Date of Award

2016

Document Type

Dissertation

Degree Name

PhD in Business

Department

Department of Mathematical Sciences: Business Analytics

First Advisor

Dominique M. Haughton

Second Advisor

Richard J. Cleary

Third Advisor

Jennifer Xu

Abstract

A consensus has been reached nowadays on the importance of data-based decision making for both data-driving pioneers such as web companies, and traditional industries or departments, such as pharmaceutical organizations, or even movie or television production companies, etc. How we collect, process, and make use of larger, fast changing, and diversified data effectively and efficiently is getting attention from statisticians, information engineers, and business analytics practitioners. Aside from thoroughly taking advantage of data on hand, data adventurers seek new data resources, which are potentially messier and more unstructured than more traditional data.

This research provides a road map to practitioners who might wish to implement our methods. We demonstrate in study 1, using a Hidden Markov Chain model, that the unknown advertising activity of a competitor can be estimated using the competitor’s sales volume together with the focal company’s sales volume and advertising expenditures. Conversely, using the same method, a competitor could estimate the focal company’s advertising activity. We introduce in this context the idea of symmetric imputation of competing marketing activity. In study 2, we propose novel methods for extracting a co-publication network of physicians from the PubMed database and for combining data from this network with a more traditional pharmaceutical marketing database. While traditional marketing predictive models assume that actors are independent, the study advances our knowledge of predictive modeling when actors are inter-dependent via a social network. Our work also suggests recommendations to the pharmaceutical industry for a more effective use of marketing dollars, which could potentially lead to reduced costs of drugs to consumers. In study 3, we conduct a text mining and sentiment analysis applied to a dataset of movie reviews. We further propose a novel measure of controversy, compute sentiment scores for each movie review and compare for each movie the numerical rating “controversy” (the standard deviation of numerical reviewer ratings) with our proposed sentiment score-based measure of “controversy”.

Download

Share

COinS