Clustering Yelp’s sentiment data through various approaches and estimating the error rate

Abstract Yelp is a platform that provides free access to use the data for educational and research purposes. The sentiment data which is used for this empirical research consists of two attributes namely sentiment text and sentiment count. The sentiment text contains the sentences and the sentiment counts contain their binary sentiment polarity as 0 and 1. 1 means the sentence is positive and 0 means negative. After performing clustering through various approaches like k-means, canopy, cobweb, etc. it is found that none of the approaches gave the exact results. Through this empirical research, we are putting forward the most appropriate approach among various approaches which gave better results for Yelp’s sentiment data and similar types of data. This research is concluded by recommending the clustering approach which is having less error rate and execution time as compared to others.