Classification of South Jakarta Language Sentence Structure (Indonesian-English Code Mixing)

Twitter can present a variety of information that makes various users from various parts of the world always update on what is being discussed (tweet). Various information is presented in various ways and delivery, one of which is by using the South Jakarta language style mix which is a code mix that mixes various vocabularies using both Indonesian and English in their conversations so that they are considered current for young people in Indonesia. Based on this phenomenon, sentence classification is made using the Support Vector Machine and Random Forest algorithm methods to classify the types of code mixing, namely alternation, insertion, and congruent lexicalization. Attributes on the data were Tag and Tag + Text. Text weighting is used with unigram, bigram, and trigram. The evaluation of the research used 10-fold cross validation for training data and confusion matrix for test data. The best data test results on Tag data obtained by Support Vector Machine using bigram obtained an accuracy value of 97.33%, while the best data test results on Tag + Text data obtained by Random Forest algorithm using unigram obtained an accuracy value of 92%.