Unmasking the Best ML Algorithm for Sentiment Analysis.

Abstract

The enormous quantity of information that we encounter every day on the Internet originates from a variety of sources: from online review platforms to social media. Such an avalanche of data conceals a valuable opportunity for text analysis-the estimation of subjective opinions, and most importantly, their sentiment or polarity. Sentiment analysis and text classification are useful in drawing valuable information out of data by classifying text into certain classes based on their content. The paper presents a comparative study of five algorithms for text classification, namely Logistic Regression, Decision Tree, Random Forest, Support Vector Machine (SVM), K-Nearest Neighbors (KNN), using a Twitter Comments dataset. We observed that a support vector machine (SVM) and Logistic Regression algorithms are doing better on all the metric scores in comparison to its peers, namely accuracy, precession, recall, and F1 score.