An Optimized CNN Based Robust Sentiment Analysis System on Big Social Data using Text Polarity Feature
Komalpreet Kaur1, Chitender Kaur2, Tarandeep Kaur Bhatia3
1Komalpreet Kaur, M. Tech Research Scholar, Department of Computer Science and Engineering, Chandigarh Engineering College, Landran (Punjab), India.
2Chitender Kaur, Assistant Professor, Department of Computer Science and Engineering, Chandigarh Engineering College, Landran (Punjab), India.
3Tarandeep Kaur Bhatia, Assistant Professor, Department of Computer Science and Engineering, Chitkara University, Rajpura (Punjab), India.
Manuscript received on 07 April 2019 | Revised Manuscript received on 20 April 2019 | Manuscript published on 30 April 2019 | PP: 1871-1877 | Volume-8 Issue-6, April 2019 | Retrieval Number: F3928048619/19©BEIESP
Open Access | Ethics and 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 (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: People express their personal views and emotions by reacting to the event, product and individuals. These reactions are very vulnerable, therefore comprehending such data and then processing it effectively, have been the content of research in various areas such as business and politics. Sentiment analysis in social media, in which the task of extracting subjective information from the e-commerce site, social sites have drawn great recognition from the Web mining community. Social media gives an invaluable insight not only into human ideas but also in the challenges incurred due to huge amounts of big data. These issues comprises of the processing of massive amounts of streaming data, as well as automatically identifying human expressions within short text messages. To solve this problem, an intelligent sentiment analysis approach is presented, which is used to extract the opinions of people from social media. Initially, a lexicon dictionary is created that comprises of positive, negative and neutral words. Then, pre-processing (normalization, punctuation removal, stop word removal and tokenization) is applied to the test data. Finally, Feature extraction, feature optimization and classification algorithms are applied to the pre-processed data. PSO is used as a feature optimization algorithm with Convolutional neural network (CNN) as a classification algorithm. CNN is trained as per the optimized features. During testing phase, sentiments such as positive, negative is identified by matching the uploaded text with the saved data. The presented model helps to identify people feeling in terms of business, comments, and reviews written on the social sites. The detection accuracy up to 98.32 % is obtained.
Keyword: Big Data, Convolutional Neural Network, Particle Swarm Optimization, Sentiment Analysis, Social Media.
Scope of the Article: Social Networks