Automatic Feature Grouping for Higher-Order Recommendation Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing recommendation systems face challenges in accurately predicting user behavior due to complex interactions between features, leading to inefficient training times and redundant interactions that compromise model accuracy.
Innovation Solution
Implement automatic feature grouping to identify important higher-order interactions, using multi-order feature interaction sets and non-linear mathematical operations to enhance prediction accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If explicit higher-order interaction model enumerates all higher-order interaction, then feature interaction coverage is improved, but training time overhead and storage overhead increase hugely
Solution Approach 1:
The patent extracts only the important higher-order feature interactions from the complete set of possible interactions. Instead of enumerating all combinations, the system selectively identifies and processes significant interactions, thereby maintaining good interaction coverage while dramatically reducing training time and storage requirements.
Solution Approach 2:
The patent applies partial action by processing only a subset of higher-order interactions rather than all possible interactions. By focusing computational resources on the most important interactions, the system achieves effective feature interaction modeling without the prohibitive cost of exhaustive enumeration.
2Measurement precision
If explicit higher-order interaction model enumerates all higher-order interaction, then feature interaction coverage is improved, but storage overhead increases hugely
Solution Approach 1:
The patent extracts and stores only the important higher-order feature interactions rather than all possible interactions. This selective extraction maintains comprehensive coverage of meaningful interactions while significantly reducing the storage burden by eliminating redundant interaction data.
3Measurement precision
If redundant feature interaction is included in model, then feature interaction coverage is improved, but model optimization difficulty increases and model effect compromises
Solution Approach 1:
The patent extracts and retains only the important feature interactions while removing redundant ones. This selective approach maintains comprehensive coverage of meaningful interactions and reduces model complexity, making optimization easier and improving overall model effectiveness.
Solution Approach 2:
The patent applies partial action by including only the necessary feature interactions in the model rather than all possible interactions. This selective inclusion reduces model complexity and optimization difficulty while maintaining sufficient interaction coverage for accurate predictions.
Data Source
AI summary
This application relates to the field of artificial intelligence. A recommendation method based on automatic feature grouping includes: obtaining a plurality of candidate recommended objects and a plurality of association features of each of the plurality of candidate recommended objects; performing multi-order automatic feature grouping on the plurality of association features of each candidate recommended object, to obtain a multi-order feature interaction set of each candidate recommended object; obtaining an interaction feature contribution value of each candidate recommended object through calculation based on the plurality of association features in the multi-order feature interaction set of each candidate recommended object; obtaining a prediction score of each candidate recommended object through calculation based on the interaction feature contribution value of each candidate recommended object; and determining one or more corresponding candidate recommended objects with a high prediction score as a target recommended object.