This study aims to: (1) implement the Twitter API to
retrieve tweet data on government policies to increase fuel prices, (2) analyze
positive and negative sentiment tweet data on government policies to increase
fuel prices using the Support Vector method. Machine. (3) implementing the
results of processing positive and negative sentiment data on government
policies in increasing fuel prices using the Support Vector Machine method to
obtain accuracy.
The method used in case studies of government policies
in increasing fuel prices begins with literature studies, needs analysis, data
collection, design, evaluation and testing of models, and analyzes models using
Support Vector Machines.
The results obtained (1) data collection was carried out
using the Google Collab extension. As for the process of collecting tweet text
data on Twitter social media regarding government policies against increasing
fuel oil taken from the results of scraping with the Twitter API with the
keyword "bbm goes up" with a total of 493 tweet data taken from
February 2023 to March 2023, (2) positive sentiment tweet data and negative
sentiment towards government policy in raising fuel prices using the Support
Vector Machine method, after being labeled, 244 data obtained positive
sentiment results, and 249 negative sentiment labeled tweet data. If presented,
the positive data is 49.49% and the negative data is 50.50%. (3) Data that has
been preprocessed is then divided into training data and test data. At a ratio
of 8:2 each for training data and test data. The amount of training data
obtained is 80% and test data is 20%. Classification in applying the Support
Vector Machine method uses a scale of 10 with training data and test data that
has been weighted using TF-IDF weighting. The best accuracy results for the linear
kernel with a ratio of 8:2 produce an accuracy of 82.88% Precision of 83.83%,
F1-Score of 83.83%, and Recall of 83.83%. Sentiment analysis of the Support
Vector Machine algorithm model produces that the left is Actual and the bottom
is the Prediction result so that it can be stated that there are 43 True
Negatives (TN), 9 False Negatives (FN), 8 False Positives (FP), 39 True
Positives (TP) ).
Keywords: Twitter
API, fuel prices, positive-negative sentiment tweets, Support Vector Machine.