Bayesian Inference for Dynamic Place Properties
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing spatial databases fail to effectively account for dynamic properties of locations, such as human activity and mobility, which are essential for accurate decision-making in various business industries. Current approaches treat these dynamic properties as quasi-static and rely on ad hoc techniques, lacking a generic solution for providing these properties.
Innovation Solution
A processor-implemented method for automatically inferring place properties based on partially observable entity spatial activity data using data-driven models. This method involves obtaining event data streams from independently controlled data sources, identifying locations, deriving contextual events, and inferring place properties using Bayesian models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If ad hoc techniques are used to estimate dynamic properties of locations, then implementation simplicity is maintained, but measurement precision and reliability of place properties deteriorate
Solution Approach 1:
The patent introduces Bayesian inference models as an intermediary layer between raw spatial activity data and place property estimates. This mediator processes partial and noisy data through probabilistic reasoning to produce accurate dynamic property estimates, resolving the contradiction by maintaining implementation simplicity while significantly improving measurement precision through sophisticated data processing.
Solution Approach 2:
The patent transforms static place property parameters into dynamic parameters that evolve over time based on observed spatial activity. By continuously updating place properties using Bayesian inference as human activity ebbs and flows, the system maintains high measurement precision while using systematic rather than ad hoc techniques.
2Measurement precision
If multiple data sources are fused to improve comprehensiveness of place properties, then measurement precision improves, but device complexity and difficulty of detecting and measuring increase
Solution Approach 1:
The patent segments the complex data fusion process into distinct functional modules: data collection from multiple independent sources, Bayesian inference processing, and place property estimation. This segmentation reduces device complexity by making each module independent and manageable while maintaining comprehensive place properties through the integration of all sources.
Solution Approach 2:
The patent creates a universal Bayesian inference framework that can process data from multiple different sources (spatial databases, data streams, sensors) through a single unified model. This multi-functional approach improves measurement precision by comprehensively integrating all available data while avoiding the complexity of source-specific processing pipelines.
3Adaptability or versatility
If real-time data processing is implemented to capture dynamic properties, then adaptability of place properties improves, but productivity and loss of time in processing increase
Solution Approach 1:
The patent implements periodic Bayesian inference updates that process spatial activity data at intervals rather than continuously. This periodic action maintains real-time adaptability of place properties by capturing dynamic changes as they occur while preserving processing efficiency by avoiding constant computation, thus resolving the contradiction between adaptability and productivity.
4Adaptability or versatility
If partial data from multiple sources is used to infer place properties, then adaptability to data availability improves, but measurement precision deteriorates due to data exhaust and noise
Solution Approach 1:
The patent converts the harmful effect of partial and noisy data into a benefit by using Bayesian inference to explicitly model data uncertainty. Instead of treating data exhaust and noise as problems to be eliminated, the system embraces them as inherent characteristics and uses probabilistic reasoning to extract accurate place property estimates, thus improving measurement precision while maintaining adaptability to partial data availability.
Data Source
AI summary
A method for determining a use of one or more locations and at least one criteria of the one or more locations using smartphones, including receiving, via a wireless network, a plurality of event data streams at different spatio-temporal resolutions from a plurality of mobile devices (e.g., at least one of the smartphones) associated with a plurality of entities, the plurality of event data streams including a plurality of unique mobile identifiers, location pings, and access pings of the plurality of mobile devices.