Engineering Module Recommendation Using Technical Attribute Embeddings
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional recommender systems face a sparsity problem in industrial engineering projects due to users interacting with non-overlapping sets of modules, leading to ineffective recommendations.
Innovation Solution
A hybrid recommender system that embeds items based on technical attributes into a low-dimensional vector space, using a graph neural network to compute user and item representations, enabling recommendations across different user groups by leveraging shared technical knowledge.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If collaborative filtering recommender systems are used to recommend modules to users, then recommendations can be made based on user behavior patterns, but the system fails when users interact with non-overlapping sets of modules leading to data sparsity
Solution Approach 1:
The patent introduces technical attributes as an intermediary between users and modules. Instead of directly comparing user interaction patterns (which fail when users interact with different modules), the system uses module technical attributes as a mediator to find recommendations. This allows users who interact with different modules to still receive relevant recommendations based on the technical characteristics of the modules rather than requiring overlapping interaction histories.
Solution Approach 2:
The patent replaces the mechanical collaborative filtering approach (which relies on direct user interaction pattern matching) with a content-based recommendation approach using technical attributes. This substitution allows the system to function effectively even when user interaction sets do not overlap, as it bases recommendations on the intrinsic technical properties of modules rather than on user behavior correlations.
2Adaptability or versatility
If users work with different suppliers or design assets with varying applications, then system adaptability increases, but the sparsity problem worsens as common configurations between users decrease
Solution Approach 1:
The patent creates a universal recommendation mechanism that works across different user groups, suppliers, and application domains by using technical attributes as a common language. The technical attributes provide a universal framework that can represent modules from different suppliers and applications in a standardized way, allowing the recommendation system to function universally across diverse contexts without requiring common interaction patterns.
Solution Approach 2:
The patent changes the fundamental parameters used for recommendation from user interaction patterns to module technical attributes. This parameter change allows the system to adapt to different users, suppliers, and applications by focusing on the intrinsic technical characteristics of modules rather than on user-specific interaction histories, thereby maintaining recommendation quality across diverse and sparse data conditions.
3Ease of operation
If conventional recommender systems are used in industrial engineering configurators, then personalized information filtering can be provided, but the systems cannot make reasonable recommendations when user item sets do not intersect
Solution Approach 1:
The patent introduces technical attributes as an intermediary that enables information filtering without requiring direct user interaction pattern matching. By using technical attributes as a mediator, the system can provide personalized information filtering based on module characteristics rather than user behavior, ensuring reliable recommendations even when users interact with completely different module sets.
Solution Approach 2:
The patent substitutes the mechanical collaborative filtering mechanism with a content-based approach using technical attributes. This replacement maintains the ease of operation for information filtering while eliminating the reliability problem that occurs when user item sets do not intersect, as the recommendation basis shifts from user interactions to module technical properties.
Data Source
AI summary
A database stores a set of items, with each item having technical attributes, and with each item representing a module that can be used in an engineering project of a first user, u1. A feature encoder embeds each item based on its technical attributes into a low-dimensional vector space. Then, in a second step, a graph neural network pools over these item embeddings to compute an updated user embedding for the first user A decoder mapping then addresses the recommendation task by outputting recommendation scores for each item. That means, heuristically speaking, that the method and system lift the recommendation task to the level of technical attributes to overcome the sparsity problem caused by item sets that are not overlapping between user groups. Thus, when matching similar users, the method does not rely on users configuring exactly the same modules but rather on configured modules that are similar from a technical point of view.