Abstract:
Sentiment analysis is one of the sciences in Text Mining which is used to classify text
documents in the form of public opinion. The text document used in this study comes
from Twitter using three positive, negative, and neutral labels. The method used was the
Support Vector Machine. This study used 2000 data using two data test scenarios,
namely, Split Validation and Cross Validation. Testing data with Split Validation was
divided into 4 variations, namely, 50%:50%, 70%:30%, 80%:20%, and 90%:10%,
while testing data with Cross Validation was divided into 10 tests. The highest accuracy
result used the Split Validation test laid in a comparison of 90%: 10%, which is 83.5%,
while data testing with Cross Validation had the highest accuracy at K-Fold 7 of
86.0%.
