| dc.description.abstract | Activities performed by an individual can reflect various emotions they possess, both directly and indirectly. Thus, emotions can influcence a person of their actions dan decision-making. Emotion analysis is a fundamental part of designing systems to understand the basic behavior of a human through emotional data such as text, voice, and facial expressions. Analyzing emotions from various text data through the internet can have diverse benefits, such as assisting organizations in analyzing comments or feedback given by customers and preventing suicide. Emotion analysis is an initial step in solving problems related or sourced from an emotion. Classification can be done by training a model using machine learning algorithms. Machine learning algorithms generally perform poorly with imbalanced training data, where the data categories are not evenly distributed. The problem of imbalanced data can be addressed by assigning different weights to each class category during model training. The aim of this research is to analyze emotions in an imbalanced dataset. The emotions analyzed in this study include love, joy, surprise, anger, sadness, and fear. Feature extraction from the text is performed using Term Frequency-Inverse Document Frequency (TF-IDF). The Class-Weighted Support Vector Machine method is used as the classification model's learning algorithm. The evaluation in this research used the Matthews Correlation Coefficient metric and obtained a value of 0.74. The conclusion is that the prediction results are quite good, approaching and agreeing with the intended target. | en_US |
