Trip Destination Inference via Open-Closed World Model Mixture
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional computing systems that predict future activities or destinations are limited by their reliance on closed-world approaches, which are constrained by incomplete data sets due to costly data collection, lack of ongoing sensing, and privacy concerns, leading to inaccurate predictions.
Innovation Solution
The integration of open-world and closed-world submodels using a submodel weight to combine probabilities associated with observed and unobserved data, allowing for predictions that consider both known and unknown data points, thereby mitigating inaccuracy from incomplete data sets.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a closed-world approach is used to generate predictions based on observed data, then the system is simpler to implement, but the prediction accuracy deteriorates due to incomplete data sets
Solution Approach 1:
The patent combines closed-world and open-world submodels into a unified prediction system. The closed-world submodel handles observed data while the open-world submodel accounts for unobserved data, and their predictions are merged through a weighted combination to produce the final prediction, thereby improving accuracy without excessive complexity
Solution Approach 2:
The patent introduces an intermediary mechanism (the open-world submodel and weighting system) that bridges the gap between limited observed data and the need for accurate predictions. This intermediary allows the system to reason about unobserved data and incorporate it into predictions, improving accuracy while maintaining manageable system complexity
2Measurement precision
If data collection is expanded to improve prediction accuracy, then the completeness of data sets improves, but the cost and complexity of data collection increases
Solution Approach 1:
The patent segments the data handling process into two distinct submodels: a closed-world submodel that processes observed data and an open-world submodel that processes unobserved data. This segmentation allows the system to work effectively with incomplete data without requiring extensive additional data collection, thereby improving prediction accuracy while avoiding the complexity and cost of comprehensive data collection
3Measurement precision
If ongoing embedded sensing is implemented to collect more data, then the completeness of data sets improves, but the system complexity and privacy concerns increase
Solution Approach 1:
The patent segments the prediction system into closed-world and open-world components, allowing it to function effectively with the data that is already collected without requiring ongoing embedded sensing. This approach improves prediction accuracy by reasoning about unobserved data while avoiding the complexity and privacy issues associated with continuous monitoring
4Quantity of substance
If the system only considers previously observed locations, then the data set remains manageable, but the prediction accuracy deteriorates when users visit new places
Solution Approach 1:
The patent introduces an open-world submodel as an intermediary that allows the system to reason about unobserved locations. This intermediary enables the system to maintain a manageable data set of observed locations while still generating accurate predictions for new places by incorporating probabilistic reasoning about unobserved data into the prediction process
Data Source
AI summary
The claimed subject matter provides systems and/or methods that facilitate generating an inference about events that may not have yet been observed. Open-world modeling can be used to take a history of observation so as to understand trends over time in the revelation of previously unseen events, and to make inferences with subsets of data that new unseen events will be seen. Thus, inaccuracies associated with predictions generated from incomplete data sets can be mitigated. To yield such predictions, open-world submodels and closed-world submodels that do not allow for previously unseen events can be combined via a model mixture methodology, which fuses inferences from the open- and close-world models.