Rapid health data repository allocation using predictive machine learning

Area of Science:

• Health Informatics
• Machine Learning
• Data Management

Background:

• Diverse health data repositories exist, including Electronic Health Records, Electronic Medical Records, Personal Health Records, and Blockchain-based systems.
• The proliferation of digital health data, driven by sources like wearable sensors, necessitates intelligent data storage solutions.
• Current storage allocation decisions are complex, especially for continuously streamed data, and patient preferences play a role.

Purpose of the Study:

• To propose a predictive model for real-time health data storage allocation.
• To develop a system that accommodates patient needs and preferences in storage decisions.
• To address the challenges of managing and storing continuously streamed health data from sources like wearable sensors.

Main Methods:

• A machine learning classifier was employed to learn the relationship between health data characteristics and storage repository features.
• A synthetic training dataset was generated based on expert correlations from small samples.
• The model was designed for rapid, real-time decision-making, even with streaming data.

Main Results:

• The evaluation demonstrated the viability and effectiveness of the proposed machine learning technique for health data storage allocation.
• The predictive model shows promise in making efficient and personalized storage decisions.
• The approach successfully handles the complexities of real-time data streaming.

Conclusions:

• Machine learning offers a viable solution for intelligent health data storage management.
• The developed model can efficiently allocate health data to appropriate repositories, considering patient needs.
• This approach is crucial for managing the increasing volume and velocity of digital health data.