Chapter 7 ML project process
Many ML projects get started the wrong way, trying a way to use data rather than using the data to fulfill a need, a need which has a benefit to the organization It is understandable that organizations want to learn from the data they have, but starting without a clear need in mind often leads to wasted efforts because sooner or later it will be discovered that the data available is not sufficient for a useful model.
At the start of a ML project there should be a clear formulated need which should be answered by the model, because ML is only a tool to help to achieve the objectives of the organization
At the beginning there is a need which ML is suitable to fulfill:
- Optimize fertilizer usage
- Improve user experience
- Reduce energy cost
- Increase milk production
Starting with the need the process can be split up in phases as shown below:
The process is not sequential but highly iterative as is described in the next chapters