Dynamic Product Recommendation Mixing Ratio
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional product recommendation systems fail to maintain user interest by displaying the same recommended products repeatedly, especially when users do not make new purchases, leading to a lack of freshness and decreased purchasing intention.
Innovation Solution
A product recommendation method that identifies relevant products based on user data and tags, extracts candidate groups of products, and determines whether they meet a threshold condition to display a diverse set of recommended products, ensuring a fresh lineup by mixing products from the candidate group with others when the threshold is not met.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the recommendation system displays products based on user purchase history and tags, then the recommendation accuracy is improved, but the diversity and freshness of recommended products deteriorates
Solution Approach 1:
The system dynamically adjusts the composition of recommended products by introducing a mixing ratio parameter. When user purchase history is available, it uses tags from purchased products to extract candidate products, but mixes them with other products at a controlled ratio (e.g., 70% candidate products, 30% other products). This dynamic adjustment maintains recommendation accuracy while ensuring product diversity and freshness in the displayed recommendations.
2Adaptability or versatility
If the system extracts tags from purchased products to identify candidate products, then the personalization of recommendations is improved, but the repetitiveness of displayed products increases
Solution Approach 1:
The system changes the parameter of product composition by introducing a mixing ratio that controls the proportion of candidate products (extracted from user purchase history) versus other products. This parameter adjustment allows the system to maintain personalization through tag-based extraction while controlling repetitiveness by limiting the proportion of candidate products and mixing in other diverse products to refresh the recommendation lineup.
3Measurement precision
If the recommendation system relies heavily on user purchase data, then the relevance of recommended products is improved, but the system fails when users do not make new purchases
Solution Approach 1:
The system prepares for the scenario where users do not make new purchases by beforehand establishing a mixing mechanism that combines candidate products (from purchase history) with other products. This cushioning approach ensures that even when purchase data is stale or unavailable, the system can still provide relevant recommendations by relying on the mixed portfolio, thus maintaining system reliability and continued product relevance.
Data Source
AI summary
A product recommendation method executed by a computer, the computer comprising: a memory storing a program; and at least one processor that executes the program, wherein the at least one processor acquires, for plurality of products, product data including a tag expressing a feature of each product; identifies a relevant product that is relevant to a user in response to a display request or an update request of a predetermined screen from the user; searches the product data and extracts the tag attached to the relevant product; extracts, based on the extracted tag, a candidate group of products to be recommended to the user; and determines whether the extracted candidate group satisfies a predetermined condition in order to identify a product to display on the predetermined screen.