Sentiment Analysis Using Hybrid Model of Stacked Auto-Encoder-Based Feature Extraction and Long Short Term Memory-Based Classification Approach
Iqra Kanwal, Fazli Wahid, Sikandar Ali, Ateeq Ur Rehman, Ahmed Hussein Alkhayyat, Akram Al-Radaei · IEEE Access · 2023
The reviews of customer about a brand or product, movie reviews, and social media reviews, all can be analyzed through sentiment analysis. Sentiment analysis is used to identify the emotional tone of language in order to comprehend the attitudes, opinions, and feelings represented in online reviews. As for large data, it is a task that can take a lot of time and can be automated as the machine will learn through training and testing of data. Previously, various standard machine learning and deep learning models namely Recurrent Neural Network (RNN), Convolutional Neural Network (CNN), Long Short Term Memory (LSTM), Naïve Bayes (NB), Support Vector Machine (SVM), Gated Recurrent Unit (GRU) have been used. The key issue on which our research is focused on is that when text is provided to LSTM directly, it cannot adequately extract informative features from the text, leading to less accurate findings. The Stacked Auto-encoder’s softmax layer, when used directly to categorize the extracted features, is power-constrained and unable to do so accurately. A hybrid of the Stacked Auto-encoder and LSTM model has been proposed. SAE is used for relevant informative feature extraction. LSTM is used for further classification of sentiments based on extracted features. The proposed model has been evaluated on an IMDB dataset by splitting it in 5 different training testing ratios using performance evaluation metrics viz. confusion matrix, classification accuracy, precision, recall, sensitivity, specificity, and F1 score. The hybrid results perform best at a ratio of 90/10, and classifying sentiments with 87% accuracy. The proposed hybrid model’s accuracy is better as compared to standard models namely RNN, CNN, LSTM, NB, SVM, and GRU.