Date of Award

11-2018

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Computer Science

First Advisor

Dr. Brian Ricks

Second Advisor

Dr. Margeret Hall

Third Advisor

Dr. Yuliya Lierler

Abstract

Analysis of Facebook posts provides helpful information for users on social media. Current papers about user engagement on social media explore methods for predicting user engagement. These analyses of Facebook posts have included text and image analysis. Yet, the studies have not incorporate both text and image data. This research explores the usefulness of incorporating image and text data to predict user engagement. The study incorporates five types of machine learning models: text-based Neural Networks (NN), image-based Convolutional Neural Networks (CNN), Word2Vec, decision trees, and a combination of text-based NN and image-based CNN. The models are unique in their use of the data. The research collects 350k Facebook posts. The models learn and test on advertisement posts in order to predict user engagement. User engagements includes share count, comment count, and comment sentiment. The study found that combining image and text data produced the best models. The research further demonstrates that combined models outperform random models.

Comments

A Thesis Presented to the College of Information Science and Technology and the Faculty of the Graduate College University of Nebraska at Omaha In Partial Fulfillment of the Requirements for the Degree Master of Science in Computer Science. Copyright 2018 Chad Crowe.

Download

Files over 3MB may be slow to open. For best results, right-click and select "save as..."

COinS