11. Utilizing Multilevel Classification to Predict Adverse Drug Effects and Reactions

Tori Puhl Butler University
Faculty Sponsor(s): Rasitha Jayasekare Butler University
Multi-class classification models are used to predict categorical response variables with more than two possible out- comes. A collection of multi-class classification techniques such as Multinomial Logistic Regression, Support Vector Machines, Naive Bayes, and k-Nearest Neighbors is used in predicting patients drug reactions and adverse drug effects based on patients demographic and drug administration. The newly released, 2018 data on drug reactions and adverse drug effects from the U.S. Food and Drug Administration are tested with the models. The applicability of model evaluation measures such as sensitivity, specificity and the receiver operating characteristics curve in multi-class settings, are also discussed.
Mathematics & Computer Science
Poster Presentation

When & Where

Irwin Library 3rd Floor