Shapley Value Vector Clustering for Decision Tree Drift Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional model drift detection methods rely on ground truth, original feature space, and distribution output scores, making it difficult to interpret and accurately monitor the performance of machine learning models over time, leading to inaccurate predictions in image classification, text classification, and data mining tasks.
Innovation Solution
Converting training data and new inputs into input vectors represented by Shapley values, clustering the data, and detecting the drift degree based on these vectors and clusters to determine changes in feature relationships, thereby ensuring accurate model performance monitoring.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional model drift detection methods rely on ground truth, original feature space, and distribution output scores, then model performance can be monitored, but the interpretation becomes difficult and accuracy deteriorates over time
Solution Approach 1:
The patent introduces Shapley values as an intermediary representation that bridges the gap between complex model internals and interpretable drift detection. By converting input features into Shapley value vectors, the method enables accurate drift measurement while maintaining interpretability through the clustering process that operates in this transformed feature space.
Solution Approach 2:
The patent transforms the feature space by converting original input features into Shapley value representations. This parameter transformation allows the drift detection to operate on a different representation that preserves accuracy while improving interpretability through the subsequent clustering analysis in the transformed space.
2Reliability
If model drift is not detected, then model performance monitoring is maintained, but inaccurate predictions occur in image classification, text classification, and data mining tasks
Solution Approach 1:
Shapley values serve as an intermediary that simplifies the measurement of model drift by transforming complex feature interactions into interpretable contribution values. This intermediary representation makes drift detection more straightforward while ensuring prediction accuracy through the clustering-based drift degree calculation.
Solution Approach 2:
The patent replaces traditional drift detection mechanisms that rely on ground truth and distribution scores with a Shapley value-based approach. This substitution simplifies the detection process by using game-theoretic feature contribution measures instead of complex statistical comparisons, reducing detection difficulty while maintaining reliability.
Data Source
AI summary
In a method for detecting a model drift provided in an illustrative embodiment of the present disclosure, training data in a training data set is converted into an input vector represented by Shapley values. A plurality of dimensions of the input vector indicates a plurality of input features of a decision tree model. The decision tree model has been trained for performing at least one of image classification, text classification, or data mining. The method also includes: clustering, on the basis of the input vector, the training data set, so as to obtain a plurality of data clusters. The method also includes: in response to receiving a first input, converting the first input into a first input vector represented by Shapley values. The method also includes: detecting a drift degree of the decision tree model on the basis of the first input vector and the plurality of data clusters.