A Logistic-Regression-Based Reduced Model for Predicting Online News Popularity

Consuming news through social media has become an integral part of our lives, thus online news popularity prediction is now a crucial issue for marketing strategies and social media platforms. This paper proposes a real-time prediction model, based on a dataset from Mashable, it uses regression analysis to find out whether an article is of high-popularity or low-popularity. In this paper, we first solve the problem of imbalanced data via Synthetic Minority Oversampling Technique (SMOTE); then various feature selection and engineering methods are employed to extract key features. A comparison experiment is conducted to verify the validity of the data enhancement. And the results show that the reduced logistic model with SMOTE and dimension reduction outperforms other models with an Accuracy of 80.6% and a Recall of 78.6%. Our work can help the software application on evaluating the popularity of online news.