Machine Learning Item Substitution Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Retailers face challenges in stocking the right assortment of goods due to space constraints and limited data on low-velocity items, leading to missed sales opportunities and a lack of suitable substitutes for out-of-stock items.
Innovation Solution
The use of trained machine learning processes, specifically leveraging Bidirectional Encoder Representations from Transformers (BERT) models, to predict item substitutions by generating features from product descriptions and substitution scores, allowing for the identification of suitable substitutes for low-velocity items based on their similarity to high-velocity items.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If variety in product assortment is increased, then customer preferences are better met, but space constraints make it difficult to stock all beneficial items
Solution Approach 1:
The patent introduces machine learning models as an intermediary system that predicts item substitutes and demand patterns. This mediator enables the retailer to stock fewer physical items while maintaining service quality by automatically suggesting substitutes from a broader catalog, thus resolving the conflict between assortment variety and store space constraints
Solution Approach 2:
The system creates virtual copies of the full product catalog through AI-generated substitute recommendations. Instead of physically stocking every possible item, the retailer maintains a comprehensive digital catalog that the machine learning model references to suggest substitutes, effectively copying the benefits of high variety without the physical space requirements
2Adaptability or versatility
If low-velocity items are stocked, then customer preferences are better met, but there is limited data available for prediction
Solution Approach 1:
The system performs preliminary actions by training machine learning models on high-velocity items with abundant data first. These pre-trained models learn substitution patterns and semantic relationships that can then be applied to low-velocity items with limited data, enabling predictions before sufficient historical data accumulates
Solution Approach 2:
The machine learning models serve multiple functions: they analyze high-velocity items with rich data, learn general substitution patterns, and then apply these learned patterns to low-velocity items with limited data. This multi-functional approach allows the same system to handle both data-rich and data-poor scenarios effectively
3Measurement precision
If machine learning models are trained on high-velocity items, then prediction accuracy improves, but the models must be adapted for low-velocity items
Solution Approach 1:
The system changes the parameter of data availability by training on high-velocity items with abundant data to achieve high accuracy, then applies the same model to low-velocity items by changing the input data parameter. The model architecture and training approach remain consistent, but the data characteristics vary, allowing the system to maintain accuracy across different item types
4Productivity
If substitute items are identified for out-of-stock items, then sales are maximized, but the complexity of the substitution system increases
Solution Approach 1:
The machine learning system performs self-service by automatically generating substitute recommendations without human intervention. The model independently analyzes product descriptions, learns substitution patterns, and generates recommendations, eliminating the need for complex manual curation systems while maintaining high sales productivity
Data Source
AI summary
This application relates to employing trained machine learning processes to determine item substitutions, such as item substitutions for low-velocity items. For example, a computing device may generate features based on item data for a pair of low-velocity items. The computing device may apply a trained machine learning process to the generated features to determine a substitution score between the pair of low-velocity items. In some examples, and based on the substitution scores, the computing device may rank the low-velocity items. The computing device may receive a request for substitute items for one of the low-velocity items, and may transmit an indication of one or more of the other low-velocity items based on the ranking. In some examples, the computing device trains the machine learning process based on item data for high-velocity items and substitute scores between the high-velocity items.