3.1 Manage Features
On the feature platform home page, you can click the VIEW link shown in the picture below (in the FEATURES section and under AVAILABLE FEATURES) to view all feature.
On the Feature Platform - All Features page, all features are listed as shown below.
To customize information you see on the screen, you can select the Edit Columns button in the upper right corner of the table to choose which column(s) need to be displayed. Click ALL to choose all fields to be displayed, click DEFAULT to go back to the default selection set by Datavisor.
For any features in the table, you can click the vertical ellipsis icon on the rightmost column, and choose any following action to take: Test, Edit, Copy, See Dependency, or Delete.
To Test the selected feature, please refer to Section 4.2: Test Features.
For any features in the table, you can click the vertical ellipsis icon on the rightmost column and choose EDIT to make any necessary modifications needed to the selected feature on the Edit Current Feature page. As shown below, the layout of the Edit Current Feature page is almost identical to the Create Feature page.ou can edit the same fields that were used to create the feature.
If the feature is created by using coding or you want to edit the source code of the feature, you can click the USE CODING button to Edit Current Feature as well.
You can Copy a feature so that you can make modifications to the feature without deleting the original version. From the vertical ellipsis dropdown menu, click Copy for the selected feature. A small window will pop up explaining that the newly copied feature must be renamed (no two features can have the same name).
Click EDIT to proceed to the Copy Feature page as shown below.
The system automatically adds a “_copy” extension to the copied feature name, but you are free to change the name as you wish. if you change the new copy feature’s name to a name that already exists, you will not be able to create the copied feature. An error will be displayed as shown below.
3.1.2 See Dependency
From the vertical ellipsis drop-down menu, you can click See Dependency to view the selected feature’s dependencies. A hierarchical graph will pop up on the right side of the screen containing the “parents” - Event attributes, templates, or features that the selected feature is derived from, and the “children” - features derived from the selected feature. The example below displays the dependencies of the feature “transaction_count_per_device_id_last_3_30_day_0”.
The Feature Dependency pop up window displays the selected feature’s details on the top and displays its dependencies below. In this example, the selected feature “transaction_count_per_device_id_last_3_30_day_0” is derived from the DataVisor event attributes (or DV provided fields “device_id” and “transaction_id”. And the selected feature derives the feature “is_new_device_id_3_days_0”. By hovering your mouse over any of the features/event attributes displayed in the graph, a small window will pop up with details as displayed below for feature “is_new_device_id_3_days_0”.
You can also click the Table option in the Feature Dependency window to view the feature dependencies in a table format as displayed below.
From the vertical ellipsis dropdown menu, you can click Delete to delete the selected feature. When you delete a feature, a pop up window(shown below will appear to ask if you want to delete the feature. Click CONFIRM to delete the feature or CANCEL to cancel the action.
In backend, the deletion function will check the dependency of the selected feature. If any other features or rules depends on this selected feature, the deletion can not be done.
3.1.4 View Source Code
You canview the source code of any feature by to the main Feature Platform - All Features page. When you click the blue underlined feature’s name under the Name column, a window containing the feature details will pop up on the right side of the screen as shown below.
Also in the same FEATURE DETAILS pop-up window, you have the option to TEST or EDIT the selected feature. Click OK to return to the Feature Platform - All Features page.
3.2 Test Features
You can test a feature by clicking on the vertical ellipsis icon on the rightmost side of the table. Then the system will direct you to the Test Feature page. Then the first step is to select an appropriate dataset to be used for testing. We will further explain with one example after the Note.
Note: In the SELECT THE DATA SOURCE field, only the validated datasets and the dataset that contains event attributes that are used in the selected feature will show up.
You can check which datasets are validated by going to the Feature Platform Dashboard and clicking VIEW under Datasets in the Features section. You will be directed to the Data Management page. Click Dataset in Validation and check the Status of each dataset listed as shown below. Datasets that are Validated With Warnings are still considered to be validated datasets.
Let’s test the “EMAIL_PREFIX_EXAMPLE” feature that we created earlier.
Find “EMAIL_PREFIX_EXAMPLE” in the Feature Platform - All Features table, and click on the vertical ellipsis icon on the rightmost side of the table to Test.
Upon entering the Test Feature page for this feature, we must first SELECT THE DATA SOURCE and find a dataset to test “EMAIL_PREFIX_EXAMPLE”. As displayed below, each of the datasets that are displayed from the drop down menu is validated and contains the event attributes used in the selected feature. In this example, because “EMAIL_PREFIX_EXAMPLE” uses the “email” event attribute to parse the email address, each of the datasets listed also contain an “email” field.
You can select any of the datasets that appear in the dropdown menu. As displayed below, some datasets may contain multiple data files that span several days; however, you may only choose just one file from one of the dates to test your feature. Next, you can Select the number of recrds to test. You can test from the Top 500 records up to the Top 10,000 records then the system will pull the first selected number of records from the selected file to test. DataVisor typically recommends choosing the Top 500 records option to run a faster test. You also have the choice of which Replay Mode to use; for a smaller number of records to test (i.e. Top 500 records), DataVisor typically recommends running in “local” mode. When testing a larger number of records, you may choose either “distribute_internal” or “distribute_external” mode.
After you click on TEST, a spinning wheel will appear. Once the test is done,the Feature Test Result table will be displayed below. The seelected feature to test will be shown in the first column with the Feature Name you specified earlier. The data will be computed row-by-row, The email prefixes of the top 500 records were extracted from the email raw feature. You can sort, search, and customize what columns should be shown on the screen. Finally, if you want to save your results, choose Export In Original Format on the top banner to export the result.
3.3 Publish Features
When you are confident that your feature is working as expected, you can Publish the feature. There are two important reasons to publish features:
- You can only use a feature in modeling if it has been published. While you can test a feature in “Draft” mode, you cannot use it for modeling.
- Publishing a feature ensures that it cannot be modified again.
In order to publish a feature, perform the following:
- Navigate to Feature Platform - All Features
- Under the Status column, select your feature.
- Select the status “Published” instead of “Draft” in the Status drop down menu.
- Click CONFIRM to proceed and publish your feature; otherwise, click CANCEL to cancel.
To quickly publish multiple features, you can click the checkboxes of the corresponding features you want to publish on the leftmost side of the table. Once you’ve checked all features to publish, you can click the PUBLISH button in the top left corner of the table. Click CONFIRM to proceed and publish all selected feature; otherwise, click CANCEL to cancel.
3.4 Export Feature Configuration
Click the EXPORT button in the Feature Configuration window to begin. Any feature that is shown in the Feature Platform - Select Features To Export table can be exported as a JSON formatted file. This file can later to used to mport into a new environment as an alternative way to create features. Click the checkboxes for any features that you would like to export, and click the EXPORT button in the upper left area above the table. You may export one or more feature configurations at a time.
The JSON formatted file that contains the feature configurations necessary to recreate that same feature in a new environment. The file contains the following keys:
- “EVENT_ATTRIBUTE_INFO”: High-level fields such as name and return type for this feature.
- “FEATURE”: Feature-level metadata regarding information about the script that generated the feature, the parameter names, etc....