Abstract

Adverse drug reactions are serious unintended effects from drug usage. It affects millions of people worldwide each year, resulting in numerous deaths and hospitalizations. Because millions of people are affected, it would be beneficial to know the frequent cases so health care practitioners can use more caution when prescribing drugs to patients. In this paper, I present a clustering model that identifies the significant groups of patients with adverse drug reactions. Clustering model is an unsupervised machine model that finds the best number of groups or classes for the instances in the data set. Using the clustering approach, I find that the optimal number of groups of patients in the data is five, meaning that there are five groups that can be identified from the data. Knowing this, when health care practitioners prescribe drugs for treatment, they would be more knowledgeable about the kinds of patients that belong to a certain group of frequent cases and use more caution to avoid high risk prescriptions.