Estimators that allows for the standardized implementation and testing of algorithms a common, lightweight interface for classifiers to follow. By using this interface, we can apply these tools to arbitrary classifiers, without needing to worry about how the algorithms work.

Estimators must have the following two important functions:

• fit(): This function performs the training of the algorithm - setting the values of internal parameters. The fit() takes two inputs, the training sample dataset and the corresponding classes for those samples.
• predict(): This the class of the testing samples that we provide as the only input. This function returns a NumPy array with the predictions of each input testing sample.

Most scikit-learn estimators use NumPy arrays or a related format for input and output. However this is by convention and not required to use the interface.

There are many estimators...

When taking measurements of real-world objects, we can often get features in different ranges. For instance, if we measure the qualities of an animal, we might have several features, as follows:

• Number of legs: This is between the range of 0-8 for most animals, while some have more! more! more!
• Weight: This is between the ranges of only a few micrograms, all the way to a blue whale with a weight of 190,000 kilograms!
• Number of hearts: This can be between zero to five, in the case of the earthworm.

For a mathematical-based algorithm to compare each of these features, the differences in the scale, range, and units can be difficult to interpret. If we used the above features in many algorithms, the weight would probably be the most influential feature due to only the larger numbers and not anything to do with the actual effectiveness of the feature.

One of the possible strategies normalizes the features...

As experiments grow, so does the complexity of the operations. We may split up our dataset, binarize features, perform feature-based scaling, perform sample-based scaling, and many more operations.

Keeping track of these operations can get quite confusing and can result in being unable to replicate the result. Problems include forgetting a step, incorrectly applying a transformation, or adding a transformation that wasn't needed.

Another issue is the order of the code. In the previous section, we created our X_transformed dataset and then created a new estimator for the cross validation.If we had multiple steps, we would need to track these changes to the dataset in code.

Pipelines are a construct that addresses these problems (and others, which we will see in the next chapter). Pipelines store the steps in your data mining workflow. They can take your raw data in, perform all the necessary transformations...

In this chapter, we used several of scikit-learn's methods for building a standard workflow to run and evaluate data mining models. We introduced the Nearest Neighbors algorithm, which is implemented in scikit-learn as an estimator. Using this class is quite easy; first, we call the fit function on our training data, and second, we use the predict function to predict the class of testing samples.

We then looked at pre-processing by fixing poor feature scaling. This was done using a Transformer object and the MinMaxScaler class. These functions also have a fit method and then a transform, which takes data of one form as an input and returns a transformed dataset as an output.

To investigate these transformations further, try swapping out the MinMaxScaler with some of the other mentioned transformers. Which is the most effective and why would this be the case?

Other transformers also exist in scikit-learn...