Resilient Behavioral Models for Dynamic Data Rights Withdrawal
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing predictive and decision models face challenges in maintaining validity and efficiency when dealing with dynamic customer data use rights, such as data license rights withdrawal or customer consent revocation, leading to the need for frequent model retraining and economic impact.
Innovation Solution
A system and method for building resilient models by discretizing features into bins, using data sampling to create model surrogate datasets, and tracking data usage rights to ensure model validity through automated data governance, coverage statistics, and estimated mean time to coverage failure.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If automatic model retraining software is used to handle data withdrawal, then model validity can be maintained, but computational resources and time are wasted due to unnecessary frequent retraining
Solution Approach 1:
The system performs preliminary actions by pre-identifying surrogate data points and pre-calculating replacement options before actual data withdrawal occurs. Coverage maps are built in advance, and when data is withdrawn, the system can immediately swap in pre-identified surrogates rather than triggering full retraining. This preliminary preparation resolves the contradiction by maintaining model validity through advance planning while avoiding unnecessary retraining time consumption.
Solution Approach 2:
The system creates copies of data points (surrogates) that replicate the statistical properties and distributions of original data points. These surrogate copies serve as replacements when original data is withdrawn, maintaining model validity without requiring retraining. The copying principle resolves the contradiction by using replicated data representations to preserve model performance while avoiding the time cost of retraining.
2Adaptability or versatility
If data points are removed from the developmental sample, then customer data use rights are respected, but the model parameters change and the model ceases to be valid
Solution Approach 1:
The system introduces surrogate data points as intermediaries between the original data and the model. When customer data must be removed for compliance reasons, surrogates act as mediators that maintain the statistical properties and coverage needed for model validity. The intermediary principle resolves the contradiction by using surrogate representations that satisfy both data right compliance requirements and model validity requirements simultaneously.
Solution Approach 2:
The system changes parameters by transforming data point representations into surrogate forms that preserve statistical properties. Instead of directly removing data points (which changes model parameters), the system transforms them into equivalent surrogate representations with identical distributions. This parameter transformation approach resolves the contradiction by maintaining model validity through parameter preservation while achieving data right compliance through representation changes.
3Measurement precision
If the entire customer dataset is used for modeling, then model accuracy is improved, but the complexity of tracking data usage rights increases
Solution Approach 1:
The system segments the customer dataset into multiple components: actual data points, surrogate data points, and coverage maps. This segmentation allows the system to track data usage rights at the surrogate level rather than managing the entire dataset directly. The segmentation principle resolves the contradiction by maintaining model accuracy through comprehensive data utilization while reducing governance complexity through structured data organization and surrogate-based tracking.
Solution Approach 2:
The system creates universal surrogate data points that can serve multiple functions: maintaining statistical coverage, representing multiple original data points, and enabling efficient rights tracking. These multi-functional surrogates allow the system to achieve model accuracy comparable to using the entire dataset while simplifying rights management. The universality principle resolves the contradiction by using versatile surrogate representations that simultaneously preserve accuracy and reduce complexity.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A system and method of building a decision or prediction model used for analyzing and scoring behavioral transactions is disclosed. A customer dataset in a model development store is used to build an original model is subject to a data right usage withdrawal, the original model having coverage over the customer dataset extract, using data sampling, a portion of the customer dataset to generate a model surrogate dataset. The system and method discretize vectors present in both the model surrogate dataset and the customer dataset, and receive data representing the data right usage withdrawal from the customer dataset. The system and method determine a depletion of the model surrogate dataset according to the data right usage withdrawal, and compute an estimated mean time to coverage failure of the original model based on the depletion of the model surrogate dataset according to the data right usage withdrawal.