Detection of ADR on Covid Vaccine Safety Data

Due to pandemic vaccines are developed at a rapid pace. There is a requirement to ensure proper post-market pharmacovigilance. The proposed model will help speed up this process by classifying the Adverse Drug Reactions (ADRs) of the vaccines based on the severity. This will help vaccine manufacturers take necessary and timely action. The model will input the patient data (such as symptoms, vaccination details, and patient health details), which will be preprocessed and cleaned. The ADR will then be classified as a minor, major, or deadly reaction. The system made use of Count Vectors (CV), Word TF-IDF, N-gram TFIDF, and Character TF-IDF feature with Naive Bayes, Random Forest, Logistic Regression, Gradient Boost, and Adaboost machine learning classifiers. Using Random Forest with word-level TF-IDF comparatively a higher accuracy of 93.83% and an F1 score of 0.92 was achieved.