Human Behavior Prediction based on Opinions using Machine Learning Techniques

Abstract

Prediction is the way of identifying the behavior of a person towards online shopping by analyzing the reviews publicly available on the web. In the present study, machine learning approaches are used to extract reviews from the web and segregate and classify them in to five categories, namely, strongly positive, positive, neutral, negative, and strongly negative, for the prediction of human behavior. Several pre-processing methods (including stop-word removal) are applied and web crawler is used to gather the data. This is followed by the application of Stanford POS tagger for tagging the reviews, which is done after stemming by using the porter stemmer algorithm. Analysis of a person’s behavior is performed and experimental results are compared with machine learning approaches.