When calling a /predict or /learn endpoint, you must provide features. Features provide information that your model can learn from and can use to make intelligent decisions.

For example, a model that predicts if it is likely to rain today might be provided with the current barometric pressure, humidity, temperature, cloudiness, and other features. Features are represented as a map data structure where a key is the name of the feature and the value is either a string or a number .

"skies": "cloudy",
"pressure": 29.87,
"visibility": "low",
"humidity": 0.76,
"temperature": 74

Feature Construction

Create features that your model can effectively learn from to produce better results.

Assigning Numerical and Categorical Features

Features are implicitly decided to be numerical or categorical. To provide a categorical feature in your map, assign the value to have at least one non-numerical character. A numerical feature can be a string or a number, but cannot contain any non-numerical characters.

"humidity": 0.76 //numerical
"pressure": "29.87" //numerical
"temperature": "74F" //categorical
"skies": "cloudy" //categorical