Movie Sentiment Analysis Using Data Augmentation And LSTM-Recurrent Neural Network

Research in the field of data mining, especially text-based, is growing with the discovery of new methods that have good performance. Sentiment analysis is a field of research that is still widely studied in text classification. Various approaches, including Machine Learning, have been tried in sentiment analysis research, but as the dataset grows, these methods do not produce optimal accuracy performance. The Deep Learning method is a solution that is often used by researchers to produce maximum performance on large datasets, but in some studies, if the number of datasets is limited, the training model cannot be carried out optimally. As a solution, data augmentation is an approach that can be used to increase the quantity of datasets so as to encourage better performance of a model. This study uses a sentiment analysis dataset regarding various films on Twitter, but with a limited amount of data, so a data augmentation approach is needed to increase the quantity. Text-based data augmentation techniques combined with the Deep Learning approach, namely the Recurrent Neural Network (RNN) and Long Short Term Memory Network (LSTM) methods are the approaches we propose in this study. RNN LSTM has the ability to Handling Long-Term Dependencies and Effective in Sequence-to-Sequence Tasks so that the performance of the model on the sentiment analysist will be more optimal The results show that the proposed model has succeeded in classifying the sentiment analysis dataset for film reviews by showing a model accuracy rate of 70.45%.