Title: A multi-level text classifier for feedback analysis using tweets to enhance product performance

Authors: Balamurugan Balusamy; Thusitha Murali; Aishwarya Thangavelu; P. Venkata Krishna

Addresses: School of Information Technology and Engineering (SITE), Institute of Technology (VIT), Vellore – 632014, India ' School of Information Technology and Engineering (SITE), Institute of Technology (VIT), Vellore – 632014, India ' School of Information Technology and Engineering (SITE), Institute of Technology (VIT), Vellore – 632014, India ' School of Computer Science and Engineering (SCSE), Vellore Institute of Technology (VIT), Vellore – 632014, India

Abstract: Big Data refers to the collection and storage of the enormous amount of data which is heterogeneous in nature. Data analysis is quite complex due to its enormous volume and its high generation speed. Big Data has many business applications such as in promotion, marketing either financially or by supporting in decision making. One such application is the sentiment analysis that paves the way for the business analysts to know the positive or negative impact over the product based on the tweets by the people. We propose a three-level text classifier with the first level as principal components analysis (PCA) followed by the support vector machine (SVM) and the conditional random fields (CRF) as the second and third level using the tweets collected. This feedback analyser would promote the sales of the product due to its high accuracy in feedback classification.

Keywords: big data; three-level text classifiers; principal components analysis; PCA; support vector machines; SVM; conditional random fields; CRF; feedback classification; tweets; product performance; sentiment analysis; Twitter; social media.

DOI: 10.1504/IJEMR.2015.073455

International Journal of Electronic Marketing and Retailing, 2015 Vol.6 No.4, pp.315 - 338

Received: 04 Feb 2015
Accepted: 07 Mar 2015
Published online: 09 Dec 2015 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article