Before you Begin

In this tutorial, you create a dataset with multiple tables and select a column for Explain machine learning to analyze. When the dataset has more than one table, Explain reviews the table containing the selected column and the other tables in the dataset for their impact on the selected column.

Background

The Oracle Analytics Explain machine learning algorithm reviews the entire dataset for patterns and facts, and generates visualizations that you can use in your workbook. Using Auto Insights as the basis for the columns used by Explain, only the most interesting columns are reviewed in the data analyses. By only looking at the most interesting columns, Explain provides analyses without compromising performance.

When using multiple table or star schema datasets with Explain you need to identify the fact table by setting the preserve grain property.

You can change the columns used in Explain by adding or removing the default column selections in Settings. Your column selections in Explain persist during the Oracle Analytics session and aren't persisted when you close the workbook.

What Do You Need?

  • Access to Oracle Analytics or Oracle Analytics Desktop

    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 SH sample schema to perform the steps in this tutorial, see Installing Sample Schemas

Create a Dataset with Multiple Tables

In this section, you create a dataset from the SH schema. By default, the Auto Join tables option uses the relationships defined in the schema to create the table joins.

This example uses the SH schema from an Oracle Database connection.

  1. Sign in to Oracle Analytics.
  2. On the Home page, click Create, and then click Dataset.
  3. In Create Dataset, select a connection that supports datasets with multiple tables to use as the source.
  4. In the Connections panel, expand the SH schema, hold down the Ctrl key, and then select the following:
    • CHANNELS
    • COUNTRIES
    • CUSTOMERS
    • PRODUCTS
    • SALES
    • TIMES
  5. Drag the tables to the Join Diagram.


    Description of sample_sales_ds.png follows
    Description of the illustration sample_sales_ds.png
  6. Right-click SALES and select Preserve Grain.


    Description of sales_fact_table.png follows
    Description of the illustration sales_fact_table.png
  7. Click Save Save icon. In Save Dataset As, enter sample_sales in Name and click OK.

Create Visualization with Explain

In this section, you create a workbook, select a column for Explain to analyze, and then examine the visualizations.

  1. Click Create Workbook.
  2. Close the Auto Insights panel.
  3. In the Data panel, expand PRODUCTS, right-click PROD_SUBCATEGORY, and then select Explain PROD_SUBCATEGORY.


    Description of basic_facts.png follows
    Description of the illustration basic_facts.png
  4. In Basic Facts about PROD_SUBCATEGORY, scroll down to view the horizontal bar chart of PROD_SUBCATEGORY by AMOUNT_SOLD.


    Description of basic_fact_amt_sold.png follows
    Description of the illustration basic_fact_amt_sold.png
  5. Click Settings.
  6. Expand SALES and remove the checks from TIME_ID levels, CHANNEL_ID, and PROMO_ID. Select QUANTITY_SOLD.


    Description of basic_facts_settings.png follows
    Description of the illustration basic_facts_settings.png
  7. Click Apply.


    Explain generates a horizontal bar visualization of PROD_SUBCATEGORY by QUANTITY_SOLD. The basic facts donut visualization doesn't change.

    Description of qnty_sold_prodsubcat.png follows
    Description of the illustration qnty_sold_prodsubcat.png
  8. Hover your cursor over the upper right side of the PROD_SUBCATEGORY by QUANTITY_SOLD and the PROD_SUBCATEGORY by AMOUNT_SOLD visualizations, and then click Select for Canvas Select canvas icon. When the check mark selected for the canvas changes to green Green check mark, click Add Selected.


    Oracle Analytics adds the selected visualizations to the canvas and closes Explain.

    Description of bf_added_to_canvas.png follows
    Description of the illustration bf_added_to_canvas.png

Examine Key Drivers, Segments, and Anomalies

  1. In the Data pane, right-click PROD_SUBCATEGORY and select Explain PROD_SUBCATEGORY.
  2. In Explain, click Key Drivers of PROD_SUBCATEGORY.


    Explain examines all tables in the dataset that are related to the PROD_SUBCATEGORY column in the PRODUCTS table. The results include the interaction with FISCAL_MONTH_NAME and CALENDAR_MONTH_NAME from the TIMES table.

    Description of key_drivers.png follows
    Description of the illustration key_drivers.png
  3. Click Segments that Explain PROD_SUBCATEGORY.

    The Explain segment analysis shows the product subcategory values and the impact by CHANNEL_CLASS and CHANNEL_DESC in the CHANNELS table.

    Description of prod_subcat_segments.png follows
    Description of the illustration prod_subcat_segments.png
  4. Click Settings. In Settings, expand COUNTRIES, select COUNTRY_NAME, and then click Apply.
  5. Click Anomalies of PROD_SUBCATEGORY. Scroll down to When COUNTRY_NAME is Germany..., click Select for Canvas Select canvas icon.


    Description of germany.png follows
    Description of the illustration germany.png
  6. In Anomalies of PROD_SUBCATEGORY, next to the message "4 updates available," click Refresh View.


    Explain adds four visualizations to Anomalies.

  7. Hover and select one or two of the visualizations for the canvas. In each visualization, click Select for Canvas Select canvas icon, and then click Add Selected.


    Oracle Analytics add another canvas to your workbook with the visualizations selected from Explain Anomalies.

    Description of explain_anomalies.png follows
    Description of the illustration explain_anomalies.png

Learn More