Bigram

Trigram

Mathematics

Naive Bayes classifier

Politics

Support vector machine

Machine learning

Random forest

Political science

Artificial intelligence

Natural language processing

Geometry

Computer science

Ensemble learning

Boosting (machine learning)

Voting

Sentiment analysis

Word (group theory)

With increasing interest in e-commerce and online shopping recently, customer reviews are one of the most important elements that determine the satisfaction of customers about the products. Also, it gives a comprehensive picture of products to business owners. Hence, the aim of this paper is to conduct a sentimental analysis approach on a set of customer reviews collected from Amazon. As will, classify each review into one of these class: positive review or negative review by using ensemble machine learning method. The ensemble machine learning method used in this research is Voting which combined five classifiers: Naive Bayes, Support Vector Machines (SVMs), Random forest, Bagging and Boosting. All the Experiment in this paper done using Weka. We test six different scenarios to evaluate our proposed model against the five classifiers. The scenarios are using unigram (with/without) stop of word removal, using bigram (with/without) stop of word removal and using trigram (with/without) stop of word removal. The result shows the random forest technique give the highest accuracy which equals to 89.87% in the case of using unigram and with a stop of word removal but voting algorithm shows the best performance on the other scenarios.

Sentiment Analysis of Customer Reviews Using Ensemble Method