Sentiment Analysis with Deep Learning Methods for Performance Assessment and Comparison

The activity of obtaining and evaluating perspectives of individuals, feelings, mindsets of others, views, and so on, toward various things including subjects, goods, and ideas is known as sentiment analysis (SA), often called sentiment mining (SM). People are producing massive quantities of thoughts and feedback regarding goods, offerings, and daily operations as a result of the quick expansion of using online applications like blogs, social media platforms, and web pages. Companies, government agencies, and institutions can collect and evaluate general population attitudes along with opinions using sentiment analysis to acquire business insight and improve decision-making. The paper represents a complete research on sentiment analysis based on DL (deep learning) approaches to give researchers an idea of the evaluation of feelings and associated disciplines. This research represents the previous studies of emotional analysis and illustrates the methodology of our work. The methodology explains data extraction, data preprocessing, text preprocessing, feature extraction, feature selection, and so on. The dataset applied in the study is an IMDb movie reviews dataset containing equal amounts of samples for training and testing. Then, we discussed sentiment analysis techniques which are Simple Neural Networks (SNN), Convolutional Neural Networks (CNN), and Recurrent Neural Networks (RNN). Using the methods, the outcome states that the simple neural network model generates an accuracy of 74.99% and a Convolutional Neural Network of 85.79%. Besides, the Recurrent Neural Network shows 86.46% which is the highest one. Furthermore, based on the results of the confusion matrix, we investigated the optimum model to attain the highest precision, recall, and F1 score.