IMPLEMENTASI METODE MACHINE LEARNING UNTUK KONTEN TEKS EMOSI

The objective of this study is to examine and identify emotions in text by
employing a CNN-BiLSTM model that has been improved with an attention
mechanism. The results suggest that the CNN-BiLSTM model successfully captures
both preceding and subsequent contexts in the text, enhancing the comprehension
of emotional information. The utilization of both CNN and BiLSTM enables the
model to acquire semantic representations of words and dependencies inside
sentences with more efficacy. Incorporating an attention layer effectively assigns
higher importance to pertinent phrases, thus enhancing the model's accuracy in
detecting emotions. This technique also offers valuable insights about the text's
most influential components for the model's decision-making process. The concept
effectively categorizes seven emotional labels: sadness, anger, joy, surprise, fear,
love, and neutral. By incorporating the "neutral" label, the model becomes capable
of identifying text that lacks intense emotions, hence enhancing the
comprehensiveness of the analysis. The CNN-BiLSTM with attention model
demonstrates superior accuracy in emotion recognition compared to other methods
such as CNNs, LSTM, SVM, and naive Bayes on the emotion detection dataset. We
assess the model's performance by utilizing metrics such as accuracy, precision,
recall, and F1-score. The model demonstrates excellent accuracy on the training
data, achieving 94.30% with a batch size of 32 and 91.05% with a batch size of 64.
The model attains an accuracy of 91% on the validation data. This study has
important implications for multiple disciplines, such as mental health and social
media analysis, as it enhances our comprehension of user preferences, opinions,
and mental states