Federated Compressed Learning Edge Computing Framework with Ensuring Data Privacy for PM2.5 Prediction in Smart City...

Area of Science:

• Environmental Science
• Computer Science
• Data Science

Background:

• Sparse data in PM2.5 air quality monitoring systems challenges smart city sensing applications.
• Existing centralized systems face data privacy concerns due to exposed sensor data.
• Inefficient deployment, communication, and fragmented records hinder high-resolution air quality prediction.

Purpose of the Study:

• To propose a novel edge computing framework, Federated Compressed Learning (FCL), for efficient and private PM2.5 prediction in smart cities.
• To develop a green energy-based wireless sensing network system utilizing the FCL framework.
• To validate the performance of FCL through prototype development and testing.

Main Methods:

• Implementing a framework combining compression techniques, regional joint learning, and secure data exchange.
• Utilizing edge computing for decentralized data processing and privacy preservation.
• Developing prototypes for a green energy-based wireless sensing network.

Main Results:

• Achieved over 95% reduction in data consumption.
• Maintained an error rate below 5% for PM2.5 predictions.
• Demonstrated secure data transmission and heavy data compaction in wireless sensing networks.

Conclusions:

• FCL effectively reduces data quantity and ensures data privacy for PM2.5 prediction in smart city sensing.
• The framework supports green energy-based wireless sensing networks and software/hardware co-design.
• While FCL shows slightly lower accuracy than centralized training, its data compaction and security benefits are significant for WSNs.