Ensembling classical machine learning and deep learning approaches for morbidity identification from clinical notes

In this study, we offer a unique technique for detecting co-morbidity in patient clinical data, employing Latent Dirichlet Allocation (LDA) for topic modelling. The LDA method is used to extract latent themes from the clinical notes. By using these selected themes as the foundation for feature representation, the data is represented in a way that is easier to understand and more concise. The possibility of LDA in identifying significant connections between subjects and health problems by using the topic-based characteristics for morbidity categorization. Our method aims to improve co-morbidity diagnosis via simplicity, interpretability, and efficacy. We want to show the effectiveness of the LDA-driven system in assisting with the precise and perceptive categorization of medical diseases from clinical data by means of a thorough study and comparison with current approaches.