A Risk-Based IoT Decision-Making Framework Based on Literature Review with Human Activity Recognition Case Studies

Area of Science:

• Computer Science
• Engineering
• Data Science

Background:

• The Internet of Things (IoT) offers significant potential for improving decision-making in critical applications.
• Uncertainties inherent in IoT infrastructure can lead to suboptimal or incorrect decisions.
• Existing decision-making frameworks often struggle to adequately address these pervasive uncertainties.

Purpose of the Study:

• To propose a novel risk-based framework for IoT decision-making that effectively manages uncertainties.
• To integrate domain knowledge into the decision-making process for enhanced rationality.
• To develop and evaluate a new data analytic approach for quantifying and managing prediction uncertainty.

Main Methods:

• A structured literature review identified risks and uncertainty sources in IoT decision-making systems.
• A novel risk-based analytics module, Calibrated Random Forest (CRF), was developed using an ensemble-based approach.
• The CRF method quantifies uncertainty using confidence scores and integrates with domain knowledge for decision rules.

Main Results:

• The proposed framework successfully converts raw sensor data into meaningful actions despite inherent uncertainties.
• Case studies in Human Activity Recognition (HAR) and remote diabetes monitoring demonstrated the framework's efficacy.
• The CRF method effectively quantifies and manages uncertainty in predictions, enhancing decision transparency.

Conclusions:

• The developed risk-based IoT decision-making framework provides a robust solution for managing uncertainties.
• The Calibrated Random Forest (CRF) method offers a significant advancement in uncertainty quantification for IoT predictions.
• The framework enables more rational, transparent, and cost-sensitive decision-making in complex IoT environments.