An adaptable approach in using Machine learning towards predicting the Popularity of news articles.

Abstract

The vast and increased utilization of the web and the arrival of the data technology field has led to a new age where individuals are starting to read online news frequently. Hence, online news has become the fundamental source of information for most individuals, and anticipating the prominence of online news has become a discussed issue that can't be neglected. It could help in assisting  authors with introducing serious and highly readable news. Number of shares an article gets is considered as one of the most obvious factors in determining its popularity. In this paper, we apply distinctive machine learning methods to anticipate the quantity of shares and categorize them as well known and unpopular. The information has been assembled from the UCI machine learning repository from Mashable. Linear regression and classification techniques like decision tree, SVM, and logistic regression are utilized to the data set. The performance of these methods is measured by their accuracy, precision, and recall measures. The aim is to find a reliable model with a prediction accuracy of 70 %. Having found a reliable prediction model, this work can then be used by online news agencies to anticipate their popularity based on content and to make changes accordingly and help news agencies in adopting promising advertising strategies.