Abstract:

Abstract This paper presents the analysis of two models namely Naïve Bayes and Logistic Regression, and a dataset for predicting emotion in a text. The experiment use emotion dataset from KAGGLE website, containing 21,459 data with two columns labelled as Text and Emotion, emotion class consists of happy, anger, sadness, love, fear, and surprise. This is to evaluate the models and dataset applied in this research if it is good and enough for predicting emotion in text. Specifically, to apply data collection, data preparation, feature engineering, model building, and model evaluation. Based on the results, we conclude that Logistic Regression Model gives the best performance. In classification report, the result shows that the accuracy of Naïve Bayes is 77 percent only while Logistic Regression is 89 percent. The result for the best model performance also has the highest percentage of accuracy obtain rather than the previous research discussed in this paper that uses different models. The result of analysis for the dataset is good when it comes for training purposes but for the real time application, the data for each emotion should be balance since the dataset utilized in this research is an imbalance dataset.

Keywords: