Association Rule Mining in Multiple, Multidimensional Time Series Medical Data.

Area of Science:

• Biomedical Engineering
• Data Science
• Signal Processing

Background:

• Time series pattern mining (TSPM) is crucial for identifying dependencies in sequential data.
• Multidimensional time series data, especially from multiple sensors like electromyogram (EMG), present complex analytical challenges.
• Integrating quantitative attributes with time series patterns is essential for comprehensive analysis in fields like healthcare.

Purpose of the Study:

• To develop an efficient method for discovering patterns and association rules in multidimensional time series data from multiple EMG sensors.
• To incorporate quantitative attributes (e.g., signal energy, onset time) into the time series pattern mining process.
• To demonstrate the scalability and speed advantages of the proposed approach compared to existing methods.

Main Methods:

• A two-stage approach for time series pattern mining was proposed.
• Stage 1: Sequential mining across time slices to discover frequent patterns in multiple time series.
• Stage 2: Focused analysis of quantitative attributes for time series identified in Stage 1.

Main Results:

• The proposed two-stage approach significantly speeds up the discovery of association rules in multidimensional EMG data.
• The method demonstrates linear scalability with respect to the number of time series involved.
• Evaluations on large EMG datasets confirm the efficiency and effectiveness of the approach.

Conclusions:

• The developed two-stage method provides an efficient and scalable solution for time series pattern mining in multidimensional sensor data.
• This approach is applicable to various medical sensor databases, aiding research in rehabilitation and prosthetic device design.
• The integration of quantitative attributes enhances the utility of TSPM in complex biological systems.