General Process for new project

Often when i get a new project in machine learning the client always ask me either to do a particular task like a prediction for one thing or give me data and ask me to find what i can do with it.

i've read the book hands-on machine learning with scikit-learn tensorflow where you can see a full process for starting a project , basically drawing plots and search in correlation matrices what is interesting.

Do you guys have a special way to do this ? what is the step by step process you use ?

Thank you.

Topic project-planning machine-learning

Category Data Science


One of the standardized process for doing data analysis / data mining is the CRISP DM (Cross-industry standard process for data mining) which is one of the most adopted norms in the industry.

The process is very straightforward and starts by an important task which is the Business understanding of the issue or data you are dealing with.

This means that a prior research should be done if the analyst is not very familiar with the industry or the problem.

What makes this standard very popular is the fact that you can iterate various times between the step of formulating the problem ( writing in simple terms), data exploration and data visualization.

Here is a a very rich document that you can follow.

It would take a bit of time to go through all of the steps, however what could be quit interesting is the practice part, where you adapt your own client's problem into a simple question and start your exploration phase right away.

CRISP DM steps

About

Geeks Mental is a community that publishes articles and tutorials about Web, Android, Data Science, new techniques and Linux security.