Machine learning solution workflow

In general, the main tasks involved in solving a machine learning problem can be summarized into four areas, as follows:

• Data preparation
• Training sets generation
• Model training, evaluation, and selection
• Deployment and monitoring

Starting from data sources and ending with the final machine learning system, a machine learning solution basically follows the paradigm shown here:

Figure 11.1: The life cycle of a machine learning solution

In the following sections, we will be learning about the typical tasks, common challenges, and best practices for each of these four stages.