Cluster-Based Recommendation Filtering for Purchase History
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing recommendation systems face challenges in accurately identifying user preferences without explicit input, often including items purchased as gifts or by multiple users sharing a computer account, leading to non-tailored recommendations due to the burden of manual collection management and changing user interests.
Innovation Solution
Implementing item clustering techniques to analyze purchase histories, exclude outlier items, and allow users to manage collections at a cluster level, with clustering algorithms selecting recommendation sources and filtering out items representing poor recommendations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If users manually review and edit their purchase histories item-by-item, then recommendation accuracy improves, but user burden and time required increase significantly
Solution Approach 1:
The patent segments the purchase history into clusters of items based on similarity characteristics. Instead of presenting individual items for review, the system groups related items into clusters, allowing users to manage preferences at a higher level of abstraction. This segmentation reduces the number of individual items users must review while maintaining recommendation accuracy.
Solution Approach 2:
The system automatically performs clustering of purchase history items without requiring user intervention. The clustering algorithm autonomously organizes items based on their characteristics, and the system uses these clusters to generate recommendations. This self-service approach eliminates the need for manual item-by-item review while still achieving accurate recommendations.
2Quantity of substance
If the system includes all items from purchase history as recommendation sources, then recommendation coverage increases, but recommendation relevance decreases due to gift purchases and multi-user accounts
Solution Approach 1:
The patent segments the purchase history into distinct clusters based on item similarity and purchasing patterns. By analyzing characteristics such as item type, category, and purchasing context, the system divides the complete purchase history into meaningful groups. This segmentation enables the system to identify which clusters represent genuine user interests versus gift purchases or multi-user accounts, allowing selective use of clusters as recommendation sources while maintaining both coverage and relevance.
Data Source
AI summary
Computer-implemented processes are disclosed for clustering items and improving the utility of item recommendations. One process involves applying a clustering algorithm to a user's collection of items. Information about the resulting clusters is then used to select items to use as recommendation sources. Another process involves displaying the clusters of items to the user via a collection management interface that enables the user to attach cluster-level metadata, such as by rating or tagging entire clusters of items. The resulting metadata may be used to improve the recommendations generated by a recommendation engine. Another process involves forming clusters of items in which a user has indicated a lack of interest, and using these clusters to filter the output of a recommendation engine. Yet another process involves applying a clustering algorithm to the output of a recommendation engine to arrange the recommended items into cluster-based categories for presentation to the user.