Stacked Machine Learning for Event-Aware Glucose Predictions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional glucose prediction systems fail to accurately predict future glucose levels due to their limited consideration of historical glucose measurements alone, neglecting events and additional factors that significantly impact glucose levels, leading to inaccurate predictions and potential health risks.
Innovation Solution
A CGM platform utilizing a stacked configuration of multiple machine learning models processes vast amounts of data from continuous glucose monitoring systems to generate glucose measurement and event predictions, incorporating additional data such as user behavior and event characteristics, and employs confidence-based input selection to enhance prediction accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional glucose prediction techniques use only historical glucose measurements as input, then the system complexity is low, but the prediction accuracy deteriorates due to inability to account for events impacting glucose levels
Solution Approach 1:
The system segments the prediction task into multiple independent machine learning models, each specializing in predicting specific events (insulin administration, exercise, meals, glucose levels). This allows each model to focus on specific patterns while collectively providing comprehensive predictions that account for all glucose-impacting events, thereby improving accuracy without proportionally increasing overall system complexity
Solution Approach 2:
The machine learning model performs multiple functions simultaneously by predicting various types of events (insulin administration, exercise, meals) and glucose levels from the same input data. This multi-functional approach allows a single system to capture diverse glucose-impacting factors, improving prediction accuracy while avoiding the need for separate specialized systems for each event type
2Loss of information
If a CGM platform processes vast amounts of continuous glucose monitoring data to identify patterns, then the prediction comprehensiveness improves, but the processing complexity becomes impossible for human analysis
Solution Approach 1:
The system replaces human analytical capabilities with automated machine learning models that can process vast amounts of CGM data. The machine learning algorithm substitutes for human pattern recognition, enabling the system to analyze comprehensive datasets including glucose measurements, user inputs, and contextual information without the limitations of human processing capacity
Solution Approach 2:
The machine learning model acts as an intermediary between raw CGM data and actionable predictions. It processes the vast amount of continuous glucose monitoring data, user inputs, and contextual information, transforming this complex dataset into meaningful event predictions and glucose level forecasts that can be used for diabetes management
3Reliability
If glucose predictions do not account for upcoming events like meals and exercise, then the system operation is simple, but the predictions become inaccurate and potentially dangerous
Solution Approach 1:
The system performs preliminary prediction of upcoming events (meals, exercise, insulin administration) before these events actually occur or before their glucose-impacting effects are observed. By predicting events in advance, the system can prepare appropriate responses and alerts, improving prediction reliability while maintaining operational simplicity through automated event detection
Data Source
AI summary
Glucose measurement and glucose-impacting event prediction using a stack of machine learning models is described. A CGM platform includes stacked machine learning models, such that an output generated by one of the machine learning models can be provided as input to another one of the machine learning models. The multiple machine learning models include at least one model trained to generate a glucose measurement prediction and another model trained to generate an event prediction, for an upcoming time interval. Each of the stacked machine learning models is configured to generate its respective output when provided as input at least one of glucose measurements provided by a CGM system worn by the user or additional data describing user behavior or other aspects that impact a person's glucose in the future. Predictions may then be output, such as via communication and/or display of a notification about the corresponding prediction.