Perbandingan Algoritma Sentimen Analisis media data Twitter Pilgub Jabar 2018

Abstract

This paper is published to review and compare the performance of the SVM algorithms and CNN algorithms as an update composition in analyzing sentiment with tweeter attributes, the comparison of these algorithms using the Python application as a tool to support machine learning. The classification of negative, neutral and positive sentiments in the tweet dataset is tested and to determine and measure the accuracy, precision, recall, f_Measure and configuration matrix weights of both the SVM algorithms and CNN algorithms. The tools used with the Python Jupyter application, Tensorflow, Noted ++ are applied to the Indonesian language Twitter classification, the measurement results are precise and accurate according to human measurement parameters related to tweet data sentiment on Twitter social media commenting on the election of the Governor of West Java with the candidate for governor and deputy governor for the period 2018-2023. The results of this study, Testing Experiments before stemming with the SVM Algorithm was carried out seven times with an average accuracy rate of around 67%, and the CNN Algorithm before stemming also with seven trials with an average accuracy of around 67%, then the Testing Experiment after stemming. with SVM conducted seven trials the average accuracy rate was around 67%, while CNN algorithms before stemming was also carried out with seven trials with an average accuracy of about 52% lower than SVM algorithms