This is a framework to predict YouTube Video Popularity with Machine Learning Models. This is also my final project at my UCLA data science certificate program, which I completed with distinction.
With the new technology and platform YouTube innovates with, it becomes interesting and important to study what makes a YouTube Video trending. How we can predict the popularity of a video? Analyzing this question can help content creators better target at viewers’ interest and generate better popularity for the creator, and in the meantime better revenue from the video.
To rephrase the question using data science language, I am trying to build up a model to predict YouTube popularity based on different features of video, such as category, number of subscribers, content title, days of trending etc. Popularity is measured by number of “views”.
First, I use a comprehensive YouTube viewership dataset from YouTube open API to explore features of trending YouTube videos. With some identified features, I am then trying to fit data into machine learning models to predict number of views. A linear model with Stochastic Gradient Descent (SGD Regressor), a decision tree model, a random forest model and a neural network model are explored. Model with “best performance” will be chosen as the final narrative for the study.