Chapter 5 - Entity Management

5.1 Package

DataVisor combines fraud templates for common fraud scenarios with fraud domain expertise to provide Feature Packages. These are subsets of our feature template library that are tailored towards individual fraud use cases.

DataVisor Packages are pre-bundled sets of feature templates for specific use cases like Account Takeover, Mass Registration, Application Fraud, etc... When you use one of our packages, mapping your dataset’s unique fields to our template fields will allow you to immediately get many related features, which will be helpful in the modeling process. We also provide the option to create your own feature packages, which allow you to select certain feature templates that can be combined to solve the fraud use case.

DataVisor offers nine packages for various industry applications:

  1. Transaction Fraud
  2. Application Fraud
  3. Content Abuse
  4. General
  5. Fake Accounts
  6. Global Intelligence Network (GIN)
  7. Promotion Abuse
  8. Anti-Money Laundering
  9. Account Take-Over (ATO)

The workflow for using a DataVisor Feature Package is as follows:

  1. Select the package that corresponds to the fraud scenario (Mass Registration, Account Takeover, etc.)
  2. Map as many fields as possible from your dataset to the DataVisor standard field templates
  3. Use these features in your modeling process

On the Feature Platform dashboard, you can begin using DV packages by clicking on the packages displayed under the Start with a package section of the Dashboard.



We have illustrated how to conveniently create features from a package in Section 3.1: Create Feature from Scratch. In this section, we will dive into how to view, create, and manage packages.


5.1.1 Create Package

Packages are composed of feature templates. To create a package, you need to access the list of feature templates (please refer to Section 6.2: Feature Template) and click CREATE NEW PACKAGE. Select which feature templates you would like in the package. Next, you can input a name and a description for the new package you are creating. The following screenshot displays how the DV_Test package is created.



5.1.2 Manage Package

Click MANAGE PACKAGES from Start with a Package. Then the manage packages screen will appear as shown below. Click on the ellipsis icon to View, Edit, Delete, or EXPORT the selected package. View Package

To access the View Package page, click View from the ellipsis icon drop down menu as illustrated above or click the package’s name under the Name column of the Packages table. In the View Package page, you can view which feature templates are a part of the package and hover over the columns for more information. Click OK to return to the Packages page. Edit Package

While DataVisor provided feature packages cannot be edited, you may Edit any of the packages created by you or others. You can edit a package by selecting or deselecting the feature templates included in the original package. While you cannot change the name of the package, you can edit the Description. Each time you edit a package or change its description, a new version of the package will be created.

Click UPDATE PACKAGE at the bottom of the page to save the edits to your package.



Once you UPDATE PACKAGE, you will be redirected to the Packages page. You can click the triangle icon next to the edited package to Select Versions of the package. As shown below, all previous versions of the package will be displayed in the Select Versions pop-up window. Click on the ellipsis icon to Export, View, Edit, or Delete any of the selected versions of the package.



Please contact DataVisor customer support for the complete list of DataVisor packages. Delete Package

While DataVisor provided feature packages cannot be deleted, you may delete the packages that have been created by you or others. Click the ellipsis button on the rightmost side of the packages table and click Delete. A confirmation window will pop up to ensure that you want to delete the package. Click CONFIRM to permanently delete the package, or click CANCEL to cancel. Export Package

You can export your packages as a .json file. Click the ellipsis icon on the right side of the table and click Export to download the package to your local drive. For packages with multiple versions, you may select which version you want to export as shown below.



5.2 Feature Template

Advanced users are well aware of the challenges of different schemas for different datasets. For such users, there is a great need to generalize the feature generation so that any dataset can generate a feature using a template, rather than modifying the fields manually each time. In DataVisor’s Feature Platform, Feature Templates created by operators solve this need.

For example, you can create a generic feature template called Timestamp that takes any time stamp input, no matter if your dataset has a field named time stamp, time_stamp, or time.  After mapping, this Timestamp always returns the same feature day_of_week. You can view and create feature templates from the dashboard, using the bottom left panel below:



5.2.1 Create New Feature Templates

Using the UI to create new templates is quite easy using DataVisor’s Feature Platform, and follows the same steps as creating a single feature. If you want to create a non-velocity feature template, do the following:

  1. Select function categories (optional)
  2. Select function or create new function
  3. Select template parameters
  4. Select template name and package
  5. Create template

In the following example, EMAIL_IS_BAD_PROVIDER is created to identify whether the email provider is bad or not, and added to the transaction fraud package.



Similar to creating a feature, if you want to create a template for a velocity feature, follow these steps:

  1. Select Velocity Function categories (optional)
  2. Select Operator Function or Create New Function
  3. Select an existing aggregator or create a new aggregator
  4. Select time period for data aggregation
  5. Create Template

In the following screenshot, a Distinct_User_IP template is created to find the distinct USER_ID grouped by IP for the last 10 days and we place it into the General package.



5.2.2 Manage Feature Templates

Selecting View All displays the list of all current feature templates, along with package information, a short description about the template, and template metadata like creator and create time. You can customize columns shown on screen, and preview, edit or delete your own feature template as you wish. But for DataVisor built-in feature template, you only can view or delete. Screenshot is shown below.
 View Feature Templates



Click on VIEW ALL in the feature templates box, all feature templates are listed to view. Click the ellipsis icon on the rightmost side of the table and click Preview to view each individual feature template.



One example is shown below. Edit Feature Templates

Click on VIEW ALL in the feature templates box, all feature templates are listed to view. Click the ellipsis icon on the rightmost side of the table and click Edit to change each individual feature template.



Upon update completion, click on UPDATE FEATURE TEMPLATE save the changes, click on DELETE to delete the feature template, or CANCEL to go back.

Note: Users can not edit Datavisor provided feature templated. They are VIEW only. Delete Feature Templates



Click on VIEW ALL in the feature templates box, all feature templates are listed to view. Select one or more feature templates, click on the DELETE button on the upper left, and the user is able to delete selected feature templates after confirmation.




Note: Users can not delete any Datavisor provided feature template. Export Feature Templates



Click on VIEW ALL in the feature templates box, all feature templates are listed to view. Select one or more feature templates, click on EXPORT button on the upper left, and the user is able to export selected feature templates after confirmation.



The download is in JSON format. Edit, Delete or Export Different Versions of Feature Template

If a feature template has multiple versions, a  icon will show on the line. By clicking on it, the ‘select versions’ window will show up. Similar to edit, delete or export feature templates, users can select a specific version and perform the operation.





5.3 Functions

We defined a function in Feature Platform as a pre-defined logic or transformation that can be applied to data fields to generate a new result. We also introduced different categories of functions and learned how to use different functions to generate custom features. In this session, we will discuss in detail how to create and explore functions in Feature Platform.  


5.3.1 Create New Functions

Functions can currently be written in Java, and both velocity and non-velocity functions are supported.

When you select Create Function, you’ll be taken to a screen that will enable you to create any type of function. After selecting Velocity function or Non-Velocity Function, you can write a function under Function Script. Creating a Non-aggregation Function

Your intended function may branch off the logic of an existing function. In order to support this, once you select Non-aggregation Function, you will reach the following page.

You can start with the Load Code of Existing Function option if you want to make edits to an existing one and its source code will populate the console. Otherwise you can start programming your function from scratch.

Similar to creating a custom feature by coding, you can assume that the coding console does not require a method signature or a class declaration, although it does require a return type.

Setting Parameters

Below the coding console is the section where parameters should be set in order to use them in your function. You will see the following section:



Here, you should give each parameter a name, a data type and a description. We currently support the following data types for parameters:

  1. String
  2. Long
  3. Integer
  4. Short
  5. Double
  6. Boolean
  7. Float
  8. Byte
  9. Character

You may add as many parameters as needed, but please note that a parameter needs to be included in Function Parameters in order for it to be referenced in the function.

Referencing Variable Names

To write a function, you will likely want to reference variables (that would typically be passed into the function). In order to do this, refer to them in your code with the following string:

Parameter_Name + “#” + Parameter_Number (0-indexed)

For example, if you have a parameter called str and it’s the first parameter in your parameter list, refer to it in the code as str#0.

An example of this usage is shown below, for a simple function that adds two numbers and returns a double for the sum:


Once you have written the function and selected your parameters, you can click on Test to see if there are any errors. If not, you can give a description and return type for the function, then create the function. Creating an aggregation Function

When creating an aggregation function, there is no need to create a parameter list. Let’s say that you want to implement the distinct count function (i.e., return the number of distinct values for some attribute). Since you don’t know what specific attribute this function will be used on currently, we need to keep this generic. An example is shown below:


Line 4 demonstrates how to filter for a specific time range -- when this function is used to create a new feature, selecting a time range at that point will fill in the exact values for start and end.

In addition to the return types supported for non-velocity functions, velocity functions also support collection attributes, including returning Lists, Sets, and HashMaps.

For a complete DataVisor function list, please refer to the appendix.


5.3.2 Manage Functions View Functions

Similar to View all Features, in order to view all functions you can use the VIEW ALL on the dashboard.



As with other tables in Feature Platform, select the Edit Column to choose which columns to display. You can also sort columns in order to provide the view that best suits your work. From this page, you can also select Create Function to immediately build a new function.

You can view all possible functions in the appendix.

By clicking on Preview, you can see the function detail. Edit Functions

By clicking on Edit, you can change the selected function with the Save or Cancel option to complete it.



Note: Users can not delete Datavisor provided functions. Delete Functions

By selecting one or multiple functions, clicking Delete in the left upper corner will delete selected functions with a confirmation.

Note: Users can not delete Datavisor provided functions. Export Functions

By selecting one or multiple functions, clicking Export in the left upper corner, you can export selected functions in JSON format.



Was this article helpful?
0 out of 0 found this helpful



Please sign in to leave a comment.