height (up to 98.4%) and weight (up to 98.2%) even with 30% data loss.
• Demonstrated robust performance with fault-tolerant prediction accuracies of 97.7% for height and 97.5% for weight.
• Results are applicable to small-to-medium tabular datasets with moderate class imbalance.

Conclusions:

• The proposed deep neural network-based control method offers an efficient and reliable framework for biometric prediction.
• The fault-tolerant mechanism ensures high accuracy and stability, making it valuable for public security technology.
• Provides a new technical path for biometric prediction and analysis with significant theoretical and practical implications.