<![CDATA[Nowadays consumers are increasingly making their opinions and experiences online. Reading those reviews are time-consuming, but, if only few reviews were read, the evaluation would be biased. Sentiment analysis aims to solve this problem by automatically classifying user reviews into positive or negative opinions. Naive Bayes classifier is a popular machine learning technique for text classification, because it is so simple, efficient and it has a great performance in many domains. However, it has a lack that it is highly sensitive to the high number of feature. Therefore, in this research the concatenation of feature selection methods is used, that is Information gain and Genetic algorithm that could increase the accuracy of Naive Bayes classifier.This research turns out text classification in the form of positive or negative from book reviews. The measurement is based on the accuracy of Naive Bayes before and after adding the feature selection method. Evaluation was performed using 10 fold cross validation. Whereas the measurement of accuracy was measured by using confusion matrix and ROC curve. The result of this research is the improvement of accuracy of Naive Bayes from 75.50% to 84.50%.]]>
Copyrights © 2014
