Revolutionizing Healthcare Decision-Making With Graph Neural Networks, Lightgbm, And Deep Q-Networks For Efficient And Accurate Predictions

Abstract

For enhanced healthcare decision-making, this research introduces a hybrid prediction model integrates LightGBM, Deep Q-Networks (DQN), and Graph Neural Networks (GNNs). DQNs employ reinforcement learning for optimized predictions, LightGBM allows for efficient feature selection and interpretability, and GNNs detect complex patterns of relationships among patient data. Together, the three enable diagnostics to be flexible, accurate, and interpretable across diverse shifting healthcare scenarios. The proposed model has better performance compared to traditional models in actual tests, with 92.8% accuracy, 91.3% precision, 93.5% recall, and 92.4% F1-score. The proposed model improves the performance of risk prediction and diagnosis by 20% through a 0.964 ROC-AUC. Through its outstanding predictive ability, scalability, and reliability, this GNN-LightGBM-DQN system offers a paradigm shift for real-time medical decision support.

Ähnliche Arbeiten