Let's assume you already have an idea for a machine learning project. How do make sure this is the righ project to work on?
We can illustrate this with an example. Let's say we want to build a board game recommendation system based on user ratings and board game descriptions or rules. We can evaluate whether this is a project worth exploring by answering a few questions.
Evaluate the feasibility and difficulty of the project
How feasible is this project? Has anyone worked on a similar project?
A quick Google search for "board game recommendation github" returns multiple related projects. Even if there were no projects specific to board games, googling "recommendation system github" returns lots of similar work in other areas.
This project seems very doable.
What kind of data do we need?
We need a dataset of user ratings of various board games and a dataset containing the descriptions and rules of those board games.
How difficult would it be to gather the dataset?
Given that we found many related projects, there are probably readily available datasets we can pull and get started. Let's assume relevant datasets were not available. In that case, we need to scrape this data from various resources. We can probably find websites that contain user ratings of board games and descriptions of those board games.
There is data available for this project out of the box. It's also feasible to build our own dataset if needed. So this is overall a feasible project. What's next?
Refine the problem statement
Ask yourself the following questions to come up with a clear problem statement:
What problem are you solving?
The problem of information overload when choosing a board game.
What does the final solution look like?
A web application that gathers user board game preferences manually or by logging user into a third-party website containing this information and returns game recommendations.
Who can benefit from the solution?
Board game stores can offer personalized recommendations to customers on their websites.
What are the benefits of using this solution?
Retailers and individual board game stores can send personalized offers to their customers and increase their sales. Board game cafes can use this tool to recommend board games to customers.
Kickstart the project
Once we have a clear problem statement, we can set up a Github repository for the project and initiate it with a README.md file containing the goal of the project and the motivation behind it as well as a plan of action.