Before you Begin

This tutorial shows you how to use Oracle Autonomous Data Warehouse AutoML in an Oracle Analytics data flow to build a prediction model without a data science background.

Background

As a business analyst, citizen data scientist, or expert data scientist, you can build and deploy machine learning models in Oracle Analytics using datasets created from Oracle Autonomous Data Warehouse. You don't need a deep understanding of specific algorithms and the modeling process to build and refine usable models.

The AutoML data flow step chooses the appropriate algorithm from the supported algorithms and identifies the features and tuning to use with your data. The selected target value determines whether the Task Type is a classification or regression type. In this example, AutoML chooses Classification for the Task Type because ATTRITION is a Text data type. The Model Ranking Metric indicates to the AutoML process which metric to use in assessing the model quality.

After generating the model from the data flow, you can use it to score your dataset using the Apply Model step in an Oracle Analytics data flow.

What Do You Need?

  • Access to Oracle Analytics
  • Access to an Oracle Autonomous Data Warehouse dataset with a user that has OML Developer and DWROLE database roles

    The Oracle Autonomous Data Warehouse user creating a dataflow using AutoML can't have administrative privileges.

Create a Dataflow

In this section, you create a dataflow using an Oracle Autonomous Data Warehouse dataset, and add the dataflow steps to complete and run the dataflow generating a prediction model.

  1. Sign in to Oracle Analytics.
  2. On the Home page, click Create, and then click Dataflow.
  3. In Add Data, select an Oracle Autonomous Data Warehouse dataset, and click Add.


    This example uses an employee attrition dataset for the AutoML example.

    Description of emp_att_dataset.png follows
    Description of the illustration emp_att_dataset.png
  4. In Data Flow Steps, double-click AutoML. In AutoML, click Select a column next to Target, and then select the column to use from the Available Data list.


    In this example, ATTRITION is the target column. AutoML chooses the best algorithm for the Task Type and Model Ranking Metric for your data. The Task Type is Classification and the Model Ranking Metric is Accuracy.

    Description of automl.png follows
    Description of the illustration automl.png
  5. In the dataflow, click the Save Model node. In Save Model, enter a name in Model name.

    In this example, the model name is employee_attrition_model.

  6. Click Save Save icon. In Save Data Flow As, enter a name, and then click OK.

    In this example, the dataflow name is emp_attr_df.

  7. Click Run Run icon to generate the prediction model.
  8. Click Go back Go back icon.

Review the Prediction Model

In this section, you review Inspect to find details about the generated model and the model's quality.

  1. On the Home page, click Machine Learning.


    Description of ml_model.png follows
    Description of the illustration ml_model.png
  2. Hover over your model generated by AutoML, click Actions Actions menu icon, and then select Inspect.


    Description of model_info.png follows
    Description of the illustration model_info.png
  3. Click Quality to view the model accuracy and precision information.


    Description of model_quality.png follows
    Description of the illustration model_quality.png
  4. Click Details to view the input columns, output columns, and parameters used in generating the model.


    Description of model_parameters.png follows
    Description of the illustration model_parameters.png
  5. Click Related to view the generated datasets.


    You can create visualizations with the generated datasets.

    Description of model_gen_data.png follows
    Description of the illustration model_gen_data.png
  6. Click More to see the generated metadata.


    Description of model_more.png follows
    Description of the illustration model_more.png
  7. Click Close.

Learn More