ResearchBib Share Your Research, Maximize Your Social Impacts
Sign for Notice Everyday Sign up >> Login

AN EFFICIENT SENTIMENT ANALYSIS BY USING HYBRID NAIVE BAYES AND SVM APPROACH IN BANKING INSTITUTIONS

Journal: International Journal of Civil Engineering and Technology (IJCIET) (Vol.8, No. 12)

Publication Date:

Authors : ; ;

Page : 373-391

Keywords : Opinion; Naïve Bayes; Positive Sentiments; Negative Sentiments and Sentiment Score.;

Source : Downloadexternal Find it from : Google Scholarexternal

Abstract

Rapid increase in internet users next to the increasing potency of online review sites as well as social media contains certain role in Emotion examination or Opinion mining that targets at shaping whatever added people believe and comment. In order to improve the facilities in banking sector, the Sentiments or Opinions are analyzed. Views encompass public generated content regarding banking services, policies, customer service, marketing, risk management. Previously, the Naïve Bayes based sentiment classification, Support Vector Machine (SVM) based sentiment classification have been developed. On the other hand, SVM is not appropriate for extensive dataset by reason of its high computational complexity as well as Naive Bayes classifier has shortage in accurateness for most of the composite real-life state of affairs where there subsists dependency amongst features. With the purpose of resolving this issue, the presented method introduced a hybrid Naïve Bayes and SVM based sentiment classification approach. In this proposed work, State bank of India (SBI) and ICICI bank reviews are taken for Sentiment analysis. It contains more reviews about E-banking, home loan registration procedure, Educational loans, customer service, interest rates, and issue with money withdrawal and deposit. In this work, the sentences in the WebPages are extracted and pos tagging is performed. By using Fuzzy Neural Networks with long short-term memory (LSTM) the Dimensionality reduction and Word Sense Disambiguation are performed. Then the word features are extracted by using Conditional Random Fields (CRF).After the feature extraction, feature selection scheme is performed by using Hidden Markov Model -Latent Dirichlet Allocation (HMM-LDA). Dependent upon the selected features sentiment classification is performed using hybrid Naïve Bayes and SVM approach. With the aim of analyzing the positive and negative score value, the sentiment polarity is computed between the candidate words and seed sentiment lexicon which is generated from corpus. Based on this positive and negative sentiment the banking industry provides efficient facilities

Last modified: 2018-05-11 23:14:38