1. Home
  2. Sample Projects
  3. Car Insurance Fraud Detection

Car Insurance Fraud Detection

◷ Reading Time: 9 minutes

This Car Insurance Fraud Detection tutorial shows how to detect a car insurance fraud using a given dataset and send an alert email along with the car registration details.

In this project, we have used the following extensions.

  • Analytics:
    The analytics extension was used to predict and detect the fraud possibility of a given car insurance claim.
  • Robotics:
    Robotics capability was used to retrieve car registration information from the public registry.

Running the Car Insurance Fraud Detection sample

1. Update the Email addresses of the sender and recipient on the page, Emails.xml.

  • SenderEmail: The alert email will be sent from this email
  • SenderEmailPassword: The password of the sender email
  • RecipientEmail: The alert email will be received at this email address
Project Explorer view showing all project documents with emails.xml selected from the Glossaries folder. Also displaying Emails.xml document open showing the name and email options

2. On the document MailFlow.xml, update the SMTP Server information if you are not using a Gmail address for SenderEmail (Currently, the sample project contains Gmail SMTP information).

Mainflow.xml diagram with the Alert Authorities node selected and the Properties view showing the email Server details highlighted

3. Open MainFlow.xml.

Project Explorer view showing the project documents and the MAinFlow.xml selected, Mainflow.xml diagram is open

4. Click on Logic Run Template.

Detail view of the Logic Run Template menu item

5. Click on one of the given templates.

Logic run template view showing available test templates

6. Click Debug.

Detail view of the Debug menu item

7. Click Next Step to go step by step.

Detail view of the Debug Next Step menu item

8. If fraud is detected, you will receive an email to the email defined as RecipientEmail from SenderEmail.

Sample email view

Car Insurance Fraud Detection process steps

The process contains a set of main steps:

  1. Detect Frauds using a dataset.
    Using data analytics, first, detects whether there is any fraud possibility in the claim request.t
  2. If fraud is detected, check car registration information from the public directory.
    Using Robotics, it checks the car registration using the car registration number.
  3. If there is an error in the registration number, send a notification.
    When the registration number cannot be found in the public directory, creates an error notification.
  4. Create the email alert message to inform the authority.
    Using the registration number details creates the email content that needs to be sent. If there is no registration, it will create a notification
  5. Send the email alert to the required authorities.
    Based on the created email content, send an alert email to a predefined email.
MainFlow showing the decision diagram and the respective nodes numerated as explained above

Project Description

The project contains the following files:

DRD (Decision Requirement Diagram)

  • DRD.xml: Shows the data flow

Flow

  • MainFlow.xml: The main flow that connects all the steps
  • Check Registration Online.xml: The flow that retrieves the car registration from the public directory using the Robotics extension

Decision Table

  • Create Email Content.xml: To create the email content according to the registration results

Business Formula

  • BusinessFormula.xml: Define the formulas that are repeatedly used in the project
  • Emails.xml: The email addresses used for the alert emails

In the Fraud Prediction folder, you will see the files related to fraud detection using analytics.

  • insurance_claims.frml: The predictive model information
  • insurance_claims.InputOutput.Concept.xml: The Concept file of the model
  • insurance_claims.DRD.xml: The DRD that calls the predictive model

The dataset used for the project is inside the folder, Dataset.

  • insurance_claims.csv: The dataset of previous frauds. This is the file that was used to create the predictive model.

The Flow Design

1. The input for this project contains claim data for car insurance which is defined in insurance_claims.InputOutput.Concept.xml. This was created automatically using the AutoML capability.

Fact concept for the project

MainFlow.xml defines the steps of the flow.

2. Next, We have used the AutoML capability of FlexRule and used the Regression method to build the model.

Main flow diagram view with the Detect Fraud node selected

The following dataset (insurance_claims.csv) was used to create the model using regression classification.

Then it automatically created the following frml file (insurance_claims.frml) with the predictive model.

It also created this DRD (insurance_claims.DRD.xml).

As well as the concept file (insurance_claims.InputOutput.Concept.xml).

3. After that, the Robotics capability was used to retrieve car registration information from the public registry.

The following Check Registration Online.xml uses Robotics Web automation capability to navigate to the vehicle registry site, enter the registration number, retrieve details, and create a message with the details.

It will search the registration number in the government vehicle registry which is a public site.

It will get the results from the Check Registration page.

If there is no data found, it will show a notification.

4. Create alert email content using the retrieved registration information

The following Decision Table Create Email Content.xml was used for that.

5. Send the alert message to the predefined recipient.

The email addresses are defined in the business formula document (Emails.xml).

6. All the expressions used in the project are defined in BusinessFormula.xml.

Download Project

Use the attachment at the end of the page to download the sample project.

Updated on January 18, 2024

Article Attachments

Was this article helpful?

Related Articles