A Study of Deep Learning Classification Methods vs. Conventional Machine Learning for Sentiment Analysis

Abstract

People nowadays are more and more turning to online social networks and services as a means of communication and advocacy. In order to ascertain the general public's stance on a certain issue, product, or subject, sentiment analysis involves cataloguing and classifying these views. With each passing day, sentiment analysis becomes more and more crucial. The capacity for computers to learn new things without being explicitly taught is the result of machine learning, a subfield of computer science. Among the many branches of machine learning, "Deep Learning" focuses on algorithms that use neural implementations, such as neural networks, neural beliefs, and so on. Evaluating feelings for a certain set of data requires the use of the most practical and precise method possible as this has implications for both producers and consumers. Various machine learning, deep learning, and hybrid approaches are compared in this project's proposed research. After comparing their accuracy for sentiment analysis, it's clear that deep learning algorithms often provide higher outcomes. On the other hand, there are situations when the disparity in accuracy between the two approaches is not significant enough to warrant using one machine learning method over the other, especially when other, more practical methods are available.