SVM Feature Contribution Reason Codes for Faster Model Interpretation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning-based classification algorithms like SVM face challenges in providing interpretable feature contributions and reason codes that are both model-specific and human-understandable, especially due to the computational intensity of methods like SHAP and the lack of clarity in conventional explanations from LIME and SHAP algorithms.
Innovation Solution
A system and method that compute feature contributions and generate human-interpretable reason codes by deriving support vector contributions, clustering similar data points, and building trepan or decision trees to provide model-specific explanations for SVM models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If SHAP algorithm is used to compute feature contributions, then global and local explanations are provided, but the computation becomes exponential in running time
Solution Approach 1:
The patent segments the feature contribution computation by identifying and utilizing support vectors from the trained SVM model. Instead of computing SHAP values for all possible data combinations (exponential complexity), the method focuses computations only on support vectors, which are the critical data points defining the decision boundary. This segmentation reduces the computational scope while maintaining explanation quality.
Solution Approach 2:
The patent performs preliminary action by pre-identifying support vectors during the SVM training phase. These support vectors are stored and reused for explanation generation, eliminating the need to recompute complex SHAP values from scratch for each explanation query. The support vector identification is done once during training, and subsequent explanations leverage this pre-computed information.
2Ease of operation
If LIME algorithm is used to provide local explanations, then explanations are generated quickly, but the explanations are not easily understood by users
Solution Approach 1:
The patent introduces an intermediary layer that translates the mathematical support vector contributions into human-interpretable reason codes. This intermediary translation layer converts complex SVM decision mathematics into natural language explanations that users can easily understand, while maintaining the computational efficiency of working with support vectors directly.
3Adaptability or versatility
If model-agnostic methods like SHAP and LIME are used, then explanations can be applied to any model, but the explanations do not capture model-specific mathematical decisions
Solution Approach 1:
The patent applies local quality by tailoring the explanation method specifically to SVM models while maintaining the ability to handle different kernel types (linear, polynomial, RBF). The explanation leverages SVM-specific mathematical properties, such as support vectors and kernel functions, to provide accurate model-specific explanations. This approach can be adapted to other models by identifying their unique mathematical characteristics.
4Loss of information
If conventional feature contribution methods are used, then coefficients are provided for every feature, but the interpretations are difficult to understand by users
Solution Approach 1:
The patent introduces an intermediary translation layer that converts the mathematical support vector contributions into human-interpretable reason codes. This intermediary translation layer explains the logical steps and reasoning used by the SVM model in natural language, making the feature contribution information accessible and understandable to end users without requiring them to understand complex mathematical formulations.
Data Source
AI summary
Systems and methods for computing feature contribution and providing hum-interpretable reason codes for a Support Vector Machine (SVM) model are disclosed, A system computes, for each data point from amongst plurality of data points indicative of plurality of features, a feature contribution of each one of the plurality of features for a SVM model used for at least one of classification decision and a regression analysis. Further, the system provides human interpretable reason code for the interpretation corresponding to at least one of classification decision and the regression analysis from the SVM model. The system outputs to the user, a feature contribution output, and the human interpretable reason codes output. The feature contribution output and the human interpretable reason codes output are indicative of an acceptable decision to be taken by the user based on the classification decision and the regression analysis received from the SVM model.