Classification and Analysis of Political Video Advertisements
Date
2021-12
Authors
Stumbaugh, Scott
Major Professor
Nordman, Daniel
Vardeman, Stephen
Advisor
Committee Member
Sabzikar, Farzad
Journal Title
Journal ISSN
Volume Title
Publisher
Authors
Research Projects
Organizational Units
Journal Issue
Is Version Of
Versions
Series
Department
Abstract
As the ever-increasing fanfare around machine learning grows, so do the seemingly endless number of methods. Without a good grasp of the statistical fundamentals, there is a tendency to think this class of models is a black box that functions as the “magic bullet,” able to blindly solve data problems in every situation. Taking time to examine the underlying principles, and break down how they steer certain model behaviors in the presence of a particular data set, helps to understand how a model will perform and why. Motivated by this mind set, the analytical approach of this paper aims to explore a novel modification to the classic decision tree model that incorporates a two-step-look-ahead approach. Included is a comparative analysis between the original algorithm (one-step-look-ahead logic) and the new one, which shows in the case of a particular simulated dataset that the two-step-look-ahead logic vastly outperforms the classic decision tree algorithm.
Comments
Description
Keywords
Citation
DOI
Source
Subject Categories
Copyright
2021