IoT Data Quality Assessment Framework Using Adaptive Weighted Estimation Fusion.

Area of Science:

• Computer Science
• Data Science
• Internet of Things (IoT)

Background:

• Effective data quality assessment is vital for IoT applications.
• Varying data quality needs across IoT applications create scalability and financial challenges.
• Current methods often require separate data pipelines for each application.

Purpose of the Study:

• To propose a novel approach for end-to-end data quality assessment in IoT.
• To integrate fusion methods into a single data pipeline for diverse applications.
• To analyze the impact of different fusion methods on data quality scores and computational efficiency.

Main Methods:

• Developed an integrated data quality assessment approach using fusion methods.
• Employed real-time and historical analytics to evaluate fusion method performance.
• Tested the approach on two real-world datasets, comparing Kalman, Adaptive weighted, and Naïve fusion.

Main Results:

• Kalman fusion demonstrated a higher overall mean data quality score compared to Adaptive weighted and Naïve fusion.
• Adaptive weighted fusion and Naïve fusion showed lower computational burden.
• The study quantified the trade-offs between data quality improvement and computational cost for each fusion method.

Conclusions:

• The proposed fusion-based approach offers a flexible and efficient solution for IoT data quality.
• A single data pipeline can effectively cater to diverse data quality requirements of multiple IoT applications.
• Choosing the appropriate fusion method depends on balancing data quality needs with system resource constraints.