IMPLEMENTASI OPINION MINING UNTUK PROVIDER INTERNET MENGGUNAKAN METODE NAIVE BAYES

The development of information technology is growing from year to year. To support the smooth flow of information, there are many internet service providers circulating in Indonesia to support their needs. Some of the largest internet service providers in Indonesia such as Indihome, First media, and Biznet Home definitely have their own advantages and disadvantages.At this time, internet provider providers only accept customer complaints or suggestions through the customer service (CS) call center. Meanwhile, many young Indonesians currently use one of the popular Social Media services, namely Twitter as a user-friendly microblogging service so that users can easily use it, especially in delivering messages in the form of tweets. Therefore, a sentiment analysis program was designed for several internet providers in Indonesia. Opinions or Opinions will be analyzed to determine public sentiment. These sentiments will be classified into 3 sentiments, namely negative, positive, and neutral sentiments. The sentiment classification process can be done manually, but if there is too much data, it requires a system equipped with a classification method, so that the determination of classification can be done quickly. The design of this program applies the Naive Bayes Classifier method. Because this method is supervised learning, it requires training datasets with labels. Labeling will be done automatically using the K-means method. K-Means will sort tweets into groups which are divided into 3 labels. The results of the K-means clustering accuracy are 73.4%. The results of this application are divided into 2 parts, namely a pie chart which is divided into slices that describe the results of the percentage of tweet classifications and a table of classification results containing the number, content of the tweet, and the results of the classification. The best level of accuracy in testing uses 220 training data, and 54 training data. The results of the accuracy of 83.3%.