Before You Begin

This 10-minute tutorial shows you how to review related datasets produced by the predictive model to determine its accuracy and implement changes to improve results.

Background

When you train and create a predictive model, Oracle Analytics generates the related datasets. These related datasets contain details about the model such as prediction rules, accuracy metrics, confusion matrix, and key drivers for prediction depending on the algorithm type. You can examine the rules used by the predictive model to help tune the model to get better results. Use this information to iteratively adjust the model settings to improve accuracy and predict better results.

This is the second tutorial in Train and Apply Predictive Models in Oracle Analytics. Read the tutorials in the order listed.

What Do You Need?

  • Access to Oracle Analytics

    When using Oracle Analytics Desktop, you must install machine learning (DVML) to use Diagnostics Analytics (Explain), Machine Learning Studio, or advanced analytics.

  • Access to the elastic_train_df data flow
  • Access to the sample_donation_data dataset

Edit the Training Model Data Flow

  1. On the Home page, click Data.
  2. In the elastic_train_df data flow, click the Actions Actions menu icon, and then select Open.
  3. Click Add a step Add a step icon on the line before the Train Numeric Prediction step, and then click Select Columns. In the Selected list, click the following:
    • PROJECTID
    • TEACHER_ACCTID
    • SCH_LATITUDE
    • SCH_LONGITUDE
    • STUDENTS_REACHED
    • NUM_DONORS
    • FUNDING_STATUS
    • DATE_COMPLETED
  4. Click Remove Selected.
  5. Click the Save Model node in the data flow. In Name, enter elastic_model_2.


    Description of elastic_model_2_df.png follows
    Description of the illustration elastic_model_2_df.png
  6. Click Save. Click Run Data Flow Run Data Flow icon.
  7. Click Go back Back icon.

Review the Revised Machine Learning Model

  1. On the Home page, click Machine Learning. In the elastic_model_2, click the Actions, and then select Inspect.
  2. In elastic_model_2, click Quality.


    The residual value distribution in elastic_model_2 is slightly different those in elastic_model_1.

    Description of elastic_model2_quality.png follows
    Description of the illustration elastic_model2_quality.png

    The Coefficient of Determination value is 2% in elastic_model_2. In elastic_model_1, the Coefficient of Determination value is 16%. Your values are different due to the random selection of donation data when running the data flow.

    Description of elastic_model_1_residual.png follows
    Description of the illustration elastic_model_1_residual.png
  3. Click Close.

Compare Predictive Model Scenarios

  1. Click Home, click Workbooks and Reports, click donations_random_sample, click the Actions Actions menu icon, and then click Open.
  2. Click Edit Edit icon.
  3. In the Data panel, click Create Scenario. In Create Scenario - Select Model, click elastic_model_2, and the click OK.
  4. In the Data panel, expand elastic_model_2, select TOTAL_DONATIONS Prediction, and drag it in Values (Y-Axis) in the Grammar panel.

    The two models produce similar results even though they use different model variables.


    Description of total_donations_model2.png follows
    Description of the illustration total_donations_model2.png
  5. Click Save.

Next Steps

Apply a Predictive Model

Learn More