Use Explain to Identify Anomalies Among Records In Your Dataset

In Oracle Analytics, right-click a target column in your dataset, select Explain, then select the Anomalies tab. Within seconds, you can see a list of anomalies, each represented as a single bar in the top bar chart.

Explain considers the attribute and date columns and identifies intersections of two or three attribute columns where the value of a measure is different to the logically expected value (regressions). The detailed results differ for measures and attributes.

If You're Explaining a Measure Column

The Anomalies tab analyzes combinations of attributes from the dimensions selected in Settings, and identifies intersections where measure values are different to the values expected by regression algorithms. A description next to the visualization summarizes the combination of attributes identified as anomalous. The visualization underneath shows the gap between actual value of the measure (bar) versus the regression expected value (flat lines). You can add the chart to your canvas and leverage it as is. You can also manually visualize the combination that was identified as an outlier. See this video.

If You're Explaining an Attribute Column

The Anomalies tab analyzes combinations of attributes from the dimensions selected in Settings and identifies intersections where the number of records is different to the number expected by regression algorithms. You can display a drop down hyperlink in the overhead text at the top of the tab showing which value of your selected attribute is displayed. Further down the page, the text above each visualization describes the combination of attributes identified as anomalous. The visualization shows the gap between the actual number of records for that intersection (bar), versus the regression expected number (flat lines). You can add the chart to your canvas and leverage it directly as is.