ANALYSIS OF EMPLOYMENT SENTIMENT IN THE INDONESIAN TELEMATICS FIELD USE MULTINOMIAL NAIVE BAYES AND VECTOR SPACE MODEL

Abstract

Indonesia in 2030 experienced a demographic bonus in the sense that Indonesia would have far more labor supply than in previous years. Then there is a discourse that this 4.0 industrial revolution will replace a lot of work, especially low-skilled work or does not require special skills and rough jobs replaced by machinery and artificial intelligent (AI). To obtain the value of the percentage of positive, negative and neutral sentiments from the public regarding the impact of the industrial revolution 4 against labor and employment on online news media sites and social media Twitter, the authors conducted a study "analysis of employment sentiment in Indonesian telematics using multinomial naïve bayes. " The author uses the preprocessing stages including the case folding, tokenizing, stopword, and stemming. Then weighting with Term Frequency - Invers Document Frequency (TF-IDF). After that the classification stage was done using the multinomial Naive Bayes Classifier method and compare it with the Vector Space model classification. The evaluation used is the Confusion Matrix evaluation method. This study produced an evaluation value in the multinomial method of Naïve Bayes for news data to produce an accuracy of 81.75%, average precision 82.77%, and the average recall of 78.15%. Whereas with the Vector Space model method for news data produces an accuracy of 67.88%, average precision 65.59%, and the average recall of 70.56%. On Twitter data with the Multinomial Naïve Bayes method resulted in an accuracy of 88.80%, average precision 93.75%, and the average recall of 74.44%. On Twitter data with the Vector Space Model method resulted in 85.60% accuracy, average precision 76.44% and average recall of 86.07%.