MUMBAI, India, Aug. 31 -- Intellectual Property India has published a patent application (202621082746 A) filed by Symbiosis International Deemed University on July 03, 2026, for A Multi-Model Graph Neural Network Engine For Interactive Link Prediction In Complex Networks.
Inventors include Dr. Deepak Suresh Asudani; Soumya Bajirao; and Janhavi Admane.
The application for the patent was published on August 28, 2026, under issue no. 35/2026.
Abstract: ABSTRACT A MULTI-MODEL GRAPH NEURAL NETWORK ENGINE FOR INTERACTIVE LINK PREDICTION IN COMPLEX NETWORKS The present invention discloses a multi-model graph neural network engine (100) for interactive link prediction in complex networks. The system comprises a dynamic data ingestion module (110) for receiving standardized citation datasets and custom user-uploaded CSV edge lists; a graph formulation module (120) for constructing adjacency and feature matrices with negative sampling; a core GNN engine comprising a GCN encoder (130), a GraphSAGE encoder (140), and a GAT encoder (150) for learning node embeddings; a dot-product decoder (155) for computing link confidence scores; an asynchronous API gateway (160) for decoupled backend processing; and an interactive visualization dashboard (170) with physics-based force-directed graph rendering for real-time prediction display. The system enables comparative ablation across architectures, achieving 98.37 percent AUC with the GAT model. [
Disclaimer: Curated by HT Syndication.
