Modified Self-Similarity for Repeat-Purchase Product Recommendations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional collaborative filtering recommendation systems are ineffective in recommending products that a consumer has already purchased, as they cannot incorporate repeat-purchase information and distinguish between frequently repurchased and never repurchased products, leading to poor recommendation performance for items like diapers, bottled water, and vitamin supplements.
Innovation Solution
A system and method that utilize modified self-similarity calculations, including repurchase rate, preference, and weight assignment, to calculate a modified self-similarity value for products, allowing for the recommendation of products already purchased by a consumer based on their repeat-purchase propensity, by incorporating repurchase rate, preference, and weight into the collaborative filtering algorithm.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional collaborative filtering recommendation technique is used, then new products can be recommended based on consumer preferences, but products that consumers have already purchased cannot be recommended
Solution Approach 1:
The patent modifies the self-similarity parameter in collaborative filtering to distinguish between products purchased once and products repurchased multiple times. By introducing a modified self-similarity calculation that incorporates repurchase rate information, the system changes the parameter values for purchased products based on their repurchase history, enabling the system to recommend both new and repeat-purchased products appropriately.
Solution Approach 2:
The patent makes the recommendation system dynamic by allowing the self-similarity values to change based on repurchase behavior over time. The modified self-similarity calculation dynamically adjusts product similarity scores based on actual repurchase rates, enabling the system to adapt to changing consumer behaviors and recommend products based on their repeat-purchase propensity.
2Ease of operation
If self-similarity value of 1 is assigned to all purchased products, then all purchased products are treated equally, but the difference in repurchase frequency between products cannot be captured
Solution Approach 1:
The patent changes the self-similarity parameter from a fixed value of 1 to a variable modified self-similarity value that incorporates repurchase rate information. This allows the system to measure and distinguish between products with different repurchase propensities while maintaining the collaborative filtering framework. The modified self-similarity calculation integrates multiple parameters including repurchase rate, preference, and weight to achieve precise measurement.
Solution Approach 2:
The patent creates a composite self-similarity metric by combining multiple factors: repurchase rate, consumer preference, and weight parameters. This composite approach allows the system to capture the nuanced differences in repurchase behavior by synthesizing multiple data dimensions into a single modified self-similarity value, achieving both measurement precision and operational feasibility.
3Device complexity
If conventional collaborative filtering is used for consumable products, then the basic recommendation structure is simple, but repeat-purchase recommendation performance is poor
Solution Approach 1:
The patent enhances the collaborative filtering algorithm by modifying the self-similarity parameter to include repurchase rate information specific to consumable products. This parameter change enables the system to capture repeat-purchase patterns without fundamentally redesigning the entire algorithm structure, maintaining simplicity while improving accuracy for products like diapers, bottled water, and vitamin supplements.
Solution Approach 2:
The patent introduces feedback mechanisms by incorporating actual repurchase rate data from consumer behavior into the self-similarity calculation. This feedback loop allows the system to continuously learn from actual repurchase patterns and adjust product recommendations accordingly, improving accuracy for consumable products where repeat purchases are common.
Data Source
AI summary
Disclosed herein are a system and method for recommending repeat-purchase products using modified self-similarity. The system is operated by a computer in an environment in which user equipment (UE), a database, and a web server are connected over a network. The system includes: a user information reception unit configured to receive user information from the database when the UE connects with the web server; a modified self-similarity calculation unit configured to calculate a modified self-similarity (MSS) using the user information received by the user information reception unit; a similarity calculation unit configured to calculate the sums of similarities between products by substituting the modified self-similarity, calculated by the modified self-similarity calculation unit, into an inter-product similarity calculation matrix; and a product recommendation unit configured to align the sums of similarities between products calculated by the similarity calculation unit and to transmit product information fulfilling preset criteria to the UE.