Telematics Outlier Reconciliation via Population Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Analyzing telematics data across diverse vehicle populations is challenging due to variations in make, model, and year of production, leading to inaccurate or erroneous data from vehicles experiencing drift or faulty components, as well as temporary deviations caused by real-world events.
Innovation Solution
A computing system generates statistical models for vehicle populations or subsets defined by make, model, and year of production to identify outliers, normalize data, and assign scores, reducing the impact of anomalous data on analysis results and reporting performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If telematics data is collected across diverse vehicle populations without statistical modeling, then data coverage is comprehensive, but measurement precision deteriorates due to variations in make, model, and year causing inaccurate or erroneous data
Solution Approach 1:
The patent segments the diverse vehicle population into distinct groups based on make, model, and year of production. Statistical models are then applied separately to each segment to identify outliers and normalize data, allowing accurate analysis within each homogeneous group while maintaining comprehensive coverage across the entire diverse population.
Solution Approach 2:
The patent changes the parameters used for data analysis by applying statistical models that transform raw telematics data into normalized values and probability distributions. This parameter transformation allows comparison across diverse vehicles while accounting for manufacturing variations, thereby improving measurement precision without sacrificing adaptability to different vehicle types.
2Reliability
If outlier observations are identified and corrected using statistical models, then reliability of analysis results is improved, but device complexity increases due to the need for statistical modeling and outlier detection systems
Solution Approach 1:
The patent implements self-service by using statistical models that automatically identify outliers and normalize data without requiring manual intervention. The system performs self-diagnosis and self-correction by comparing each vehicle's data against population-based probability distributions, thereby improving reliability while keeping the system relatively simple and automated.
Solution Approach 2:
The patent uses feedback mechanisms where statistical models continuously monitor telematics data, identify deviations from expected patterns, and adjust normalization parameters accordingly. This feedback loop improves reliability by systematically correcting outliers while maintaining a manageable level of complexity through automated iterative refinement.
Data Source
AI summary
A computing system implements a telematics service that obtains a set of vehicle telematics data for each vehicle of a population. Within each set of vehicle telematics data, a set of time-based measurements for a measurement type is identified. The set of time-based measurements identified for each vehicle are combined to obtain a combined set of time-based observations for the measurement type across the population of vehicles or a sub-set of the population defined by vehicle make, model, and/or year of production. An outlier observation is identified from among the combined set of time-based observations. A determination is made whether the outlier observation is part of a temporary deviation or a persistent deviation. For a temporary deviation, an impact of the outlier observation on the set of time-based measurements is reduced. For a persistent deviation, the outlier observation is programmatically characterized.