Item Feature Ranking via Shapley Value Correlation Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Users face difficulty in selecting items with the best trade-off between a determinative characteristic and feature set when browsing electronic interfaces, as the features most strongly linked to user actions are not prominently provided or organized.
Innovation Solution
The method involves determining item features using machine learning models to calculate correlations with determinative characteristics, calculating Shapley values, and organizing the electronic interface based on the most strongly correlated features, which can be displayed or used for filtering.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If all item features are displayed in the electronic interface, then users have complete information for decision-making, but the interface becomes cluttered and navigation becomes difficult
Solution Approach 1:
The patent extracts and highlights only the most relevant features (those with highest Shapley values) for prominent display, while keeping other features accessible but less prominent. This selective extraction resolves the contradiction by presenting essential information upfront without overwhelming the interface.
Solution Approach 2:
The patent applies different display qualities to different features based on their relevance scores. High-relevance features receive prominent positioning and visual emphasis, while lower-relevance features are displayed in less prominent locations. This local differentiation maintains information completeness while improving navigation ease.
2Measurement precision
If multiple machine learning models are used to calculate feature correlations, then correlation accuracy improves, but computational complexity and processing time increase
Solution Approach 1:
The patent combines multiple machine learning models (e.g., linear regression, tree-based algorithms) to leverage their complementary strengths. Each model contributes to the overall correlation assessment, with results aggregated through Shapley value calculations. This merging approach improves measurement precision while distributing computational load across specialized models.
Solution Approach 2:
The patent calculates Shapley values to determine the marginal contribution of each feature, focusing computational resources on features with highest impact. Rather than uniformly processing all features with equal depth, the system applies partial action by prioritizing calculations for most influential features, thereby improving accuracy where it matters most while controlling overall complexity.
3Measurement precision
If features are organized based on Shapley values from machine learning models, then user selection accuracy improves, but computational processing time increases
Solution Approach 1:
The patent performs preliminary computation of Shapley values and feature relevance scores during off-peak periods or in batch processing mode. These pre-computed relevance metrics are then reused for multiple user queries, improving user selection accuracy while amortizing the computational time cost across many requests rather than calculating everything in real-time for each user.
Solution Approach 2:
The patent dynamically adjusts the granularity of Shapley value calculations based on query characteristics and system load. For simple queries or high-load periods, it uses pre-computed or coarser-grained feature importance metrics. For complex queries or low-load periods, it performs more detailed calculations. This parameter adjustment balances selection accuracy with processing time requirements.
Data Source
AI summary
A method of determining item features and organizing an interface according to the features includes determining a plurality of features of a plurality of items, the plurality of items accessible through an electronic interface, applying a plurality of machine learning models to the determined features, wherein each of the machine learning models calculates a correlation of each feature to characteristic determinative of user selection on the electronic interface, calculating respective Shapley values of each correlation determined by each of the plurality of machine learning models, determining one or more of the item features that are most strongly correlated with the determinative characteristic according to the respective Shapley values, and causing the electronic interface to be organized according to the determined most strongly correlated item features.