This project aims to predict the popularity of an online news article before it is published. Such a tool will help publishers and editors in maximizing the popularity of their articles and sell advertisement.
- The dataset I have analysed belongs to Mashable Inc. — which is considered to be a one-stop shop for social media.
- It contains information of 40,000 articles published between 2013 and 2015.
- With the expansion of the Internet, more and more people enjoy reading and sharing online news articles.
- The number of shares under a news article indicates how popular the news is.
In this project, we intend to find the best model and set of features to predict the popularity of online news, using machine learning techniques.
Our data comes from Mashable, a well-known online news website. The performances are recorded and compared. Feature selection methods are used to improve performance and reduce features.
Random Forest turns out to be the best model for prediction, and it can achieve an accuracy of 70% with optimal parameters.
The data set can be summarized as follows:
- Data Set Characteristics: Multivariate
- Number of Instances:39797
- Attribute Characteristics: Integer, Real
- Number of Attributes: 61
Our work can help online news companies to predict news popularity before publication.