What are the key criteria to consider when evaluating datasets?

In this section, we will understand what the key criteria are when it comes to evaluating datasets.

Data quantity

Is there sufficient data to train an accurate model or to make inferences about a wider population if you’re working with a data sample? As mentioned in the previous chapter, in statistics, you must often work with a limited sample of data, and the ability of that sample to represent the wider population often depends on the size of the sample. Within machine learning, models trained on larger datasets perform much better than those trained on a small sample. There are more advanced techniques, such as data augmentation and transfer learning, that can help in this situation and will be covered later, but an initial consideration is whether there is enough data available to meet business requirements around accuracy.

Consider, for instance, a customer churn model designed to predict which customers...