Document Type
Thesis - Open Access
Award Date
2022
Degree Name
Master of Science (MS)
Department / School
Electrical Engineering and Computer Science
First Advisor
Kaiqun Fu
Abstract
This paper presents and explores a novel way to determine the sentiment of a Steam game review based on the predicted recommendation of the review, testing different regression models on a combination of Term Frequency-Inverse Document Frequency (TF-IDF) and Latent Dirichlet Allocation (LDA) features. A dataset of Steam game reviews extracted from the Programming games genre consisting of 21 games along with other significant features such as the number of helpful likes on the recommendation, number of hours played, and others. Based on the features, they are grouped into three datasets: 1) either having keyword features only, 2) keyword features with the numerical features, and 3) numerical features only. The three datasets were trained using five different regression models: Multilinear Regression, Lasso Regression, Ridge Regression, Support Vector Regression, and Multi-layer Perceptron Regression, which were then evaluated using RMSE, MAE, and MAPE. The review recommendation was predicted from each model, and the accuracy of the predictions were measured using the different error rates. The results of this research may prove helpful in the convergence of Machine Learning and Educational Games.
Library of Congress Subject Headings
Educational games.
Video games -- Reviews.
Sentiment analysis.
Natural language processing (Computer science)
Number of Pages
44
Publisher
South Dakota State University
Recommended Citation
Zhang, Anqi, "Sentiment without Sentiment Analysis: Using the Recommendation Outcome of Steam Game Reviews as Sentiment Predictor" (2022). Electronic Theses and Dissertations. 490.
https://openprairie.sdstate.edu/etd2/490
Rights Statement
