Edge4TSC: Binary Distribution Tree-Enabled Time Series Classification in Edge Environment.

Area of Science:

• Computer Science
• Data Science
• Artificial Intelligence

Background:

• Time series data generation is rapidly increasing across diverse fields.
• Time Series Classification (TSC) is crucial for applications like human activity recognition and smart city governance.
• There's a growing need for timely TSC without manual feature engineering.

Purpose of the Study:

• To propose Edge4TSC, a framework for edge-based time series processing and classification.
• To leverage deep learning for automatic feature extraction in TSC.
• To introduce a new time series representation method to enhance classification accuracy.

Main Methods:

• Developed the Edge4TSC framework for edge computing environments.
• Applied deep learning techniques for automated feature extraction.
• Introduced a novel time series representation using a binary distribution tree.

Main Results:

• Edge4TSC processes time series data in the edge environment for instant results.
• Deep learning methods achieved competitive performance compared to state-of-the-art TSC solutions.
• The binary distribution tree representation significantly improved classification accuracy on challenging datasets.

Conclusions:

• Edge4TSC offers an efficient solution for real-time TSC on edge devices.
• The proposed representation method effectively addresses accuracy concerns in TSC.
• The framework demonstrates strong generalization ability and accuracy improvement validated through experiments.