Interpretable User Segmentation via Predictive Model Mediation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing predictive models, such as neural networks, do not provide insights into the relevant user attributes that influence their predictions, making it challenging for operators to configure computing systems effectively for targeted content delivery.
Innovation Solution
A data analysis application determines user segments by generating conditions from training data, assessing their relevance, and presenting them to operators, allowing for better understanding and configuration of predictive models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If predictive models use multiple hidden layers to extract features, then prediction accuracy is improved, but interpretability of features deteriorates
Solution Approach 1:
The patent introduces an intermediary component that translates the internal numerical features of the neural network into human-interpretable conditions. This mediator layer converts the black-box feature representations into understandable rules that operators can use, bridging the gap between accurate predictions and interpretability without modifying the core predictive model.
2Measurement precision
If the computing system runs the predictive model for every user, then prediction accuracy is maintained, but computational burden increases
Solution Approach 1:
The patent segments the user population into distinct groups based on the interpretable conditions derived from the predictive model. Instead of running the full predictive model for every individual user, the system applies pre-defined segment rules to rapidly classify users, significantly reducing computational burden while maintaining accurate prediction capabilities through the segment assignments.
3Ease of operation
If the operator configures the computing system without knowledge of relevant attributes, then system simplicity is maintained, but targeted content delivery deteriorates
Solution Approach 1:
The system automatically generates and provides the relevant attribute knowledge to the operator through the interpretable conditions. Rather than requiring the operator to manually identify important features or configure complex parameters, the system self-generates the actionable insights from the training data and presents them in an easily consumable format, enabling effective configuration without expert knowledge.
Data Source
AI summary
Systems and methods are disclosed herein for determining user segments created by a predictive model based on user behavioral data. A data analysis application executing on a computing device receives training data and a user input defining an outcome of interest. The data analysis application trains a predictive model with the training data and the outcome of interest. The data analysis application generates input data for each of a set of conditions determined from the training data. The data analysis application receives predicted outcome from the predictive model based on the input data. The data analysis application determines the relevance of the condition based on a comparison of the predicted outcome and the outcome of interest. The data analysis application generates a user segment that comprises a condition from the set of conditions based on the relevance of the condition.