Perbandingan Fitur Ekstraksi Untuk Klasifikasi Emosi Pada Sosial Media

Abstract

Emotions are complex conscious experiences characterizing mental states, such as excitement, anger, love, fear, and so on, as part of important human nature. Nowadays, many people express themselves as a reflection of their personality using social media. Social Media grows and becomes a method for social interaction and information sharing. Based on that, researchers tried to use social media data to classify someone's emotions. Emotional detection of text from social media is a field of research that is gaining a high interest, especially for the sake of emotional analysis. To be able to classify such emotions, researchers use comparative feature comparison and algorithms classification. The comparison of features in this research is the extraction features TF-IDF and N-gram which are then classified using Naïve bayes algorithm. However, before the extraction feature is applied, there is a pre-processing text technique using several methods: Case Folding, Stopword  Removal, and Stemming. Based on this research, techniques of extraction features in this research generating the highest accuracy value after the classification method is the TF-IDF with an accuracy value of 80%, 98% for the highest value of precision in measurement of “pleasure” emotion, 99% for the highest recall value is in “happy” emotion, and 95% for the highest F1-score value in “pleasure” emotion.