Introducing model evaluation
Model evaluation is the process of assessing the performance of one or more data science models to decide which is the best one to solve a given task. Model evaluation is an iterative task because we run it over and over again, until we reach a satisfactory model.
Model evaluation depends on the task we want to solve. In general, there are two types of tasks:
- Supervised learning – You train a model with some labeled data, you test the model on other labeled data, and then you try to predict the target value for unseen and unlabelled data. In this case, model evaluation is simple because, during the testing phase, you can compare the output produced by the model with the labeled testing data.
- Unsupervised learning – You do not have any labeled data, but you try to predict the output on the basis of some criteria, such as data similarity. In this case, model evaluation is quite complicated because you do not have any testing...