Device-Dwell Graphs for Noisy Geolocation Analytics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing geolocation analytics systems face challenges in accurately determining places of interest visited by mobile devices due to noisy and low-accuracy geolocation measurements, which are exacerbated by issues like GPS shadows and varying signal strengths, leading to difficulties in identifying precise locations and analyzing user behavior at scale.
Innovation Solution
The development of a device-dwell graph system that organizes spatial and temporal data to accurately determine likely dwell locations by using a dynamic distancing model, which cleanses noisy data and attributes device presence to specific places, even with inaccurate geolocation reports, and probabilistically matches devices to individuals across multiple devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional geolocation analytics systems use reported geolocation data from mobile devices, then they can track user movement and identify places of interest, but the accuracy of location determination deteriorates due to noisy measurements, GPS shadows, and signal variations
Solution Approach 1:
The patent introduces an intermediary processing layer that receives noisy geolocation reports and transforms them into reliable location determinations. This layer uses multiple data sources (cell tower triangulation, WiFi positioning, GPS) and applies filtering algorithms to produce a consolidated, more accurate location estimate that mediates between the unreliable individual measurements and the final location analysis
Solution Approach 2:
The system merges multiple geolocation measurement techniques (cellular triangulation, GPS signals, Bluetooth beacons, WiFi positioning) into a unified location determination process. By combining these diverse data sources and cross-validating their results, the system achieves higher reliability than any single method could provide alone
2Productivity
If the system processes geolocation data at scale with millions of devices and hundreds of thousands of places of interest, then it can provide comprehensive analytics coverage, but the computational complexity and difficulty of detecting precise locations increases
Solution Approach 1:
The patent segments the large-scale geolocation data processing into hierarchical levels: local processing at device level (filtering and preliminary analysis), regional processing at server level (aggregating device data within geographic zones), and global processing (overall pattern recognition). This segmentation allows the system to handle millions of devices by processing data in manageable chunks rather than as a monolithic problem
Solution Approach 2:
The system adds temporal dimensionality to the geolocation data processing, analyzing location patterns over time windows rather than treating each location point independently. This transforms the problem from a static spatial analysis to a dynamic spatiotemporal analysis, enabling the system to distinguish between transient location noise and genuine place visits through temporal pattern recognition
3Measurement precision
If the system uses multiple significant digits for geolocation reporting, then it appears to provide high precision, but the actual measurement accuracy remains limited by inherent limitations of the measurement techniques
Solution Approach 1:
The system changes the parameter representation by transitioning from raw coordinate values with false precision to probability distributions that explicitly model measurement uncertainty. Instead of reporting a single point with many decimal places, the system reports likelihood ranges and confidence intervals, transforming the data representation to reflect the true state of measurement uncertainty
Data Source
AI summary
Provided is a process that determines a device-dwell graph based on noisy reported geolocations for mobile computing devices.