Please use this identifier to cite or link to this item:
https://knowledgecommons.lakeheadu.ca/handle/2453/815
Title: | Improved Sentiment Classification by Multi-model Fusion |
Authors: | Gan, Lige |
Keywords: | Sentiment analysis;Natural language processing;Machine learning;Classification;Multi-model;Data fusion |
Issue Date: | 2016 |
Abstract: | Sentiment Analysis (SA), also Opinion Mining, is a sub-field of Data Mining. It aims at studying and analyzing human’s sentiments, opinions, emotions or attitudes through their written text. Since all sentiment information are hiding in the text content, a way for acquiring them is using Natural Language Processing (NLP) techniques. On the other hand, with the development of Artificial Intelligence (AI), more Machine Learning (ML) algorithms have been developed. In recent years, these ML algorithms are widely applied in SA field for classifying the text instead of traditional methods. However, selecting an appropriate ML algorithm is a controversial topic in SA research. In this study, we investigated nine commonly used algorithms such as Naïve Bayes (NB), Support Vector Machine (SVM) and Logistic Regression (LR). A comprehensive comparison of the nine ML algorithms using different metrics enabled to develop a merging model for deriving an optimum algorithm for a specific SA task. The proposed merging model, also called the multi-model, combines multiple ML algorithms’ results by using some fusion method to get the best performance out of these algorithms. The performance of the multi-model has also been evaluated and compared to the single ML algorithms. |
URI: | http://knowledgecommons.lakeheadu.ca/handle/2453/815 |
metadata.etd.degree.discipline: | Engineering : Electrical & Computer |
metadata.etd.degree.name: | Master of Science |
metadata.etd.degree.level: | Master |
metadata.dc.contributor.advisor: | Khoury, Richard Benlamri, Rachid |
metadata.dc.contributor.committeemember: | Atoofian, Ehsan Fiaidhi, Jinan Christoffersen, Carlos |
Appears in Collections: | Electronic Theses and Dissertations from 2009 |
Files in This Item:
File | Description | Size | Format | |
---|---|---|---|---|
GanL2016m-1a.pdf | 1.53 MB | Adobe PDF | View/Open |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.