Using scikit learn to build a simple logistic regression model from the Iris data set and persist the model with pickle. Pickle is a Python module that allows us to serialize a Python object structure. In this case, we will pickle the model we generated with scikit learn. We then use Flask as our api backend to serve our pickled model. Once our backend is up and running, we can make calls against the api on the /predict route to generate responses from our model.
Our predict endpoint will return a JSON object with the class probabilities of our input variables.
Use the accompanying Jupyter notebook to make requests against the endpoint.
Sample input:
(Adjust input values to get your own unique response against the model)
input_data = {"sepalLength": 5.2,
"sepalWidth": 4.5,
"petalLength": 1.4,
"petalWidth": 4.2}
Sample output:
[{'name': 'Iris-Setosa', 'value': 93.48},
{'name': 'Iris-Versicolour', 'value': 0.01},
{'name': 'Iris-Virginica', 'value': 6.5}]