Drug Satisfaction Analytics: Investigating the Effectiveness of Various Embedding Techniques

In the healthcare economy, medication satisfaction can be defined as a patient reported upshot. Gaining insights into side effects of medication from individuals who have previously encountered them is crucial. This study concentrates on predicting satisfaction rating of drugs by extracting reviews from social healthcare forums for five distinct medications, each addressing a different health condition. Experimental analysis is conducted to assess the performance of Random Forest classifier integrated with three unique embeddings namely BERT, BioBERT and S- BERT. The results demonstrates that the fusion of Random Forest with BERT surpasses other approaches in terms of evaluation metrics. This superior performance is particularly evident in the range of F1 scores achieved across five different datasets, where this model combination scored between 0.91 and 0.98.