Fintech Sector

Loan Application Fraud Detection

Fraud Detection Predictive Modeling

Developing a model that can identify fraudulent loan applications for a financial institution?
Challenge Accepted!

Loan Application Fraud Detection

Process & Story

A company from the banking sector approached us with a significant problem. They wanted to develop an AI model that could identify fraudulent loan applications. Currently, banks and lenders struggle with this issue, generating substantial losses to them.

Problem

Loan and mortgage fraud are standard practices in the financial world. Application fraud is a type of criminal activity, in which a dishonest individual uses a stolen or contrived ID to apply for a loan or line of credit with the goal of not repaying it. The con artist establishes credible-looking credit and account activity over time to gain access to larger amounts of loans and higher lines of credit. Fraudsters utilize a break-out technique, which is a wide method of fraud that involves committing numerous instances of application fraud. Over time, the fraudster builds up a number of credit lines. The perpetrator then maxes out all of the credit lines in rapid succession before disappearing.

Identifying these cases manually is challenging because of the substantial volume of applications to be reviewed each day. Fraudulent applications take a long time to detect and even longer to deny, resulting in more significant revenue loss. Our client, just like other financial institutions, was experiencing the same issue.

Solution

The solution was to create a model that could sift through all loan applications and sort out the probable fraudulent ones from the genuine ones within minutes.

Our team started with a thorough understanding of the business process. We learned how both the client and its customers interact in this transaction, including deciding whether to offer a loan, write a contract, apply for a credit score, process loans, and other related activities.

We started by researching what kind of data is available and performing initial data exploratory tasks. As loan fraud at the application stage is not that common, we struggled with significant data imbalance. We decided to build a logistic regression model with the available features like loan transaction details, bank statements, credit score, history on banking activities and credit history, bank account details, borrower’s identity, age, income and asset information, and other public information.

The model based on logistic regression classified applications into binary categories of regular and fraudulent ones. This approach enabled fast case processing and immediate fraud alerts. Thanks to the model, bank operators can notice suspicious applications and immediately take action to protect their customers and the bank itself.

Case Study Schema Loan Application Fraud Detection

Tools

Google Cloud Platform

Google Cloud Platform

Tensorflow

Tensorflow

Python

Python

Have a similar project in mind?

LET'S TALK!

Challenges

Defining what to consider a fraud
The line between fraud and non-fraud can be very thin and the ML model has to incorporate both business knowledge, company policy, and statistical features. Sometimes even a good definition of fraud is a challenge. Building a good model requires iterations and very close collaboration with the client.

Datasets imbalance
First, since there are very few loan and mortgage fraud cases in general, we had fewer samples for creating a predictive model. The scarcity of examples was problematic because it imbalanced our training set with many examples from the positive category and very few examples from the fraudulent one. For this reason, we used various sampling methods to even out these numbers and reach an expected accuracy level for default prediction.

Model explainability and interpretability
Various AI-based applications used in the fintech industry are required by law to be explainable, i.e., banks and loan originators must explain their decisions to the regulators or auditors. This is why we decided to utilize logistic regression modeling, which is explainable and generates results that both humans and machines can easily understand.

Architecture design
The model had to work with a large dataset in order to automate the loan underwriting process. Predictions should be quick enough for bank representatives to process loan applications in minutes and accurately detect mortgage fraud. The architecture we designed could meet these requirements and generate accurate results.

Our other projects:

Insurance Sector

Auto Insurance Fraud Claim Detection

Developing an AI model that can detect car insurance fraud based on car accident claims?
Challenge Accepted!

Fraud Detection Predictive Modeling

Fintech Sector

Predicting Bank Loan Defaults for Profit Maximization

Developing an AI model for loan default prediction and using it to increase bank profits?
Challenge Accepted!

Predictive Modeling

Social Media Platform

Topic extraction and personalization models for idea exchange system

Developing an AI-based system that is able to extract and classify interesting ideas from articles and books and recommend them to social network users?
Challenge Accepted!

Natural Language Processing Recommender System

See more projects

Find us on

Need help with implementing AI in your business?

Let's talk blue circle

This site uses cookies for analytical purposes.

Accept Privacy Policy

In the interests of your safety and to implement the principle of lawful, reliable and transparent processing of your personal data when using our services, we developed this document called the Privacy Policy. This document regulates the processing and protection of Users’ personal data in connection with their use of the Website and has been prepared by Nexocode.

To ensure the protection of Users' personal data, Nexocode applies appropriate organizational and technical solutions to prevent privacy breaches. Nexocode implements measures to ensure security at the level which ensures compliance with applicable Polish and European laws such as:

  1. Regulation (EU) 2016/679 of the European Parliament and of the Council of 27 April 2016 on the protection of natural persons with regard to the processing of personal data and on the free movement of such data, and repealing Directive 95/46/EC (General Data Protection Regulation) (published in the Official Journal of the European Union L 119, p 1); Act of 10 May 2018 on personal data protection (published in the Journal of Laws of 2018, item 1000);
  2. Act of 18 July 2002 on providing services by electronic means;
  3. Telecommunications Law of 16 July 2004.

The Website is secured by the SSL protocol, which provides secure data transmission on the Internet.

1. Definitions

  1. User – a person that uses the Website, i.e. a natural person with full legal capacity, a legal person, or an organizational unit which is not a legal person to which specific provisions grant legal capacity.
  2. Nexocode – NEXOCODE sp. z o.o. with its registered office in Kraków, ul. Wadowicka 7, 30-347 Kraków, entered into the Register of Entrepreneurs of the National Court Register kept by the District Court for Kraków-Śródmieście in Kraków, 11th Commercial Department of the National Court Register, under the KRS number: 0000686992, NIP: 6762533324.
  3. Website – website run by Nexocode, at the URL: nexocode.com whose content is available to authorized persons.
  4. Cookies – small files saved by the server on the User's computer, which the server can read when when the website is accessed from the computer.
  5. SSL protocol – a special standard for transmitting data on the Internet which unlike ordinary methods of data transmission encrypts data transmission.
  6. System log – the information that the User's computer transmits to the server which may contain various data (e.g. the user’s IP number), allowing to determine the approximate location where the connection came from.
  7. IP address – individual number which is usually assigned to every computer connected to the Internet. The IP number can be permanently associated with the computer (static) or assigned to a given connection (dynamic).
  8. GDPR – Regulation 2016/679 of the European Parliament and of the Council of 27 April 2016 on the protection of individuals regarding the processing of personal data and onthe free transmission of such data, repealing Directive 95/46 / EC (General Data Protection Regulation).
  9. Personal data – information about an identified or identifiable natural person ("data subject"). An identifiable natural person is a person who can be directly or indirectly identified, in particular on the basis of identifiers such as name, identification number, location data, online identifiers or one or more specific factors determining the physical, physiological, genetic, mental, economic, cultural or social identity of a natural person.
  10. Processing – any operations performed on personal data, such as collecting, recording, storing, developing, modifying, sharing, and deleting, especially when performed in IT systems.

2. Cookies

The Website is secured by the SSL protocol, which provides secure data transmission on the Internet. The Website, in accordance with art. 173 of the Telecommunications Act of 16 July 2004 of the Republic of Poland, uses Cookies, i.e. data, in particular text files, stored on the User's end device.
Cookies are used to:

  1. improve user experience and facilitate navigation on the site;
  2. help to identify returning Users who access the website using the device on which Cookies were saved;
  3. creating statistics which help to understand how the Users use websites, which allows to improve their structure and content;
  4. adjusting the content of the Website pages to specific User’s preferences and optimizing the websites website experience to the each User's individual needs.

Cookies usually contain the name of the website from which they originate, their storage time on the end device and a unique number. On our Website, we use the following types of Cookies:

  • "Session" – cookie files stored on the User's end device until the Uses logs out, leaves the website or turns off the web browser;
  • "Persistent" – cookie files stored on the User's end device for the time specified in the Cookie file parameters or until they are deleted by the User;
  • "Performance" – cookies used specifically for gathering data on how visitors use a website to measure the performance of a website;
  • "Strictly necessary" – essential for browsing the website and using its features, such as accessing secure areas of the site;
  • "Functional" – cookies enabling remembering the settings selected by the User and personalizing the User interface;
  • "First-party" – cookies stored by the Website;
  • "Third-party" – cookies derived from a website other than the Website;
  • "Facebook cookies" – You should read Facebook cookies policy: https://www.facebook.com/policy/cookies
  • "Other Google cookies" – Refer to Google cookie policy: www.google.com/policies/technologies/types/

3. How System Logs work on the Website

User's activity on the Website, including the User’s Personal Data, is recorded in System Logs. The information collected in the Logs is processed primarily for purposes related to the provision of services, i.e. for the purposes of:

  • analytics – to improve the quality of services provided by us as part of the Website and adapt its functionalities to the needs of the Users. The legal basis for processing in this case is the legitimate interest of Nexocode consisting in analyzing Users' activities and their preferences;
  • fraud detection, identification and countering threats to stability and correct operation of the Website.

4. Cookie mechanism on the Website

Our site uses basic cookies that facilitate the use of its resources. Cookies contain useful information and are stored on the User's computer – our server can read them when connecting to this computer again. Most web browsers allow cookies to be stored on the User's end device by default. Each User can change their Cookie settings in the web browser settings menu: Google ChromeOpen the menu (click the three-dot icon in the upper right corner), Settings > Advanced. In the "Privacy and security" section, click the Content Settings button. In the "Cookies and site date" section you can change the following Cookie settings:

  • Deleting cookies,
  • Blocking cookies by default,
  • Default permission for cookies,
  • Saving Cookies and website data by default and clearing them when the browser is closed,
  • Specifying exceptions for Cookies for specific websites or domains

Internet Explorer 6.0 and 7.0
From the browser menu (upper right corner): Tools > Internet Options > Privacy, click the Sites button. Use the slider to set the desired level, confirm the change with the OK button.

Mozilla Firefox
browser menu: Tools > Options > Privacy and security. Activate the “Custom” field. From there, you can check a relevant field to decide whether or not to accept cookies.

Opera
Open the browser’s settings menu: Go to the Advanced section > Site Settings > Cookies and site data. From there, adjust the setting: Allow sites to save and read cookie data

Safari
In the Safari drop-down menu, select Preferences and click the Security icon.From there, select the desired security level in the "Accept cookies" area.

Disabling Cookies in your browser does not deprive you of access to the resources of the Website. Web browsers, by default, allow storing Cookies on the User's end device. Website Users can freely adjust cookie settings. The web browser allows you to delete cookies. It is also possible to automatically block cookies. Detailed information on this subject is provided in the help or documentation of the specific web browser used by the User. The User can decide not to receive Cookies by changing browser settings. However, disabling Cookies necessary for authentication, security or remembering User preferences may impact user experience, or even make the Website unusable.

5. Additional information

External links may be placed on the Website enabling Users to directly reach other website. Also, while using the Website, cookies may also be placed on the User’s device from other entities, in particular from third parties such as Google, in order to enable the use the functionalities of the Website integrated with these third parties. Each of such providers sets out the rules for the use of cookies in their privacy policy, so for security reasons we recommend that you read the privacy policy document before using these pages. We reserve the right to change this privacy policy at any time by publishing an updated version on our Website. After making the change, the privacy policy will be published on the page with a new date. For more information on the conditions of providing services, in particular the rules of using the Website, contracting, as well as the conditions of accessing content and using the Website, please refer to the the Website’s Terms and Conditions.

Nexocode Team