Hybrid transformer and graph neural network framework for sustainable development goals classification: dataset and code

This repository contains the dataset and implementation code for the paper "Context-aware sustainable development goals classification of scientific literature using a hybrid domain-specific transformer and graph neural network framework" by Jalal Sadoon Hameed Al-Bayati and Mohammed Al-Shammaa, University of Baghdad. The repository includes: (1) 17 CSV files (sdg1.csv–sdg17.csv) containing 19,354 Scopus-indexed peer-reviewed publications from the University of Baghdad (2015–2026), labeled according to the 17 United Nations Sustainable Development Goals; (2) the hybrid model training code (05_Hybrid_8GB.ipynb) implementing the SciBERT–GraphSAGE architecture; and (3) baseline experiment code (SDG_Class_Expv2.ipynb) for TF-IDF+SVM, BERT, and SciBERT baselines.