Probit Regressive Preprocessing Based Stochastic Gradient Decision Stump Tree Boosting Sentiment Classification for Recommendation System

Abstract

Sentiment classification plays a significant role on service recommendation. Few research works have been designed for classifying the customer reviews with help of different classification algorithms. However, accuracy of conventional sentiment classification algorithm was not sufficient. In order to overcome such limitations, a Probit Regressive Preprocessing based Stochastic Gradient Decision Stump Tree Boosting Sentiment Classification (PRP-SGDSTBSC) Method is proposed. In PRP-SGDSTBSC Method, customer reviews are taken as an input from large database. After that, data preprocessing is carried out in PRP-SGDSTBSC Method by using Probit Regression model to analyze terms in customer reviews and to remove the stop words. After that, Stochastic Gradient Decision Stump Tree Boosting Sentiment Classification (SGDSTBSC) is carried out in PRP-SGDSTBSC Method to classify the customer reviews as positive and negative sentiment with higher accuracy and lesser time. The designed SGDSTBSC classifier model is an ensemble of several weak classifiers (i.e., decision stump tree). For every weak learner, preprocessed customer reviews are considered as training samples. Then, the weak classifiers are combined to form strong classifier to provide the final results as positive sentiment or negative sentiment. After obtaining the classification results, the recommendation is given to the user for particular item. Experimental evaluation is carried out on factors such as preprocessing time, classification accuracy, error rate and computational time with respect to number of customer reviews.