Machine Learning with Scikit-Learn - [Part 42]

In this tutorial we're diving into another section of machine learning, which is automatic feature selection.

One of the major reasons to use automatic feature selection with your datasets is to reduce the dimensionality of your data, which is very desired outcome in many projects - one possible reason being the lack of computational resources or time to train an algorithm or a neural network.

Some of the methods used of automatic feature selection include:

• univariate statistics
• model-based selection
• iterative selection

In this specific tutorial we're going to start with univariate statistics, and two methods than can be implemented in scikit-learn for this purpose are: SelectKBest and SelectPercentile.

Here we're going to use the SelectPercentile and we're going to work on the preloaded cancer dataset, that comes with scikit-learn. To the original dataset we're going to add additional noise features, after which we apply SelectPercentile.

As seen in the tutorial, this method reduces the features by about 50%.

Please watch the video below for a complete walk through:

---

To stay in touch with me, follow @cristi

---

Cristi Vlad Self-Experimenter and Author