Add MWAA app that trains a generic model and then deploys it as a Lambda function

[robertlcx]

Description

Usage

A dataset config secret is required. The make run already adds one:

{
    "url": "https://gist.githubusercontent.com/netj/8836201/raw/6f9306ad21398ea43cba4f7d537619d0e07d5ae3/iris.csv",
    "name": "iris.data",
    "feature_columns": ["sepal.length", "sepal.width", "petal.length", "petal.width"],
    "target_column": "variety"
}

This can technically be set to anything as long as classifiers are being trained/deployed.

Considerations

1. Using the local executor in Airflow for ease of use. A follow-up more-complex DAG using solely SageMaker will be added.
2. The workflow can accept any dataset that has a classifying target column.
3. There's a fan-out on 3 different algorithms: LogisticRegression, KNeighborsClassifier, DecisionTreeClassifier. The model with the best accuracy wins.

Running it

make start
make run
make stop