news_popularity2

Predict the popularity (number of shares on social media, on log-scale) of Mashable.com news articles based on the text of their title, as well as auxiliary numerical features like the number of words in the article, its average token length, and how many keywords are listed, etc. This dataset represents a very difficult prediction problem with only weak signal offered by the observed features. It is fundamentally hard to forecast how popular an article will be based only on its title and crude numerical summary statistics. To be comprehensive, an AutoML benchmark should contain at least one challenging problem like this. While pop stems from the same original data source as channel, the two have different labels to predict and do not share exactly the same set of features. Dataset found from the paper: Benchmarking multimodal automl for tabular data with text fields. arXiv preprint arXiv:2111.02705.