Using Diverse Feature for Opinion Mining of “Kerala Floods 2018”
S. Fouzia Sayeedunnisa1, Nagaratna P Hegde2, Khaleel Ur Rahman Khan3

1S. Fouzia Sayeedunnisa, Department of IT, M.J. College of Engineering and Technology, Hyderabad (Telangana), India.
2Dr. Nagaratna P Hegde, Department of CSE, Vasavi College of Engineering, Hyderabad (Telangana), India.
3Dr. Khaleel Ur Rahman Khan, Department of CSE, ACE Engineering College, Hyderabad (Telangana), India.
Manuscript received on 26 February 2019 | Revised Manuscript received on 13 March 2019 | Manuscript Published on 17 March 2019 | PP: 23-27 | Volume-7 Issue-ICETESM18, March 2019 | Retrieval Number: ICETESM07|19©BEIESP
Open Access | Editorial and Publishing Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (

Abstract: With the intensive use of social media for communication among people, the outspread of information on these platforms has increased the demand to perform mining of user insights on various topics to gain knowledge .This mining of user reviews to know the positive and negative opinion of people towards a topic, product, brand is Sentiment Analysis. A subject of special interest Sentiment Analysis finds application in various fields viz. Brand Monitoring, Voice of Employee, Social media monitoring .In this paper mining of tweets from Social Media site Twitter is done to analyze the public opinion on “Kerala Floods 2018 “.Twitter generates 500 million messages called ‘tweets’ per day. Processing these gargantuan tweets is time consuming; this paper aims in processing these tweets into set of features by using Sentiment Lexicons and then applying a filtering method to extract those features which are of high value and discarding all the low value features. Emoticons, slang, hash tags plays a vital role in conversation among people to express their opinions. The extracted features which are huge in number are reduced using the feature selection method Information Gain (IG). The selected high value features through IG are applied to Bayes classifier for classification of opinions. It is apparent from the results that analyzing sentiments using emoticons, slang and hash tag features as one among the features is better than using conventional n- gram features. This manuscript uses Accuracy, Precision, Recall, F- measure and Time for processing to analyze the performance of high value words using IG.
Keywords: Emoticons, n- Grams, Slang, Social Network Twitter.
Scope of the Article: Data Mining and Warehousing