An Adapative Neuro Fuzzy Inference Sentiment Analysis using Nonlinear SVM
Abstract
In this paper we compute the sentiment of social media posts using a set of fuzzy rules involving two lexicons and datasets. The proposed Hybrid adaptive fuzzy system integrates NLP ,word sense disambuigation and SVM-Nonlinear. Using four unsupervised fuzzy rule based system to classify the tweet posts into positive, negative sentiment. we used SVM-Nonlinear to optimize the rule set generated for two classes by ANFIS approach. We perform a comparative analysis of our method on two public twitter dataset, two sentiment lexicons and three approaches (ANFIS,ANFIS-Linear SVM,ANFIS-Nonlinear SVM)for unsupervised sentiment analysis. The fusion of Hybrid Adaptive neuro fuzzy system with SVM-Nonlinear for sentiment classification provides a new paradigm in sentiment analysis. ANFIS-Nonlinear outshines ANFIS and ANFIS-linear in classification of individual tweets to positive ,negative. this classification helps in producing predictions of political assessment with 90% accuracy.
Keywords: Social media, Twitter, Support vector machine, Adaptive neuro-fuzzy inference system, Political analysis.