Compound Suitability Prediction in Blockchain-Based Cross-Domain Supply Chain using AdamW5 Optimized FT-Transformer

The safe and effective use of chemical compounds across agriculture, pharmaceuticals, and food industries is vital for public health and environmental sustainability.However, current systems struggle with predicting compound suitability, limiting traceability and data integrity.This study presents an integrated AI-driven blockchain framework to predict the suitability of chemical compounds across these sectors, ensuring transparent traceability in supply chains.The research develops a robust multilabel classification model using an optimised FT-Transformer, which effectively handles complex molecular descriptor data.To improve classification performance, the study introduces the AdamW5 optimiser, which integrates multiple optimisation strategies to enhance convergence stability and generalisation performance.The suitability predictions are securely stored on a Hyperledger Fabric blockchain, providing immutable records, automated access control, and verifiable audit trails.This framework achieved a classification accuracy of 98.45%, with a precision of 98.17% and an F1-score of 97.95%, significantly outperforming baseline models, while ensuring secure and tamper-proof data storage.