Date of Award

3-2022

Document Type

Thesis

Degree Name

Master of Science

Department

Department of Operational Sciences

First Advisor

Phillip M. LaCasse, PhD

Abstract

Emotion classification can be a powerful tool to derive narratives from social media data. Recurrent Neural Networks (RNN) can meet or exceed the performance of state-of-the-art traditional machine learning techniques using exclusively open-source data and models. Specifically, these results show that RNN variants can produce more than an 8% gain in accuracy in comparison to Logistic Regression and SVM techniques and a 15% gain over Random Forest when using FastText embeddings. This research found a statistical significance in the performance of a single layer Bi-directional Long Short-Term Memory (Bi-LSTM) model over a 2-layer stacked Bi-LSTM model. This research also found that a single layer Bi-LSTM RNN met the performance of a state-of-the-art Logistic Regression model with supplemental closed-source features from a study by Saputri et al. (2018) when classifying the emotion of Indonesian Tweets. This model can be provided to operational units within the INDOPACOM theater giving them the ability to identify social media posts based on predicted emotion class - allowing them to gauge public reaction to military exercises in theater.

AFIT Designator

AFIT-ENS-MS-22-M-131

DTIC Accession Number

AD1172390

Share

COinS