Emotions Comparison of Commuter Line and Transjakarta Users Based on Twitter Using Multinomial Naïve Bayes Classifier
Denis Eka Cahyani, Fransiska Meilani, Sari Widya Sihwi · 2019 5th International Conference on Science and Technology (ICST) · 2019
The availability of public transportation is a serious concern of the DKI Jakarta Government. The government provides a variety of public transports such as Commuter line and Transjakarta. User satisfaction with transportation services can display in social media such as Twitter. Tweets about service satisfaction can be emotional or not, this can be known through classification. One classification method that can be used is Multinomial Naive Bayes (MNB). The research stages in this classification include data collection, data, text preprocessing, TF-IDF weighting, MNB classification using 10-fold cross validation, and evaluation. The classification consists of two stages, the first stage to classify tweets contains emotions or not, the second stage classifies emotionally tweets into five classes namely happy, angry, sad, fear and surprised. The accuracy result of first stage classification for Commuter line, Transjakarta and mixed data were 84.40%, 81.78%, and 83.52%. The accuracy results of the second stage classification for the data of Commuter line, Transjakarta and mixed data were 91.85%, 91.21%, and 91.53%. The accuracy of balanced data with Random Oversampling (ROS) for Commuter line, Transjakarta, and mixed data were 96.18%, 96.55%, and 95.63%. Furthermore, conduct a comparative analysis of emotions based on the results of accuracy, word clouds, and time series graphics. The time series results show that the increase in Transjakarta angry emotions is higher than Commuter line, this is due to more factors in Transjakarta disruption. And the word cloud result shows the most common term used for the two data set is 'tunggu' or 'wait', this indicates that both transportations have the same problem, which is a long waiting time for transportation.