Date of Award
2-22-2016
Document Type
Thesis
Degree Name
Master of Science (MS)
Department
Computer Science
First Advisor
Kenji Yoshigoe
Abstract
Healthcare organizations generate massive quantities of data which are often convolved with valuable hidden insights. In order to discover new adverse drug events (ADEs) the data mining techniques are employed extensively, but the computational complexity of the traditional data analytics platforms makes the need to migrate to a scalable analytics platform. The goal in experiments is predicting the possibility of certain interactions given a prescribed set of drugs. In this work, we used Spark platform and MLlib library as an alternative scalable solution that we found them scalable, easy to work with and with reasonable modeling accuracy. We utilized Hadoop and Spark to process the data and MLlib to perform the predictive modeling task and as result we could have the modeling accuracy as good as 70%. Also to tackle curse of dimensionality and feature selection we used Logistic regression with L1 Norm that led us with improvement in accuracy and modeling time up to 5% and 5 time faster respectively.
Recommended Citation
Rezaie, Mohammadreza, "Discovering Adverse Drug Interaction Using Convex Optimization Techniques" (2016). Theses and Dissertations. 663.
https://research.ualr.edu/etd/663
