Knowledge Graph Embedding for Prediction Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for building and utilizing knowledge graphs, such as those between venture capital companies and startup companies, face challenges in accurately modeling relationships and predicting future actions due to limitations in human networks and communication, as well as inaccuracies in peer groupings based on node attributes.
Innovation Solution
The method involves embedding a graph using an embedding model to generate node and edge vectors, forming peer node groups, and training a machine learning model to predict future actions, which are then used to generate insights and stored in an insight database, with dynamic embedding and exposure of node and edge attributes to improve accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional attribute-based methods are used to build knowledge graphs, then the implementation is simple, but the prediction accuracy and relationship modeling are insufficient
Solution Approach 1:
The patent transforms graph data from traditional attribute-based representation to embedding vector representation. Nodes and edges are converted into dense vector embeddings that capture structural and semantic information, fundamentally changing the parameter space from discrete attributes to continuous vector spaces, thereby improving prediction accuracy while managing complexity through learned representations
Solution Approach 2:
The patent replaces traditional rule-based and attribute-matching mechanisms with machine learning models. Instead of manually defining relationships based on node attributes, the system uses trained ML models that automatically learn patterns from embedded representations, substituting mechanical rule-based systems with adaptive learning-based systems
2Measurement precision
If graph embedding and machine learning models are used, then prediction accuracy improves, but computational complexity and processing time increase
Solution Approach 1:
The patent performs graph embedding and model training in advance as a preliminary action. The embedding model pre-processes the entire graph structure to generate node and edge embeddings, and the machine learning model is trained beforehand on historical data. This pre-computation stores learned representations that can be quickly applied to new predictions, reducing real-time processing time while maintaining high accuracy
Solution Approach 2:
The patent creates embedded vector copies of nodes and edges that capture essential graph structure information. Instead of repeatedly processing the original complex graph data during predictions, the system uses these pre-generated embedding vectors as compressed representations, enabling faster computation while preserving the structural relationships needed for accurate predictions
3Adaptability or versatility
If dynamic embedding with historical graphs is implemented, then the model adapts to temporal changes, but computational resources and processing overhead increase
Solution Approach 1:
The patent implements dynamic embedding by periodically re-processing historical graphs at defined intervals. Instead of continuously updating embeddings in real-time, the system performs embedding updates periodically on historical data, allowing the model to adapt to temporal changes in the graph structure while avoiding constant computational overhead. This periodic approach balances adaptability with resource conservation
Data Source
AI summary
In some aspects, the techniques described herein relate to a method including: embedding a graph with an embedding model, wherein the embedding generates node and edge vectors for each node and each edge, respectively, in the graph; generating peer node groups based on the node vectors and the edge vectors; generating edge predictions based on the node vectors and the edge vectors; training a machine learning model based on the node vectors and the edge vectors; inputting the node vectors and the edge vectors into the machine learning model; receiving, as output from the machine learning model, a prediction of a future action based on the node vectors and the edge vectors; generating a plurality of insights based on the peer node groups, the edge predictions, and the prediction of a future action; storing the plurality of insights in an insight database, and providing an interface to the insight database.