PREDIKSI PROFESI BERDASARKAN MODEL BAHASA PADA TWEETS

Abstract

With the advance of social media, people tends to be  very reactive on issues which are happening around the  globe. Everybody can show their opinions freely, and  sometimes uncontrollable, no matters what their job is.  This research investigates the tendency of words choice  in someone’s job based on the style of language he/she  used in his/her twitter account. It is assume that most of  the people in a specific job has the same language used  on social media. The analyses of the study is performed  by using Naïve Bayes classifiers for around 30,000  tweets. The text processing are divided into three main  parts, i.e.: retrieval and grouping of the data, data  processing, and evaluation. The type of jobs which are  analyzed, consists of: politicians, actresses/actors,  musicians, and students, through their official twitter  accounts. The experimental results show that  multinomial Bayes classifiers are more reliable than the  binomial classifiers. Further investigation shows that the  best accuracy is achieved by the unigram model, which  has a mean of 0.73±0.127 in a 5 cross validation setting.  This fact reveals that there is no direct relatioship  between someone’s word choice and his/her profession.