Aspect Level Sentiment Analysis using Stacked Auto Encoder with Random Forest

Sentiment analysis is termed as recognition of emotions extracted from textual features and termed as one of the prominent part of opinion mining. The existing deep learning approach had showed good performance. But, to improve this performance level, a hybrid framework is proposed by combining lexicon features as well as deep aspect features with application of autoencoders in order to solve the limitations of earlier methods. Evaluating and testing of the designed framework flexibility for multiple domains. For the performance evaluation of proposed model, IMDB dataset is used. The simulation is performed on MATLAB platform and this proposed hybrid lexicon and deep aspect level feature extraction model represents better results as compared to other existing works.