An Analysis of Machine Learning Techniques for Predicting and Evaluating the Popularity of Online News Articles

In the world today, online news content's reading and sharing are increasing with each passing day. Shares, likes, comments, and subscriptions for any news article showcase the popularity of the news. Popular news websites like “www.newindianexpress.com”, “www.hindustantimes.com”, and “www.timesofindia.indiatimes.com” release a lot of news articles every day. Each of these articles belongs to a different category like technology, entertainment, sports, business, etc., and are different every weekday. In this paper, the model and constraints to guess the acceptance of news articles published using Machine Learning (ML) techniques are studied. Dataset is downloaded from Mashable, which is a well-known news website. The dataset of popular hashtags is used from social media to counter the relevancy problem. The problem with the current methods is that we cannot guess the acceptance of news articles to be published in the very far future. To overcome this, we have used Logistic Regression, Random Forest, and Ada Boost algorithms which reduce features and enhance performance. Comparison of the algorithms is done for Accuracy, F1-Score, and Area Under Curve. We have compared the proposed work with existing work to analyze and validate it. The results show that random forests outperform the other algorithms in terms of prediction accuracy and generalizability. Top of Form