07/05/2022
By Mary Lou Kelly

The Manning School of Business, Department of Operations & Information Systems, invites you to attend a doctoral dissertation defense by Murtaza Nasir on “Essays on Data Science & Analytics: Methodology Development and Applications.”

Candidate name: Murtaza Nasir
Defense Date: Friday, July 15, 2022
Time: 3-5 p.m.
Location: Pulichino Tong Business Center, Room 462
Dissertation Title: Essays on Data Science & Analytics: Methodology Development and Applications

Advisors: Asil Oztekin, Ph.D. and Nichalin Summerfield, Ph.D.

Committee Members:
Asil Oztekin, Ph.D. (Chair), Operations & Information Systems Department, UMass Lowell
Nichalin Summerfield, Ph.D. (Co-Chair), Operations & Information Systems Department, UMass Lowell
Yao Chen,, Ph.D., Operations & Information Systems Department, UMass Lowell
Serhat Simsek, Ph.D., Montclair State University

Abstract:

Machine Learning (ML) has become increasingly important in the domain of business decision making processes and understanding business problems. In this work we propose some novel machine learning methodologies that enable us to overcome some challenges with working with imbalanced and complex datasets. Furthermore, we present some works that show applications of these as well as some existing ML methodologies. The first methodology, titled “SANSA, a synthetic average neighborhood sampling algorithm,” improves on existing synthetic data generation algorithms for imbalanced learning by providing a novel tuning parameter that allows the algorithm to adapt to a wide variety of datasets, as opposed to existing methods. The second methodology presented, titled “SIFT, a simple interaction finding technique,” presents a novel methodology to make black-box machine learning models more transparent by analyzing and presenting how the any given trained model makes decisions. The methodology can be used to identify novel interactions between predictors as well as any potential biases in the model. In the remaining chapters of this work, we show some applications of machine learning algorithms, including SANSA and SIFT. We present chapters on using machine-learning based analytics to understand important factors and their interactions for successful completion of substance use disorder treatment as well as for various COVID-19 diagnostic as well as prognostic outcomes.