Operators
Introduction
In Einblick, operators are used to transform, analyze, or visualize incoming data. Each operator represents a distinct step in your analysis. They appear as rectangles on the space you're working on. Much of the power of Einblick lies in the ability to customize and combine operators as you see fit. A few example operators are shown in the image below.
Einblick comes with a set of built-in operators that expose various kinds of functionality. In the image above, the filter and table operators are built-in operators and are always available. In the Available Operators section below, you can find a list of all currently available operators.
You can also create and share your own custom operators, called user-defined operators (UDOs). The 2D box plot operator in the image above is an example of a UDO, with a specification defined within Einblick. To learn more about UDOs, visit the User-Defined Operators page.
Adding Operators to Canvases
The primary way to add operators to canvases is by dragging them from the operators menu at the left of the canvas, as seen in the image below. You can place them anywhere on the canvas.
A few built-in operators can be added through shortcuts. For example, dragging a dataframe onto the canvas produces a new table operator with that dataframe as an input.
Operator Inputs and Outputs
Most operators take some sort of input data to act upon. In general, the input data to an operator is represented by an input dataframe at the top-left of the operator, while it's output appears as a dataframe at its bottom-right. The input data to an operator can also be filtered by attaching a filter link to one of the operator's input dataframes. In the image below, the output dataframe of the top-left operator is used as the input to the bottom operator, and the filter output of the top-right operator is used as a filter input of the bottom dataframe. To learn more about connecting operators in this manner, visit the Data Links and Filter Links pages further on in this section.
Operator Input Menus
Most operators have customizable options that determine their behavior. Depending on the operator, specifying some of these options may be necessary before running it. These options can always be found on the left edge of the operator.
The image below shows the input menus of the table operator, circled in red. The attributes menu, which allows you to specify which attributes (i.e. columns) to use in the table, is expanded.
Running Operators
Once the adequate settings in the input menus have been set, you can run an operator by clicking the blue play button at the bottom left corner of the operator. If the inputs have not been set appropriately, the play button will be grayed out and disabled. In the image below, the left table operator has a disabled play button and cannot be run, while the right table operator can be re-run at any time.
By default, certain operators run as soon as the appropriate inputs have been set, and whenever a change has been detected in its inputs. However, you can always run (or re-run) an operator using the play button.
If an operator is currently running, the play button will be replaced by a stop button (see below image.) Clicking this button will stop the currently running operator. You can then click the play button again at any point to run the operator again.
Miscellaneous Operator Menus
At the bottom-left corner of an operator, you will see several icons that expose additional functionality. In the previous section, we covered the first of these: the play button. The others are described below.
Operator Options
Clicking the triple-dots icon will display a series of options available for the operator. The specific set of available options will depend on several factors, but some common operations include duplicate, delete, and add to dashboard.
Sample Information
Clicking the sample information icon will show information about the input data of the operator. For example, in the below image, we can see that all 595 rows of the input were sampled, and that the input data was sampled random randomly. For large datasets, this information will update as the sampled data changes over time.
Help
In the help menu, you will see a short description of the operator (if provided), along with the option to show a wizard (if applicable; see next section) and accompanying documentation (if applicable).
Wizards
Many of the core operators come with wizards, which guide you through the process of setting up the operator and using it effectively. For example, the automl wizard helps you decide, among other things, which scoring metric to use in your AutoML task. Toggle the wizard by clicking the show/hide wizard option.
A step of the automl wizard is shown below.
Lock Operator
To keep an operator in place on the canvas, preventing any movement or resizing, you can "lock" the operator by clicking the lock button. Click the icon again to "unlock" it.
Available Operators
Below is a list of all available operators currently in Einblick.
Data Manipulation
| Operator | Description |
|---|---|
| Filter | Filter a dataframe with an arbitrary number of conditions |
| Join | Join two dataframe on one or multiple keys |
| Python | Write and execute your own python snippets |
| Expression | Manipulate your dataframe (create/modify columns, hot-encode) |
| Transpose | Transpose a dataframe of event-style data |
Descriptive Analytics
| Operator | Description |
|---|---|
| Correlation | Compute the correlation between Numerical and/or Ordinal Attributes |
| KDA (Key Driver Analysis) | Automatically find interesting differences between two populations |
| Chart | Graphical Pivot Table |
| Profiler | Summarize the statistics of a data frame |
| Statistical Test | Run T-test or Chi-squared test depending on the data input type |
| Table | Spreadsheet visualization of your data frame with support for filtering and aggregation |
Predictive & Prescriptive Analytics
| Operator | Description |
|---|---|
| Forecast | Build univariate and/or multivariate forecast model on time-series data |
| Mutual Information | Model agnostic feature importance |
| AutoML | Build end-2-end regression or classification predictive models |
| Text Featurizer | Extract embedding from text columns |
| What-if | Simulate scenarios and optimize with respect to an objective function |
User-Defined Operators
See User-Defined Operators (UDOs) for more information.