Skip to main content
Version: 0.21.2

Detect

Detect lets you leverage DataChat's powerful machine learning capabilities to analyze your data. With Detect, you can identify:

  • Cyclical trends in a column using a datetime column as a reference.
  • Outliers in a numerical column.
  • Outliers in a numerical dataset.

Note that Detect works only on numerical columns. If any categorical or string columns are detected, you can choose to either keep them in the analysis or remove them. Removing them might make the analysis quicker.

Format

  • Detect outliers in the columns <columns> (using the method <method>)
  • Detect outliers in the dataset <dataset> (using the method <method>)
  • Detect any cyclical trends in the column <column> using the column <datetime column>

Parameters

Detect uses the following parameters:

  • column (required). The column in which to detect cyclical trends.
  • datetime column (required). The datetime column to use as a reference when detecting cyclical trends in the columns above.
  • columns (required). A comma-separated list of numeric columns to analyze for outliers.
  • dataset (required). The dataset to analyze for outliers.
  • method (optional). The method to use when detecting outliers. By default, the isolation forest method is used.

Output

If any outliers are detected, a success message appears in the chat history and a table appears in the display panel and becomes [dataset]_Outliers. The table includes the values that were identified as outliers, their score, and their ranking.

If any cyclical trends are detected, a success message appears in the chat history and a bar chart appears in the display panel. The chart illustrates the top five periods with a positive correlation and the confidence of each interval.

Examples

To detect cyclical trends in a column called "Temperature" using a column called "Date" as a reference, enter Detect any cyclical trends in the column Temperature using the column Date.

To detect outliers in a column called "Age," enter Detect outliers in the columns Age.

To detect outliers in the columns called "Age" and "TicketPrice," enter Detect outliers in the columns Age, TicketPrice.