A systematic review of sentiment analysis classification; Bi-LSTM, LSTM, SVM and artificial bee colony (ABC)

The sentiment classification is a real time process of automatically detecting the reviews in text and classifiedthose emotions such as negative, positive, or neutral. Most of the research focuses on sentiment analysis based on neural network methods for extracting the text from the reviews. The sentiment analysis is similar to the classification, where the extracted features are fed to the classifier as an input to predict an output. In this study, the hybrid method of two deep learning models are long short term memory (LSTM) and Convolution neural network (CNN) applied in the classification of sentiment reviews. The big data of sentiment analysis based on the topics of sensitive information. The hybrid method of Multi-head Attention (MHAT) with Bidirectional Long-Short Term Memory (BiLSTM) based on Chinese text social media for sentiment analysis. The Gated recurrent unit (GRU) has the limitations such as less convergence rate, decreases the learning efficiency and under fitting problem. The SVM has the limitation of large dataset is not suitable, more complexity and noise and the RNN model has the exploding vanishing gradient issues.