PERBANDINGAN ALGORITMA KLASIFIKASI NAÃVE BAYES DAN SUPPORT VECTOR MACHINE UNTUK SENTIMEN ANALISIS APLIKASI MYPERTAMINA
Total View This Week0
Total View This Week0
Institusion
INSTITUT TEKNOLOGI DIRGANTARA ADISUTJIPTO
Author
Ariwardana Putra, Resnu
Subject
Q Science (General)
Datestamp
2024-02-27 03:21:01
Abstract :
Pertamina opens a lot of input for services provided in order to provide services to consumers, ranging from social media accounts to applications provided by Pertamina, namely MyPertamina as a place for customer activities in making transactions. Currently, MyPertamina on the Google Play Store has been downloaded with a rating of 2.4 and 302 thousand reviews (September, 2022). A fairly low rating accompanied by various negative and positive reviews shows that the services provided by MyPertamina have not fully met the expectations of MyPertamina users. The comparison at this stage is comparing the accuracy results of each method, namely the Support Vector Machine and Naïve Bayes methods. The test method for calculating accuracy for both Support Vector Machine and Naïve Bayes data was performed using a confusion matrix by comparing all testing labels with training. In the Support Vector Machine algorithm, the precision value for the positive class is 100%, the negative value is 94.52%, and the Naïve Bayes algorithm has positive precision of 36.84%, negative is 98.78%. The success rate of the system in rediscovering information for the positive class Support Vector Machine algorithm was 3.53%, negative class by 100%, and negative class Naïve Bayes algorithm by 97.03%, positive 28.82% system performance was very low in terms of system success in rediscovering positive classes in the data. The review data collected was 3000 after removing duplicates into 2998 reviews latest July 19, 2023 and back on the MyPertamina application from the Google Play site. Based on the sentiment class classification process, the number of negative class reviews was 2,828 and positive reviews 170 reviews. Based on the analysis conducted with Rapidminer using the Support Vector Machine algorithm has an accuracy of 94.53% better than Naïve Bayes 93.16%.
Pertamina opens a lot of input for services provided in order to provide services to consumers, ranging from social media accounts to applications provided by Pertamina, namely MyPertamina as a place for customer activities in making transactions. Currently, MyPertamina on the Google Play Store has been downloaded with a rating of 2.4 and 302 thousand reviews (September, 2022). A fairly low rating accompanied by various negative and positive reviews shows that the services provided by MyPertamina have not fully met the expectations of MyPertamina users. The comparison at this stage is comparing the accuracy results of each method, namely the Support Vector Machine and Naïve Bayes methods. The test method for calculating accuracy for both Support Vector Machine and Naïve Bayes data was performed using a confusion matrix by comparing all testing labels with training. In the Support Vector Machine algorithm, the precision value for the positive class is 100%, the negative value is 94.52%, and the Naïve Bayes algorithm has positive precision of 36.84%, negative is 98.78%. The success rate of the system in rediscovering information for the positive class Support Vector Machine algorithm was 3.53%, negative class by 100%, and negative class Naïve Bayes algorithm by 97.03%, positive 28.82% system performance was very low in terms of system success in rediscovering positive classes in the data. The review data collected was 3000 after removing duplicates into 2998 reviews latest July 19, 2023 and back on the MyPertamina application from the Google Play site. Based on the sentiment class classification process, the number of negative class reviews was 2,828 and positive reviews 170 reviews. Based on the analysis conducted with Rapidminer using the Support Vector Machine algorithm has an accuracy of 94.53% better than Naïve Bayes 93.16%.
Download
book
BibTex
Latex, Jabref
cloud_download
Original Resource
url resource Institution
