Retail Recommendation Domain Model for Granular Data Normalization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional data mining techniques for generating retail recommendations lack granularity, leading to limited pattern correlation and increased complexity as the number and complexity of association rules grow, especially when dealing with complex products and configurations, and require substantial programming effort and maintenance, making them inflexible and costly.
Innovation Solution
A system and method using a recommendation domain model to normalize customer purchase information, catalog data, and historical transaction data, enabling the generation of targeted selling points by transforming data into a shared domain representation that captures key concepts and attributes, allowing for more accurate and flexible association rule generation independent of SKU references.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional data mining techniques use generic item descriptions (SKU level) to identify frequent sets, then the frequent sets are not large and power-set/rule generation is tractable, but the lack of granularity diminishes the quality of association rules and results in limited pattern correlation
Solution Approach 1:
The patent segments the item description into multiple hierarchical levels: generic item level (for tractable rule generation) and detailed attribute level (for high-quality associations). This segmentation allows the system to work with both coarse-grained SKUs for computational efficiency and fine-grained attributes for pattern discovery, resolving the contradiction between granularity and complexity.
Solution Approach 2:
The patent adds a new dimension to item representation by introducing attribute-based descriptions alongside traditional SKU identifiers. This dimensional expansion allows the system to maintain the computational simplicity of SKU-level processing while simultaneously enabling detailed pattern correlation through attribute-level analysis, effectively resolving the granularity-complexity tradeoff.
2Reliability
If the number and complexity of mined association rules increases to provide better recommendations, then recommendation quality improves, but programming complexity and maintenance effort increase substantially
Solution Approach 1:
The patent implements self-service through automated attribute extraction and association rule generation. The system automatically extracts attributes from transaction data, builds the item attribute model, and generates association rules without requiring substantial manual programming. This automation maintains high recommendation quality while dramatically reducing programming complexity and maintenance burden.
Solution Approach 2:
The patent changes the parameters of association rule generation by shifting from traditional item-based rules to attribute-based rules. This parameter change allows the system to generate meaningful associations even with moderate data volumes, improving recommendation quality without proportionally increasing system complexity or maintenance requirements.
3Extent of automation
If conventional configuration systems are developed to automate configuration selection, then recommendation automation improves, but the expense of creating and maintaining such systems increases substantially
Solution Approach 1:
The patent creates a universal item attribute model that can be applied across multiple products and services within a vendor's portfolio. This universal model serves multiple functions: it enables association rule generation, supports configuration recommendations, and facilitates cross-selling opportunities. By making the system multi-functional, the patent reduces the need for separate configuration systems for different products, thereby lowering overall complexity and maintenance costs while maintaining high automation levels.
Data Source
AI summary
A data processing system normalizes data sets (such as low-resolution transaction data) into high-resolution data sets by mapping generic information into attribute-based specific information that is stored in a database. By establishing a shared domain model for representing items in the recommendation context, catalog and quote history with common terms and concepts, a recommendation engine operating in the shared domain may process the attribute-based representations to make specific and relevant recommendations to the customer. In addition, when certain attribute values are normalized over time, recommendations derived from past order history can be intelligently applied to current orders. The normalized representation of elements in the shared domain may also be used to generate compelling selling point text for each recommendation that is specific to the marketing objectives of the seller and identifies the objectives of the buyer.