Fine-grained sentiment classification of social media

This research is conducted to enhance the sentiment analysis for classifying data that are collected from Twitter into binary classes positive and negative. The project starts with the implementation of valence-based and rule-based method to improve the current simple polarity-based method. The impl...

Full description

Saved in:
Bibliographic Details
Main Author: Le Thi, Nhu Y
Other Authors: Lin Zhiping
Format: Final Year Project
Language:English
Published: 2017
Subjects:
Online Access:http://hdl.handle.net/10356/71001
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Nanyang Technological University
Language: English
id sg-ntu-dr.10356-71001
record_format dspace
spelling sg-ntu-dr.10356-710012023-07-07T16:35:25Z Fine-grained sentiment classification of social media Le Thi, Nhu Y Lin Zhiping School of Electrical and Electronic Engineering A*STAR Institute of High Performance Computing Wang Zhaoxia DRNTU::Engineering::Electrical and electronic engineering This research is conducted to enhance the sentiment analysis for classifying data that are collected from Twitter into binary classes positive and negative. The project starts with the implementation of valence-based and rule-based method to improve the current simple polarity-based method. The implementation includes the tuning method for determining threshold value that gives the best classification results. Then, the results are compared and discussed, which concludes that the valence-based method performs better than the polarity-based method in various datasets. In addition, the Random Forest classifier with word frequency as feature is implemented and evaluated in comparison with other machine learning classifiers consisting of Support Vector Machine, Naïve Bayes, Maximum Entropy and Extreme Learning Machine. The tuning method of hyperparameters for Random Forest in different datasets is also explained, and an idea is introduced about the impact of parameters on its performance as well as its prospective application. The result has shown that with the proper tuning of Random Forest hyperparameters, including the number of decision trees and the maximum number of random features, it can give the highest accuracy for larger datasets in all the five classifiers discussed in this report. Bachelor of Engineering 2017-05-12T07:09:08Z 2017-05-12T07:09:08Z 2017 Final Year Project (FYP) http://hdl.handle.net/10356/71001 en Nanyang Technological University 55 p. application/pdf
institution Nanyang Technological University
building NTU Library
continent Asia
country Singapore
Singapore
content_provider NTU Library
collection DR-NTU
language English
topic DRNTU::Engineering::Electrical and electronic engineering
spellingShingle DRNTU::Engineering::Electrical and electronic engineering
Le Thi, Nhu Y
Fine-grained sentiment classification of social media
description This research is conducted to enhance the sentiment analysis for classifying data that are collected from Twitter into binary classes positive and negative. The project starts with the implementation of valence-based and rule-based method to improve the current simple polarity-based method. The implementation includes the tuning method for determining threshold value that gives the best classification results. Then, the results are compared and discussed, which concludes that the valence-based method performs better than the polarity-based method in various datasets. In addition, the Random Forest classifier with word frequency as feature is implemented and evaluated in comparison with other machine learning classifiers consisting of Support Vector Machine, Naïve Bayes, Maximum Entropy and Extreme Learning Machine. The tuning method of hyperparameters for Random Forest in different datasets is also explained, and an idea is introduced about the impact of parameters on its performance as well as its prospective application. The result has shown that with the proper tuning of Random Forest hyperparameters, including the number of decision trees and the maximum number of random features, it can give the highest accuracy for larger datasets in all the five classifiers discussed in this report.
author2 Lin Zhiping
author_facet Lin Zhiping
Le Thi, Nhu Y
format Final Year Project
author Le Thi, Nhu Y
author_sort Le Thi, Nhu Y
title Fine-grained sentiment classification of social media
title_short Fine-grained sentiment classification of social media
title_full Fine-grained sentiment classification of social media
title_fullStr Fine-grained sentiment classification of social media
title_full_unstemmed Fine-grained sentiment classification of social media
title_sort fine-grained sentiment classification of social media
publishDate 2017
url http://hdl.handle.net/10356/71001
_version_ 1772826929735401472