Implementation of Naive Bayes Algorithm on Sentiment Analysis Application

Abstract

Natural Language Processing (NLP) is one of the branches of artificial intelligence science where this branch science is the basis for developing sentiment analysis. The application in NLP in sentiment analysis includes Pre-processing text consisting of featured selection and tokenization. For the classification process, the determination of the algorithm is determined by comparing the results of the classification predictions of naïve Bayes, Weighted Instances, and Zero-R with the data that has been calculated for its frequency terms. The results of the testing analysis showed naïve Bayes had a stable accuracy after being tested with an accuracy value of 99.62% in the training data and 94% in the Testing Data, with an average classification failure of 0.13%. The results of the acquisition of words are used as a corpus for the construction of sentiment-level sentence analysis applications. The application development by the Naive Bayes algorithm was built using the PHP programming language and literary library. The method of application development using the waterfall starts from the analysis process to the application implementation. Based on testing the accuracy of 30 comments classified by the system, it produces an accuracy value of 86.66%. However, the accuracy of comments that have been classified as applications were retested using machine learning Weka which resulting in an accuracy value of 93.33%. The difference in accuracy is due to the Naive Bayes algorithm in utilizing the appearance of words to form a sentiment classification.