New Ensemble Approach to Analyze User Sentiments from Social Media Twitter Data

Abstract

The Sentiment analysis helps to identify and classify the opinions or the sentiments which are expressed by means of source text. Social media have gained more attention nowadays. Public and private opinions about varied subjects are expressed and spread continually via numerous social media. Twitter is one of the social media that is gaining popularity. Thus, the twitter is widely used for analyzing the sentiment of the huge groups. It is difficult to perform sentiment analysis in twitter in comparison with general sentiment analysis since the twitter contains many special characters, symbols, slang words and also has the limitation of 140 words. Knowledge base approach and Machine Learning approach are the two paths for analyzing the text. This paper proposes a twitter sentiment analysis that can spot the interest and opinion of people in regard to the products such as e-Commerce sites, network operators and social media. Feature for classification of the tweets as negative, neutral and positive that depends on views of people and the identification of accuracy by various algorithmic approaches. It also involves the emotional classification of tweets into joy, fear, sadness, disgust, anger, surprise and unknown. The word cloud has been identified for the usage of most repeated word. This paper proposes new ensemble approach for sentiment analysis. Then the algorithms such as Naïve Bayes, Random Forest and kNN have been employed in Rapidminer to find the accuracy of the sentiment. Results indicate that the proposed approach gives good accuracy than traditional algorithms.