Time-Series Product Recommendation Using Evolving Customer Attributes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing product recommendation systems fail to accurately capture time-series changes in customer and product attributes, leading to low precision in recommending products.
Innovation Solution
A product recommendation system that utilizes an estimation model to analyze time-series changes in customer purchase behavior and attributes, employing algorithms like TGFN, STAR, and Netwalk to generate a trained model for precise product recommendations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a typical system for estimating a product to be recommended to a customer is used, then the system structure is simple, but the precision in grasping time-series change features in customer purchase activity is insufficient
Solution Approach 1:
The patent applies dynamics by transitioning from static customer attribute analysis to dynamic time-series analysis of purchase behavior. The system processes sequential purchase records and attribute changes over time, capturing temporal evolution patterns that static systems cannot detect. This enables the model to adapt to changing customer preferences and behaviors dynamically.
Solution Approach 2:
The patent adds a temporal dimension to the recommendation system by incorporating time-series data of purchase behavior and attribute changes. Instead of analyzing single-point customer profiles, the system analyzes trajectories of customer behavior over time, adding the time dimension to the feature space and enabling detection of trends and patterns across multiple time points.
2Measurement precision
If complex purchase factors including time-series changes are considered, then the precision of product recommendation is improved, but the complexity of analysis increases
Solution Approach 1:
The patent introduces an intermediary estimation model that bridges raw time-series purchase behavior data and final product recommendations. This model acts as a mediator that processes complex temporal patterns, customer attribute evolutions, and purchase histories, transforming them into actionable recommendation scores. The intermediary model simplifies the overall system architecture by centralizing the complex analysis in a dedicated component.
Solution Approach 2:
The patent utilizes parameter changes by modeling the evolution of customer attributes and purchase behavior as changing parameters over time. The system tracks how customer preferences, purchasing frequency, and product category interests change as parameters, allowing the recommendation engine to adapt to these parameter evolutions and provide more accurate predictions based on current trends rather than historical averages.
Data Source
AI summary
A product recommendation system is provided with: an estimation model, which expresses a relationship between purchase behavior information pertaining to a subject customer during a first period (expressing a time-series change in product purchase activity on the part of the subject customer) and subject customer attribute information (expressing a time-series change in an attribute of the subject customer), and a product purchase record for the subject customer after the first period; and an estimation unit which, on the basis of purchase behavior information and subject customer attribute information during a second period after the first period, estimates a recommended product for the subject customer after the second period.