A Comparative Study of Supervised and Unsupervised Machine Learning Algorithms on Consumer Reviews

Abstract

For any organization involving consumers, reviews and feedbacks are quite important. For this purpose, the bulk of data is generated from various social networking sites in terms of reviews and feedbacks. In order to understand consumer’s perception about an item, this research scrutinizes various supervised and unsupervised machine learning algorithms on two data sets. A comparative analysis is made for deliberating the efficiency of these algorithms on distinct datasets for text classification. This research is an attempt to find the best fit classifier for consumer’s perception using sentiment analysis. So, in order to accomplish this objective, firstly text preprocessing techniques are applied on datasets then feature extraction techniques are applied on the processed data. Thereafter, classification and clustering are applied using supervised and unsupervised machine learning algorithms respectively. Further, these algorithms are evaluated and the result reveals that supervised machine learning algorithms especially Support Vector Machine (SVM) outperforms unsupervised machine learning algorithms for garments dataset. And Naive Bayes (NB), Logistic Regression (LR) outperforms unsupervised machine learning algorithms for restaurant dataset.