Graph-Based User Action Inference for Out-of-Pattern Behavior
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems for inferring user activity based on electronic device signals are inaccurate due to reliance on historical usage patterns and lack of consideration for situational parameters, leading to inadequate content recommendations and user experience.
Innovation Solution
A method using a graph-based approach to dynamically determine user-specific contexts and behaviors, incorporating weighted indices such as average usage, context affinity, and confidence, to identify current user actions and perform relevant device actions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If systems rely on historical usage patterns and content correlations for predicting user actions, then they can provide recommendations based on past behavior, but the accuracy deteriorates when users exhibit deviation-of-interest or out-of-pattern behavior
Solution Approach 1:
The system dynamically determines user-specific contexts and behaviors in real-time rather than relying on static historical patterns. The graph-based approach continuously updates user models with current contextual information, enabling accurate predictions even when user behavior deviates from historical norms. The system adapts to changing user preferences and situations through dynamic context aggregation and scenario evaluation.
Solution Approach 2:
The system changes the parameters used for prediction from purely historical usage patterns to a combination of current contextual parameters and historical data. By incorporating situational parameters (time, location, device state) and dynamically weighting historical versus current information, the system maintains accuracy while adapting to out-of-pattern behavior. The weighted indices allow flexible adjustment of how much historical data versus current context influences predictions.
2Quantity of substance
If systems use entirely historical usage history for content recommendations, then they can leverage accumulated data, but the recommendations become vulnerable to errors when user behavior does not match historic patterns
Solution Approach 1:
The system segments historical usage data into meaningful contexts and scenarios rather than treating it as a monolithic history. By organizing data into discrete contextual nodes (base contexts and augmented contexts) within a graph structure, the system can selectively apply historical patterns where appropriate while remaining flexible when current situations differ from historical norms. This segmentation allows the system to leverage accumulated data without being constrained by it.
Solution Approach 2:
The system introduces current contextual information as an intermediary between historical data and prediction output. Rather than directly applying historical patterns to current situations, the system uses real-time context (time, location, device state, user actions) as a mediator that filters and adapts historical data to current conditions. This intermediary layer enables the system to maintain historical knowledge while adapting to present circumstances.
3Device complexity
If systems do not account for current or future situation parameters, then they can simplify recommendation calculations, but the system becomes more vulnerable and error-prone in handling user behavior not matching historic patterns
Solution Approach 1:
The system adds new dimensions to the recommendation calculation by incorporating situational parameters (time, location, device state, user actions) alongside historical data. Rather than remaining in a single-dimensional historical framework, the system operates in a multi-dimensional space where current context and historical patterns interact. This dimensional expansion increases computational complexity but significantly improves reliability for situation-driven behavior.
Data Source
AI summary
In a system which predicts useful actions for a user, a graph is used to permit better suggested actions. The graph includes base contexts which are related to time, place and occasion and augmented contexts which are related to device state and user actions. A base context together with one or more augmented contexts may provide a suggested action. Several alternative groupings of base contexts and augmented contexts is a scenario. A high-scoring suggested action from one of the scenarios is provided as the suggested action.