Feature Engineering⚙️🔧

When the ultimate aim of your Machine Learning project is to get the best possible prediction results, Feature Engineering plays a crucial role. Feeding the best features possible to the model will not only help getting better predictions, but also make the model simpler.:bulb:

Feature Engineering includes getting the importance of each feature. The features that influence the target the most are selected and rest are dropped. In other words, features with maximum information are retained.:repeat_one:

It also includes building new features by combining, splitting or transforming multiple features. For numerical data, linear combinations, polynomial features, binning, transformations like Box-Cox are carried out to get the most out of data. And it totally depends on the data and the domain as to which operations are required.:thinking:

Feature extraction is another method to create new features by extracting the relevant information from all the features.

For example, PCA, constructs new orthogonal features which contain the maximum information out of all the features. This also results in dimensionality reduction, but as a downside, interpretability is lost as you no longer can define what the features are showing.:bulb:

#machinelearning #datascience