Predictive Model for Suspect Meter Read Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for detecting exceptional meter reads in managed print services rely on outdated average meter rate calculations, which fail to adapt to changing behavior and often flag legitimate usage as suspect, leading to unnecessary reconciliation efforts and costs.
Innovation Solution
A system and method using a predictive model to determine anticipated meter read values and forecast error values, with a sensitivity adjustment, to set a threshold for flagging suspect meter reads, allowing for real-time adaptation to changing usage patterns.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If an unweighted average meter rate is used to detect exceptional meter reads, then the detection method is simple to implement, but it fails to adapt to changing usage patterns and produces false positives
Solution Approach 1:
The patent applies dynamics by transitioning from a static unweighted average meter rate to a dynamic predictive model that continuously adapts to changing usage patterns. The system uses historical meter read data to build predictive models that evolve over time, allowing the detection threshold to dynamically adjust to legitimate behavior changes while maintaining implementation feasibility through automated model training and updating.
2Ease of operation
If a constant average meter rate is presumed between invoices, then the calculation method is straightforward, but it cannot accommodate legitimately changing behavior
Solution Approach 1:
The patent applies parameter changes by modifying the fundamental assumption from a constant average meter rate to a time-varying predictive model. The system changes the parameters used for detection by incorporating historical data patterns, seasonal variations, and usage trends into the predictive model, allowing the detection criteria to adapt to legitimately changing behavior while maintaining computational tractability through efficient model updating mechanisms.
3Ease of manufacture
If equal weighting is given to all historical activity, then the average calculation is simple, but recent activity is not properly emphasized
Solution Approach 1:
The patent applies local quality by implementing differential weighting where different portions of historical data receive different weights based on their relevance. The predictive model assigns higher weights to recent activity patterns and lower weights to older data, creating a localized emphasis on current usage trends. This approach maintains calculation feasibility while significantly improving the accuracy of usage pattern representation by focusing on the most relevant historical periods.
Data Source
AI summary
A system for detecting suspect meter reads in a print environment may include a computing device and a computer-readable storage medium in communication with the computing device. The computer-readable storage medium may include one or more programming instructions for receiving historical meter read values associated with a print-related service, selecting a model set including one or more of the historical meter read values, using a predictive model to determine an anticipated meter read value and a corresponding forecast error value from the model set, determining an updated forecast error value, determining a threshold value, identifying an actual meter read value, determining an average rate associated with the actual meter read value, and flagging the actual meter read value as suspect based on a comparison of the average rate and the threshold value.