Ensemble Classifier for Praise or Complaint Classification and Visualization from Big Data

With the advent in Big Data Analytics, IoT and Machine Learning newer opportunities are created for Business organizations to analyze, monitor and mine user-generated contents in real time for business intelligence using cognitive IoT. Customers share their opinions online through social media platforms like review sites, Twitter and Facebook, etc. Sentiment analysis combined with real-time reporting can provide precise valuable contextual insights enabling more improved decision making. The existing sentiment analysis techniques identify only positive, negative or neutral sentiments and do not consider informativeness of reviews while analyzing the sentiments. The extreme opinions like praise and complaint sentences are informative subsets of positive and negative sentences and are very difficult to find. This chapter proposes the Ensemble classifier using linguistic features for praise or complaint classification from big customer review datasets and visualization of it. The Praise and Complaint sentences are further classified based on aspect and analysis at aspect level is presented from business intelligence point of view. The performance of the four different supervised machine learning classifiers, namely Random forest, SVC, KNeighbours, MLP with linguistic hybrid features and Ensemble of above algorithms is evaluated on Hotel and Amazon product reviews dataset using parameters Accuracy, Precision, Recall, and F1-score. The proposed approach has given excellent 99.7% Accuracy and 99.6% F1-Measure and outperforms existing approaches.