@phdthesis{digilib47012,
           month = {July},
           title = {PENERAPAN ANALISIS SENTIMEN PADA PENGGUNA TWITTER MENGGUNAKAN METODE CONVOLUTIONAL NEURAL NETWORK DAN NAIVE BAYES (STUDI KASUS : KARTU PRAKERJA)},
          school = {UIN SUNAN KALIJAGA YOGYAKARTA},
          author = {NIM. 17106050009 PAHLEVI WAHYU HARDJITA},
            year = {2021},
            note = {Nurochman, S.Kom., M.Kom},
        keywords = {Sentiment Analysis, CNN, NBC, Prakerja, Twitter},
             url = {https://digilib.uin-suka.ac.id/id/eprint/47012/},
        abstract = {During pandemic, the government is promoting the Prakerja Card program. Many Indonesian people have voiced their opinions through social media Twitter regarding the Prakerja Card. This results in data that can be processed to be used as useful information in sentiment analysis. The main task in sentiment analysis is analyzing the data and then classifying the tweet text into one of the classes, namely positive, negative or neutral. Therefore, this research tries to analyze sentiment about the prakerja card program using the Convolutional Neural Network and Naive Bayes methods.
This research aims to determine the best Convolutional Neural Network architecture, apply the Convolutional Neural Network (CNN) and Naive Bayes (NBC) methods and compare the performance of the Convolutional Neural Network method with the Naive Bayes method on sentiment analysis about Prakerja cards. CNN is a deep learning algorithm that has many types of functions and layers so that it can be used in sentiment analysis, while NBC is a supervised learning algorithm that is often used in sentiment analysis. The data used is tweet data about the Prakerja Card program obtained by crawling. Data analysis use confusion matrix and k-fold.
In this research, it was concluded that the CNN model with the GlobalMaxPooling layer was the best model of the other two CNN models. Sentiment analysis got the best accuracy of 79\% on the CNN method and NBC got 76\% accuracy. In K-fold with 5 classes, the best accuracy result is 81.9\% on the CNN model with a learning rate optimization of 0.00158. While the average accuracy on NBC only reached 75.3\%.}
}