Xarray¶

Setup¶

In [ ]:

pip install ydf xarray -U

pip install ydf xarray -U

In [1]:

import ydf
import xarray as xr

import ydf import xarray as xr

Xarrays¶

Xarray is a great way to train and use YDF models. YDF can consume Xarrays directly for all its operations.

Let's define a Xarray dataset:

In [2]:

dataset = xr.Dataset({
    "feature_1": ("example", [0.1, 0.2, 0.3, 0.4]),
    "feature_2": ("example", ["X", "X", "Y", "Y"]),
    "feature_3": (("example", "pixel"), [[1, 2], [3, 4], [5, 6], [7, 8]]),
    "label": ("example", ["A", "B", "A", "B"]),
})

dataset = xr.Dataset({ "feature_1": ("example", [0.1, 0.2, 0.3, 0.4]), "feature_2": ("example", ["X", "X", "Y", "Y"]), "feature_3": (("example", "pixel"), [[1, 2], [3, 4], [5, 6], [7, 8]]), "label": ("example", ["A", "B", "A", "B"]), })

This dataset contains 4 examples. feature1 is a single-dimensional numerical feature. feature_2 is a categorical feature. feature_3 is a multi-dimensional feature. Finally, label is the label.

We can then train a model on this (very simple) dataset.

In [4]:

model = ydf.RandomForestLearner(label="label").train(dataset)

model = ydf.RandomForestLearner(label="label").train(dataset)

Train model on 4 examples
Model trained in 0:00:00.006497

We can make predictions with the model.

In [5]:

model.predict(dataset)

model.predict(dataset)

Out[5]:

array([0.2999998, 0.2999998, 0.2999998, 0.2999998], dtype=float32)