Feature-Based Sentimental Analysis On Product Review System Using CUDA-BB Algorithm

Abstract

Sentiment analysis methods are classified as feature extraction method, i.e., view's, review's or sentence to predict the emotion of a sentence or a text using natural languages processing(NLP). The analysis of sentiment involves classification of the text into three phases "positive," "negative" or "The Neutral." It Analyzes the data and labels and classified as either "good" "better" "best" or "bad", "worse" based on the emotions or feeling given by customer and finally classified as positive and negative or neutral respectively. So, in regard for the past few years, The World Wide Web (WWW) has become a vast source in providing raw data which is is in the form of opinions or emotions or reviews given by the user or customer about a particular product. E -commerce, has grabbed the attention of the business people to improve the quality of their product by taking the review's from different Social media websites like Facebook , Twitter, Amazon, Flipkart, etc Sentimental analysis or Opinion mining is one of the major challenges of NLP (natural language processing).Business Analytics plays a very important role in the current scenario. In particular, these people rely on the feedback of their products given by the customer's to withstand the competition and knowledge mining that can give them an outstanding view of what to expect in the future..Classification and rule induction describes the main topics in the field of decision making and knowledge discovery. In this paper, we propose CUDABB (CUDA Bag-Boost) algorithm is used to find the overall star rating of Smartphone's using GPU parallel computing. This strategy not only reduces variance and bias but also this approach allows to produce better predictive performance compared to a single model. And the results obtained is compared with SLIQ and MMDBM using pycuda and GPU with computed acceleration rate(speedup) time using Amazon mobile review dataset. The aim of GPU mining technique is to enhance the execution speed with less handling time. Finally, we conclude that the proposed method achieves better accuracy.