Demand-Supply Agent Pair Recommendation via Deep Neural Network Vectorization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for matching demand and supply agents in B2B transactions face challenges such as the 'cold start' problem, where historical transaction data is lacking, leading to an intractable search space and poor candidate generation, especially when evaluating performance in cold start scenarios.
Innovation Solution
A system utilizing deep neural networks to produce demand agent vectors and select candidate supply agents based on future transaction likelihood, combined with k-nearest neighbor analysis and machine learning modeling to recommend demand-supply agent pairs, effectively reducing the search space and addressing data imbalance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If brute force approach is used to create records for every customer-partner combination, then complete candidate generation is achieved, but search space becomes intractable and computational complexity increases
Solution Approach 1:
The patent segments the candidate generation process into multiple stages: initial candidate generation based on limited features, then progressive refinement through additional features and machine learning models. This divides the intractable brute force search into manageable segments that can be processed sequentially, reducing overall computational complexity while maintaining completeness.
Solution Approach 2:
The patent performs preliminary candidate generation using available historical transaction data and basic features before applying complex machine learning models. This preliminary filtering reduces the search space early in the process, avoiding the need to evaluate all possible customer-partner combinations from scratch and significantly reducing computational burden.
2Measurement precision
If historical transaction data is used for matching, then accuracy improves, but cold start problem occurs when no historical data is available
Solution Approach 1:
The patent introduces intermediary elements such as industry categories, product categories, and intermediary scoring mechanisms that bridge the gap when direct historical transaction data is unavailable. These intermediaries allow the system to make informed recommendations for cold start scenarios by leveraging indirect signals and patterns from related entities, rather than relying solely on direct historical data.
Solution Approach 2:
The patent dynamically adjusts model parameters and data weighting based on data availability. When historical transaction data is scarce (cold start), the system shifts weight to alternative features like company profiles, industry relationships, and intermediary signals. As more historical data becomes available, the system transitions to relying more on direct transaction patterns, thus adapting to different data regimes.
3Measurement precision
If classification algorithms are used with historical transaction data, then existing patterns are captured, but cold start, candidate generation and imbalance problems persist
Solution Approach 1:
The patent employs a composite approach by combining multiple types of data sources and model types. Instead of relying on a single classification algorithm, the system integrates historical transaction data, company profiles, industry relationships, and intermediary signals into a unified framework. This composite structure allows the system to leverage pattern recognition from classification algorithms while simultaneously addressing cold start and candidate generation through diverse data sources and multiple modeling strategies.
Data Source
AI summary
A system and method for recommending demand-supply agent pairs for transactions uses a deep neural network on data of demand agents to produce a demand agent vector, which is used to select supply agents based on their likelihood of future transaction and to find k nearest neighbor demand agents for each of the demand agents. The candidate supply agents and the k nearest neighbor demand agents are then combined to produce candidate demand-supply agent pairs, which are used to find recommended demand-supply agent pairs by applying modeling using machine learning.