SENTIMENT ANALYSIS USING NEURAL NETWORK

Abstract

In last few years, there has been a rapid growth in the use of social networking sites so this become a significant medium for people to express their views or opinions. Sentiment Analysis is used to identifying and classifying the polarity of an opinion, such as positive, neutral or negative. Twitter is one of the social media site which is gaining the popularity. Twitter Sentiment Analysis has application such as political sentiment analysis, companies getting their customer's perspective on their products or opinions on current issue, movie reviews. Recent research has involved looking at text from online tweets, online movie reviews, etc. to try and classify the text as being positive, negative, or neutral. For this research work, a Supervised Machine Learning Based approach is realised for Sentiment Analysis of tweets. The approach used here is Probabilistic Neural Network because PNN is a multi-layered feedforward neural network due to which it is fast as compare to other approaches. PNN is developed based on an estimation method that is probability density function for Parzen window and Bayes classification rules. Probabilistic Neural Network has feature of adaptive learning, fault tolerance, parallelism and generalization which provides a superior performance. Smoothing parameter of PNN plays a great role for predicting an accurate class of classifier. So a self-adaptive algorithm is used to calculate and optimize the smoothing parameter. Training and Testing dataset is collected from twitter using Twitter API based on a keyword “#NSG”. This classifier is used to determine the opinion of a tweet whether this tweet is about to good effort of Indian Prime Minister or a diplomatic failure of not getting a NSG membership. Probabilistic Neural Network is trained with the smoothing parameter which is adjusted self-adaptively so that best parameter is selected and according to that results are obtained for testing data set after that accuracy of classifier is determined.