Graph-Based Asset Recommendations from Client Trading Relationships
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for providing and recommending alternative assets to clients are manual, time-consuming, and inefficient, relying on criteria like maturity, rating, sector, and currency, without leveraging graph relationships for automated recommendations.
Innovation Solution
A computer program monitors transactions, updates a heterogeneous graph with asset and client data, trains a graph model, queries it for recommendations, and outputs a ranked list of clients likely to trade specific assets, using graph neural networks and natural language processing to enhance prediction accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If manual filtering and distribution methods are used, then sales can identify alternative assets based on criteria, but the process is time-consuming and inefficient
Solution Approach 1:
The patent replaces the manual mechanical process of filtering and distributing assets with an automated graph neural network system. The GNN model automatically processes transaction data from messaging interfaces, builds heterogeneous graphs representing client-asset relationships, and generates recommendations without manual intervention, thereby substituting the mechanical manual workflow with an automated computational system
Solution Approach 2:
The system enables self-service by automatically monitoring messaging interfaces for transactions, updating the heterogeneous graph with new data, training the graph model, and generating recommendations without requiring sales personnel to manually perform these tasks. The system serves itself by continuously learning from new transaction data and autonomously producing updated recommendations
2Extent of automation
If sales manually create client lists based on daily axed assets, then recommendations can be sent, but the process lacks automation and precision
Solution Approach 1:
The patent introduces a heterogeneous graph as an intermediary structure between raw transaction data and final recommendations. The graph serves as a mediator that organizes complex client-asset relationships, captures nuanced interactions through named entity recognition, and enables the graph neural network to process information in a structured manner, thereby managing system complexity through intermediate representation
Solution Approach 2:
The system segments the recommendation process into distinct modular components: (1) monitoring messaging interfaces for transactions, (2) extracting entities via named entity recognition, (3) updating the heterogeneous graph structure, (4) training the graph neural network model, and (5) generating recommendations. This segmentation allows each component to be independently developed, maintained, and optimized
3Measurement precision
If traditional filtering criteria are used, then assets can be categorized, but graph relationships and trading history are not leveraged for predictions
Solution Approach 1:
The system performs preliminary action by continuously pre-processing and storing transaction data in the heterogeneous graph structure before recommendations are needed. The graph is continuously updated with new transactions, and the model is pre-trained on historical data, so when recommendations are required, the system can quickly query the pre-processed graph without performing heavy computation at recommendation time
Solution Approach 2:
The patent transitions from traditional one-dimensional filtering criteria (maturity, rating, sector, currency) to a multi-dimensional graph-based representation that captures complex relationships between clients and assets. The heterogeneous graph adds new dimensions by representing entities as nodes and relationships as edges, enabling the system to analyze trading history and interaction patterns that cannot be captured by simple categorical filters
Data Source
AI summary
Systems and methods for predicting recommendations using graph relationships are disclosed. According to an embodiment, a method may include: (1) monitoring, by a computer program, a messaging interface for transactions; (2) updating, by the computer program, a heterogeneous graph with data from the transactions, wherein the heterogeneous graph identifies a plurality of assets and a plurality of clients; (3) training, by the computer program, a graph model with the heterogeneous graph; (4) querying, by the computer program, the graph model with one of the plurality of assets, wherein the graph model returns a recommendation that identifies a subset of the plurality of clients for the asset; and (5) outputting, by the computer program, the recommendation.