The Judgement of the Reviews of the Movie Oppenheimer by the LSTM Model in Deep Learning

Abstract

The main purpose of this paper is to make the LSTM model predict whether 100 reviews of the movie Oppenheimer are positive or negative. We obtained 100 reviews of the movie Oppenheimer in the portal site Naver in 2024. While we classified 70 reviews into train data, we classified 30 reviews into test data. Also, we classified 20% of train data into validation data. We used a hidden layer to increase the accuracy of the LSTM model. We used sigmoid as activation and rmsprop as optmizer. A major point of this paper is that there was a gradual fall in the loss of train data. More specifically, there was a continual decline in the loss of train data from epoch 1 to epoch 12. However, there was a slight rise in the loss, but there was a steady decrease in the figure from epoch 14 to epoch 25. A point to note is that the accuracy rate of the LSTM model reached a peak when epoch was 7 (92.86%). A further point to note is that the val_loss of validation data increased to 0.8977 when epoch was 12. It is worthwhile pointing out that the val_accuracy of validation data did not improve even though learning took place 25 times. It increased to 64.64%, but there was a slight decline in the val_loss of validation data when epoch was 12.