Rule Formation Support Apparatus for Target User Preference Modeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In online transactions, existing technologies face challenges in creating rules that accurately reflect the tendencies of specific users due to the overwhelming influence of general user data, making it difficult to tailor recommendations effectively.
Innovation Solution
A rule formation support program and apparatus that compares and integrates the selection frequency data of general users with that of a target user, normalizing and weighting the data to highlight attributes with significant differences, thereby generating a model that reflects the target user's preferences.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If general user data is used to create transaction rules, then the model covers broad user behaviors, but the target user's specific tendencies are overwhelmed and lost
Solution Approach 1:
The patent segments the tendency data into two distinct components: general user tendency data and target user tendency data. By separating these data sources and processing them through different weighting mechanisms, the system preserves both the broad coverage from general data and the precision of individual user tendencies, resolving the contradiction between model coverage and target user accuracy.
Solution Approach 2:
The patent applies local quality by differentiating the treatment of tendency data based on its source. General user data receives one type of processing while target user specific data receives different processing (higher weighting). This localized differentiation ensures that each data source contributes appropriately to the final model, maintaining both overall coverage and individual precision.
2Measurement precision
If only target user data is used to create rules, then the recommendations are highly personalized, but the model lacks robustness from broader user patterns
Solution Approach 1:
The patent merges general user tendency data and target user tendency data into an integrated model. By combining these data sources with appropriate weighting (where target user data has higher weight), the system achieves both personalization accuracy and model robustness, resolving the contradiction between user-specific precision and overall reliability.
Solution Approach 2:
The patent changes the parameter of data weighting dynamically based on the data source. Target user tendency data is assigned a higher weight parameter while general user data receives a lower weight parameter. This parameter adjustment allows the model to prioritize personalization while still incorporating broader patterns for robustness, resolving the contradiction between personalization accuracy and model reliability.
Data Source
AI summary
A process includes obtaining, based on history information of transactions by each user, a first tendency that indicates a selection frequency of each attribute in the transactions by a plurality of users and a second tendency that indicates a selection frequency of each attribute in the transactions by a predetermined target user, specifying an attribute that includes a difference equal to or greater than a predetermined value in the selection frequency of each attribute by comparing the first tendency and the second tendency, and when generating data to be used for one of creating and updating a model that corresponds to a rule of an attribute presented to the target user by integrating the first tendency and the second tendency, performing integration by using the second tendency for the specified attribute.