What are good Machine Learning problems?

Thanks to the article How to Automate Tasks on GitHub With Machine Learning for Fun and Profit

by Hamel Husain

A good machine learning problem can be defined as:

  • There is an abundance of data, which is already labeled or where labels can be inferred.

  • The data can be used to solve real problems.

  • The problem relates to a domain you are passionate about or the problem you want to solve is your own and you can be your first customer.

  • There is a platform where your data product can reach a massive audience, with mechanisms to gather feedback and improve.

  • You can create this with minimal expense and time, hopefully using languages and tools you are familiar with.

  • There is a way to monetize your product if it becomes successful.