<![CDATA[In today's era, transportation is one of the most vital needs. according to BPS Indonesia (badan pusat statistik) in 2017, motorbikes are the most owned form of transportation in Indonesia. one of the most commonly used motorbike brand is honda. astra honda motor provides a range of services available to its customers. of all their services, honda's customers would surely have a number of opinions or feedbacks regarding their services, whether they are positive or negative. in order to classify and identify aspects of the public opinion, naive bayes with lexicon-based features are implemented to classify them and dbscan is implemented to cluster them by taking the top 3 terms generated from each document. the dataset used in this paper consists of 100 datas divided into an equal set of each class, and 25 datas for testing in which 13 are classified positive and 12 are negative. the result of the classification process applying naive bayes with lexicon-based features is a precision value of 53%, recall of 63%, f-measure of 58% and 60% of accuracy. While, the result of the aspect identification came down to 41,4% for precision, 80,5% for recall, 54,6% for f-measure and a mere 37,6% for its accuracy level. as for the cluster evaluation with silhouette coefficient, the best parameter values using dbscan is an epsilon of 0.1 along with minpts of 1.]]>
Copyrights © 2020
