Supervised Machine Learning

Supervised machine learning refers to training algorithms to learn from labeled data sets. We provided labeled examples, and then optimize an algorithm against the ground truth labeled data sets.

Classification and Regression

Two examples of supervised machine learning tasks that we can solve are classification and regression.

Classification refers to the prediction of a category, or class, given a set of input features. The target will be a categorical variable. Classification problems can be binary (predicting one of two classes), or multi-class (predicting one of several mutually-exclusive classes).

Regression refers to the prediction of a real number, given a set of input features. The target will be a numerical variable.

Examples of categorical variables in environmental and climate science

A number of different problems in environmental science can be viewed as classification problems. For example, hurricanes and tornadoes are classified according to severity scale, and air quality can be classified in terms of how healthy or hazardous it is. We also cast problems where we are asking whether an event occurred or not as a binary classification problem (for example, was there a wildfire or not?)

In addition, many in situ and remote sensing applications require classifying observations into a variety of classes. For example, land usage or cloud type can be categorized using satellite images.