A review of subsequence time series clustering

Area of Science:

• Computer Science
• Data Mining
• Pattern Recognition

Background:

• Subsequence time series clustering is a critical but unresolved issue in data analysis.
• It has diverse applications, including e-commerce, outlier detection, speech recognition, and DNA recognition.

Purpose of the Study:

• To provide a comprehensive review of subsequence time series clustering techniques.
• To categorize existing literature into preproof, interproof, and postproof periods.
• To evaluate the strengths and weaknesses of current state-of-the-art methods.

Main Methods:

• Literature review and categorization of subsequence time series clustering approaches.
• Analysis of methods based on their historical development (preproof, interproof, postproof).
• Evaluation of the performance and limitations of various clustering algorithms.

Main Results:

• The paper categorizes existing literature into three distinct periods.
• It discusses various state-of-the-art subsequence time series clustering methods within these categories.
• Identified strengths and weaknesses of employed methods provide insights for future research.

Conclusions:

• Subsequence time series clustering is an evolving field with ongoing challenges.
• A structured review of historical and current methods is essential for progress.
• Identifying limitations of existing techniques will guide future research and development.