An Efficient Pattern Mining Approach for Event Detection in Multivariate Temporal Data

Area of Science:

• Data Science
• Machine Learning
• Biomedical Informatics

Background:

• Complex multivariate temporal data, like electronic health records, present challenges for event detection.
• Accurate event detection is crucial for intelligent patient monitoring and clinical decision support.

Purpose of the Study:

• To propose a novel pattern mining approach for learning event detection models from complex temporal data.
• To develop an efficient method for finding predictive patterns in time series data.
• To create a framework for selecting minimal, predictive, and non-spurious patterns.

Main Methods:

• Developed Recent Temporal Pattern mining to convert time series data into temporal abstraction sequences.
• Constructed complex time-interval patterns using temporal operators, working backward in time.
• Introduced the Minimal Predictive Recent Temporal Patterns framework for pattern selection.

Main Results:

• Applied the methods to predict adverse medical events using real-world clinical data.
• Demonstrated the effectiveness of the approach in learning accurate event detection models.
• Showcased the benefits for developing intelligent patient monitoring and decision support systems.

Conclusions:

• The proposed pattern mining approach is effective for event detection in complex temporal data.
• This method advances the development of intelligent systems for healthcare.
• Accurate event prediction is a key step towards improved patient care and safety.