Analysis on Selection of Presidents and Vice Presidents of the Republic of Indonesia 2019 Using Support Vector Machine Method

Abstract

2019 is a political year in Indonesia, because in that year the General Elections of the President and
Vice President of the Republic of Indonesia were held for 2019-2024. Ahead of the election will be a lot of
information and opinions about the two pairs of candidates. Twitter social media is generally often used by the
public to express opinions about things, including for presidential and vice presidential elections. To find out
the community's sentiments, it is necessary to do a management process of the many existing tweets, it can be
done with a sentiment analysis system. Sentiment analysis in this study was made using a machine learning
approach. The sentiment analysis system developed in this study consists of three stages, namely preprocessing,
feature extraction and classification. In the preprocessing phase consists of cleansing, case folding, spell
checking, stopword removal, stemming, convert negation and tokenization, then for feature extraction stage the
TF-IDF (Term Frequency - Inverse Document Frequency) algorithm is used, which is widely used in
combination with LSA (Latent) Semantic Analysis). With the use of LSA which contains SVD (Singular Value
Decomposition) algorithm, it can decipher sparse matrices obtained by eigenvectors from the TF-IDF
algorithm, thereby reducing text storage space, and increasing the efficiency of text classification. The last stage
is the classification process to calcify data into positive or negative sentiments, the algorithm used is Support
Vector Machine. From the results of the research, the system was built to carry out the classification process as
planned. And from the test results obtained an average result for accuracy, precision, recall and f1-score around
0.7.