Recommendation Engine Using Customer Household Graphs for Sparse Data

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing recommendation engines struggle to provide personalized recommendations to customers without prior purchase history, especially when customers do not enroll in loyalty programs or pay with cash, leading to sparse data and scalability issues.

Innovation Solution

A system that generates customer household graphs using transaction data from a data repository, including enterprise and customer ontologies, to infer user nodes and user groups, allowing for recommendations based on product metadata, spatial, and temporal data, even without prior purchase history.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional recommendation engines use prior purchase history and loyalty program data, then recommendation accuracy is improved, but coverage of customers without purchase history is lost

Engineering Contradiction:

Improverecommendation accuracyVSAvoidcustomer coverage

Core Design Contradiction:

Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The system performs preliminary actions by pre-processing transaction data to extract product metadata, categories, and relationships before actual recommendation generation. This allows the system to build knowledge bases and clustering models in advance, enabling accurate recommendations even for customers without prior purchase history by leveraging pre-computed product relationships and patterns.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent introduces intermediary structures including product metadata layers, clustering models, and graph representations that mediate between raw transaction data and final recommendations. These intermediaries transform individual customer transactions into generalized patterns and insights, enabling the system to infer preferences for customers without direct purchase history through their associated product relationships.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Ease of operation

If the system processes detailed transaction data for each customer, then recommendation personalization is improved, but system complexity and data processing requirements increase

Engineering Contradiction:

Improverecommendation personalizationVSAvoidsystem complexity

Core Design Contradiction:

Ease of operationVSDevice complexity

Solution Approach 1:

The system segments the complex task of personalized recommendation into distinct modules: data extraction, product metadata generation, clustering model processing, and recommendation generation. Each module handles a specific aspect of the data processing pipeline independently, reducing overall system complexity while maintaining personalization capabilities through coordinated operation of these segmented functions.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent implements universal data structures and processing frameworks that serve multiple functions. The same product metadata extraction and clustering mechanisms work for both customers with and without purchase history, and the graph representation system handles various data types (transactional, product, customer) uniformly, reducing system complexity through multi-functionality.

Inventive Principle:

Principle #6Universality (Multi-functionality)

3Productivity

If the system uses clustering and graph representations to aggregate data, then scalability is improved, but data processing time increases

Engineering Contradiction:

ImprovescalabilityVSAvoiddata processing time

Core Design Contradiction:

ProductivityVSLoss of time

Solution Approach 1:

The system performs preliminary clustering and graph construction on product and transaction data before actual recommendation queries. By pre-computing cluster assignments, product relationships, and graph structures in advance, the system reduces real-time processing requirements during recommendation generation, improving scalability without significant time loss during actual customer interactions.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent applies partial clustering and graph processing only where necessary rather than processing all data uniformly. The system processes detailed data for high-value segments (customers with purchase history) while using aggregated cluster information for broader segments (customers without history), reducing overall processing time while maintaining scalability through selective application of computational intensity.

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS10445811B2Recommendation engine comprising an inference module for associating users, households, user groups, product metadata and transaction data and generating aggregated graphs using clustering

Publication Date: 2019.10.15 TATA CONSULTANCY SERVICES LTD

AI summary

A method for providing recommendations for customers is described. The method comprises obtaining transaction data associated with one or more products being purchased by a customer. The method further comprises generating a customer household graph, wherein the customer household graph comprises a customer household node associated with the customer, a product superset node, and one or more product nodes associated with the product superset node. Further, at least one product metadata node associated with each of the one or more product nodes is determined. Further, at least one user node is associated with the customer household node based on the at least one product metadata node. Further, a user group node associated with the customer household node is determined based on the at least one user node. Further, one or more recommendations to be provided to the user are ascertained.