Predicting the Popularity of TED Talks

In this paper, we explore how to predict a TED talk’s popularity by its inherent features via machine learning techniques. we quantify a TED talk’s popularity as logarithmic transformation of its daily views and daily comments and include 43 features as predictors. We find that the ordinary least squares regression, ridge regression, and LASSO regression models perform well, and predictors such as a talk’s number of language translations, average Internet development environment when published, duration, main speaker’s occupation, as well as the timing it being uploaded have essential effects on its popularity. In the end, we also provide our suggestion on how to improve TED talks’ popularity within and beyond the scope of machine learning. All the materials in this research (including but not limited to original and processed data, figures, tables, etc.) are under the Creative Commons License: BY-NC-ND 3.0.Master of Science in Information Scienc