Deep and Transfer Learning-based Research Article Recommendation System for Healthcare Services

To address the growing volume of confidential biomedical data in the healthcare sector, we propose an inventive Deep Learning (DL) approach for suggesting relevant articles, tailored specifically for the healthcare field. The proposed approach addresses the challenge of information curation by presenting a personalized assistant that aids both patients and medical professionals in navigating the extensive scholarly landscape. In situations where individuals face limitations in accessing healthcare services, such as remote or restricted circumstances, our approach demonstrates significant value in providing direct medical consultation. By acquiring insights into symptoms and potential diseases through medical articles, it provides a foundation for informed decision making. It harnesses advanced pre-trained models including Bidirectional Encoder Representations from Transformers (BERT), Robustly Optimized BERT Pretraining Approach (RoBERTa), and eXtreme Learning Machine Network (XLNet). These models undergo rigorous training using diverse medical article datasets, extracting nuanced semantics through tokenization, embeddings, self-attention mechanisms, and transformations. The system's reliability is enhanced by conducting a thorough assessment of optimizers, including Root Mean Square Propagation (RMSprop), AdamW, and Stochastic Gradient Descent (SGD). Finally, the proposed approach is evaluated against various matrices such as accuracy and loss curves, F1, recall, and precision comparison.