Given the plethora of available resources online, it can be disorienting trying to figure out which resources to focus on. The Kaggle’s state of data science and machine learning survey 2017 asked respondents about the learning methods that they found the most helpful. Here is a visualization of the responses, created by my previous colleague Jack Cook.
The most effective learning methods are doing projects, taking courses, and just spending a ton of time on StackOverFlow (SO). Kaggle competitions rank high on the list but since the respondents are Kaggle users, their answers are biased. A college education is perceived as slightly more useful than watching YouTube tutorials and reading blogs. The least useful methods in this survey are podcasts, newsletters, and conferences.
Attending conferences might not be useful for building your skill sets, but very useful for building up your network. Getting published at conferences is a great way to put your name out there and signal that your knowledge of the field is in-depth enough to come up with original research ideas.