Performance analysis of Naive Bayes and support vector machine algorithm on classification of Nepali opinion text

Abstract

Opinion is a subjective expression of individual on something. These are views, emotions or sentiments. The opinion helps individual and organization to make decision about the certain things. The opinion classification is the process of analyzing the view or opinion using the natural language processing techniques. The Naïve Bayes and Support Vector Machine (SVM) algorithm are supervised machine learning algorithm for classification. Most of the researches in opinion classification are done in English language but it is important to perform the opinion classification in Nepali language as the amount of data in Nepali is increasing rapidly in the form of blog, review, opinion column in newspaper. Nepali sentences were collected from the opinion section of different online portal of national newspaper in this study. The python programming language was used for implementing both algorithms with NLTK library and output were analyzed on the basis of performance metrics. The accuracy of SVM is 85% which is higher than accuracy of Naïve Bayes algorithm i.e. 83% on preprocessed the data. The accuracy of both algorithms was improved after preprocessing as compared to without preprocessing the data. The Study concluded SVM model was the best model with higher values of performance metrics and is recommended for opinion classification of Nepali text data over the Naïve Bayes algorithm.