Assorted Model of Sentiment using Mapreduce Framework

Abstract

Social networking sites platforms, such as Facebook and Twitter, are being broadly used by community to share their feelings on different matters. Consequently, social networking site becomes an admirable and major open source for collecting public opinion. To perform sentiment analysis on such huge data, computational assorted models of single node are ineffective. Two ways to grip data that are big ,either by using super computers or by using parallel processing or by distributed processing. Where it is costly to use super computer, most models of parallel processing such as MPI, are difficult to implement and scaling, MapReduce is one of the parallel processing models that is highly scalable, tolerant to fault, and easy for using, in this research paper, we have proposed assorted model of sentiment analysis for twitter using MapReduce Framework. mapreduce based naïve bayes training algorithm was proposed for this purpose. Only single mapreduce job is executed for this algorithm which makes it different from earlier previous work. Training model is deployed to to classify million of tweets of twitter computers are costly, most parallel programming models, such as MPI, are difficult to use and scale. MapReduce is one of the parallel programming models that is highly scalable, fault tolerant and easy to use. This paper proposes a scalable framework for sentiment analysis of Twitter using MapReduce model. For this purpose a MapReduce based Na¨ıve Bayes training algorithm is proposed, this algorithm uses only one MapReduce job which makes it different from previous works. The trained model is deployed to classify millions of tweets. Accuracy and Scalabilty of our propsed model is well compared to previous models.