Inventory Prediction Using Cross-Category Graph Neural Networks

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional inventory management systems are unable to identify missing items in an individual's inventory list, as they rely on comprehensive and complete data that is often incomplete or unavailable for non-commercial entities, such as individuals, who do not maintain detailed records of their possessions.

Innovation Solution

An inventory prediction system utilizing a cross-category directional graph, where items with similar attributes are grouped into category nodes, and connections between nodes are weighted based on user data, allowing for the prediction of missing items and generation of listing recommendations using machine learning models.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional inventory systems rely on comprehensive current inventory information as ground truth, then prediction accuracy is improved, but the system cannot identify missing items in incomplete inventory lists

Engineering Contradiction:

Improveprediction accuracyVSAvoidmissing items in inventory list

Core Design Contradiction:

Measurement precisionVSLoss of information

Solution Approach 1:

Instead of using current inventory as ground truth to predict future needs, the system inverts the approach by using historical transaction data and graph-based relationships to predict what items should currently be in inventory, thereby identifying missing items through the discrepancy between predicted and actual inventory lists

Inventive Principle:

Principle #13The other way round (Inversion)

Solution Approach 2:

The patent introduces a graph-based intermediary structure where items are nodes and relationships are edges, with a graph neural network serving as a mediator to propagate information across the graph. This intermediary enables the system to infer missing items by leveraging relationships between items even when direct inventory data is incomplete

Inventive Principle:

Principle #24Intermediary (Mediator)

2Loss of time

If conventional inventory systems use historical transaction information to forecast demand, then future inventory needs can be predicted, but the systems cannot identify items missing from current inventory records

Engineering Contradiction:

Improveresponse time for inventory predictionVSAvoidcompleteness of inventory list

Core Design Contradiction:

Loss of timeVSLoss of information

Solution Approach 1:

The system performs preliminary action by building a comprehensive graph structure from historical transaction data in advance, encoding item relationships and co-occurrence patterns before prediction is needed. This pre-computed graph structure enables rapid prediction while capturing complex relationships that would be difficult to compute in real-time

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent replaces traditional mechanical inventory tracking systems with a data-driven graph neural network approach. Instead of relying on physical inventory records and simple forecasting algorithms, the system uses machine learning models that process graph-structured data to infer missing items, substituting computational intelligence for conventional inventory management mechanics

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

3Quantity of substance

If detailed inventory records are maintained for all items, then complete inventory information is available, but non-commercial entities such as individuals cannot maintain such comprehensive records

Engineering Contradiction:

Improvecompleteness of inventory dataVSAvoidease of maintaining inventory records

Core Design Contradiction:

Quantity of substanceVSEase of operation

Solution Approach 1:

The system enables self-service by automatically generating complete inventory predictions without requiring users to manually maintain detailed records. The graph neural network autonomously infers missing items based on available data and learned relationships, eliminating the burden of comprehensive record-keeping while still achieving complete inventory information

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The patent creates a universal system that works for both commercial and non-commercial entities. The graph-based approach and neural network model are domain-agnostic, able to handle inventory prediction for businesses with transaction systems as well as individuals with minimal records, making the solution universally applicable across different entity types

Inventive Principle:

Principle #6Universality (Multi-functionality)

Data Source

PatentUS20240394625A1Inventory Item Prediction and Listing Recommendation

Publication Date: 2024.11.28 EBAY INC

AI summary

An inventory prediction system is described that outputs a predicted inventory item not included in a user's known inventory using a cross-category directional graph that represents item categories as nodes. The inventory prediction system implements a prediction model trained using machine learning to output the predicted inventory item using the graph and at least one item from the user's known inventory. The inventory prediction system is further configured to generate a listing recommendation for the predicted inventory item. To do so, the inventory prediction system implements a logistic regression model trained using machine learning to calculate a probability that the listing recommendation should be generated using attributes of the predicted inventory item and attributes of currently trending items. The listing recommendation is generated to include a description of, and estimated value for, the predicted inventory item, together with an option to generate a sale listing for the predicted inventory item.