Adjustable Data Record Auditing with Regression-Based Drift Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current auditing processes for data records with value increases are manual and inefficient, leading to over-auditing certain parts of the population while missing actual outliers due to systemic drift caused by factors like employee bonuses and stock grants.
Innovation Solution
Applying regression modeling, specifically linear regression, to identify and correct for systemic drift in adjustable data records by normalizing data to a mean deviation of 0, allowing for the use of statistical tools like standard deviation to select audit candidates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual auditing is used with a fixed percentage threshold (e.g., 10% increase), then the process is simple to implement, but it leads to over-auditing certain populations while missing actual outliers due to systemic drift
Solution Approach 1:
The patent transforms the auditing approach by changing from a fixed percentage threshold parameter to a dynamic parameter system that includes regression models and standardized residuals. The system calculates expected values based on historical data and identifies outliers by comparing actual values to these dynamic expectations, thereby improving measurement precision while managing complexity through automated computational processes.
Solution Approach 2:
The patent replaces the manual mechanical auditing process with an automated computational system. Instead of auditors manually reviewing records above a fixed threshold, the system uses regression analysis, calculates standardized residuals, and automatically identifies outliers through statistical computations, thereby improving accuracy while reducing reliance on manual processes.
2Productivity
If a fixed percentage threshold (e.g., 10% increase) is used to trigger audits, then the criteria are easy to apply, but it causes over-auditing of certain populations due to systemic drift from factors like bonuses and stock grants
Solution Approach 1:
The patent implements a feedback mechanism where the system continuously learns from historical audit data and adjusts its expectations accordingly. By using regression models that incorporate historical patterns and calculating standardized residuals, the system provides feedback on what constitutes normal variation versus actual outliers, thereby improving both efficiency and accuracy in identifying audit candidates.
Solution Approach 2:
The patent performs preliminary analysis by establishing regression models and calculating expected values before the actual auditing process. This preliminary action includes analyzing historical data, identifying patterns, and setting dynamic thresholds based on statistical measures such as standardized residuals, which prepares the system to efficiently and accurately identify outliers without manual intervention during the audit execution phase.
3Reliability
If auditors manually review all data records with value increases, then all potential candidates are examined, but the limited time of auditors restricts the number of audits that can be performed
Solution Approach 1:
The patent extracts and isolates the outlier detection function from the overall auditing process. By using regression analysis and standardized residuals to identify specific outlier records, the system separates the identification of high-priority audit candidates from the manual review process, thereby maintaining reliability for critical cases while significantly reducing the time required by extracting only the most suspicious records for auditor attention.
Solution Approach 2:
The patent applies partial action by focusing auditing resources on the most significant outliers rather than examining all records with value increases. By calculating standardized residuals and identifying records that fall outside normal statistical variation, the system performs a targeted partial audit on high-risk candidates, achieving sufficient reliability for detecting significant issues while minimizing time loss compared to comprehensive manual review.
Data Source
AI summary
Embodiments describe techniques for identifying candidates for a value increase audit. The techniques described apply regression models to identify and correct for systemic drift in adjustable data records. In some embodiments, the result is an approximate normal distribution of variations from 0 which allows the use of statistical tools such as standard deviation.