Starting from business requirements
A typical ML process starts by defining business requirements. Follow the following steps to define the business requirements of the problem:
- Clearly define the business outcome that your ML solution is supposed to achieve, among all the stakeholders. For example, for a prediction ML problem, we need to define a range of accuracy that is acceptable by the business and agreed upon by all the stakeholders.
- Clearly define the data source of the ML problem. All ML projects are based on loads of data. You need to clearly define what the reliable data sources are, including training data, evaluation data, testing data, and a feed of regularly updated data.
- Clearly define the frequency of ML model updating (since data distributions drift over time), and the strategies for maintaining production during the model updating times.
- Clearly define the financial indications of the ML product or project. Understand any limitations such as resource...
