Technology investigation on time series classification and prediction.

Area of Science:

• Data Science
• Machine Learning
• Statistics

Background:

• Time series data is crucial across scientific fields.
• Recent research (2017-2021) heavily focuses on time series classification and prediction.
• A comprehensive review of 120,000 publications indicates these two areas dominate time series research.

Purpose of the Study:

• To analyze the technical development of time series classification and prediction algorithms.
• To provide a reference base for researchers by examining 87 high-impact publications.
• To identify current trends and potential future research avenues in time series analysis.

Main Methods:

• Systematic review of 87 highly relevant and cited literature on time series analysis.
• Categorization of time series classification into supervised, semi-supervised, and early classification methods.
• Exploration of time series prediction techniques, from classical statistics to deep learning and transfer learning.

Main Results:

• Time series classification research is advancing with supervised, semi-supervised, and early classification approaches.
• Time series prediction encompasses a range of methods, including statistical, neural network, fuzzy modeling, and transfer learning.
• The analysis identifies key evolutionary paths for both classification and prediction algorithms.

Conclusions:

• The study offers insights into the current state of time series classification and prediction.
• It suggests future research directions, including interpretability and online learning in time series analysis.
• This work serves as a valuable resource for researchers in the field of time series.