Indian Journal of Science and Technology
DOI: 10.17485/ijst/2016/v9iS1/109897
Year: 2016, Volume: 9, Issue: Special Issue 1, Pages: 1-5
Original Article
Yong-Wook Nam and Yong-Hyuk Kim*
Department of Computer Science, Kwangwoon University, 20 Kwangwoon-ro, Nowon-gu, Seoul 139-701, Korea; [email protected]
[email protected]
*Author for correspondence
Yong-Hyuk Kim
Department of Computer Science
Email:[email protected]
Objectives: This paper aims to gather and categorize valuable tweets that are shared by many people regarding the opinions expressed in Social Network Services (SNS)in real time. Methods/Statistical Analysis: Among many SNS, we have targeted Twitter which has excellent data accessibility. To find the comments on the current hot issue keywords, Google and Twitter Trends Keywords were utilized in the search. At first, the most retweeted tweets were gathered, but contrary to our expectations, most of them were general news that did not require analysis or marketing-related advertisements, etc., so they were classified. We solved this issue by making use of machine learning. Findings: Since media and celebrities have many followers, more of their tweets are retweeted compared to the average number of retweets per account. Therefore, because opinions should not be distinguished as influential just by their retweet numbers, the model made from the training data gathered in this study classified the tweets to analyzing the opinion mining. The evaluation of classification results showed 84.8% accuracy, and the evaluation of new tweets showed an accuracy of 84%. It seems that much more accurate results could be predicted with more training data. Improvements/Applications: A program allowing users to find influential opinions about real-time trending search terms has been developed. |
Keywords: Influential opinion, Machine Learning, SNS, Spam Detection, Twitter
Subscribe now for latest articles and news.