An Improved Grey Wolf Optimization Algorithm for Enhancing Two-Pass Opinion Mining Classifier using Patient Reviews

The Sentiment analysis on various product reviews is a fascinating area of natural language processing and web text mining. Our Objective is to analyze the effect of an artificial neural network based method for drug based opinion classification. In the research that has done so far on sentiment analysis, ANNs have been considered rarely. In this work feed forward neural network architecture has been studied for sentiment classification. When back-propagation learning algorithms are used to train neural networks, they are typically slow and higher learning rates are needed. The selection of an appropriate learning rate is a complex issue. Proper learning rate is necessary to overcome the problem connected with back-propagation algorithms. More effort need to be put forth in the training of a neural network due to its nonlinear nature .Also identifying the unknown best set of main controlling parameters (weights) is a task where a quicker learning rate will cause unsteady learning. At the same time, a slower learning rate causes a disproportionately longer training time. It is quite demanding to discover an easy method for decide on the learning rate. Forgetting the optimal solution, Usually flawless weight connections are not formed by the back-propagation learning algorithm In this study, a Grey Wolf Optimizer algorithm was revised. For optimizing the weight of a neural network, this modified version was applied to the two pass opinion classifier.. When NN is weight optimized by IGWO, our Two Pass classifier is found to perform better and yields higher accuracy in classification. Accordingly, in my previous research, an efficient two-pass classifier system for patient opinion mining to analyze drugs satisfaction is proposed. Then, the selected features are given to optimal two-pass classifier. The two-pass classifier is a combination of support vector machine (SVM) and neural network (NN). Here, the weight value of (SVM-NN) classifier is optimally chosen with the help of improved gray wolf optimization (IGWO) algorithm. Finally, in the classification stage, we attain an opinion of corresponding drugs. The performance of proposed methodology is evaluated in terms of precision, recall and F-measure.