A Bayesian Deep Learning Model for Sentiment Analysis Based Product Review Prediction

Abstract Identifying the nature and polarity of customer reviews is critically important for businesses, typically based on opinion mining and sentiment analysis. It has a wide range of applications in information systems, including classifying reviews, summarizing review and other real time applications. There are promising possibilities to use sentiment analysis in real time business models. This work analyses the use of sentiment analysis for analysing customer reviews. Customer reviews are a valuable source of feedback for businesses. In this approach, the customer review dataset from Amazon reviews has been analyzed. Pre-processing of raw data has been done prior to using it to train a neural network. The regularization based Bayes Optimized Deep Neural Network has been used for sentiment classification from social media dataset. To compare the performance of the proposed system against existing research in the domain, the predication error metric has been computed. From the performance of the proposed system, it can be observed that the proposed system clearly outperforms the existing approaches in terms of prediction error and accuracy. Keywords: Machine Learning, Sentiment Analysis, Deep Neural Networks, Regulrization, Mean Absolute Error, Accuracy.