A Global Award-Winning AutoML Platform – DriverlessAI for Financial Institutions

It’s great to take the world by the storm with our partners H2O.ai. Their Driverless AI platform combined with our BFSI intensive domain expertise is enabling banks and financial institutions to maximize their ROI’s.

What Does Driverless AI Do?

Driverless AI makes automating the rote data science process feasible. With Driverless AI, data scientists can work on projects faster and more efficiently by using automation to accomplish key machine learning models in just minutes or hours.

Its objective is to achieve the highest predictive accuracy in a much shorter time. This state-of-the-art automatic machine learning platform delivers unique and advanced functionality for model validation, feature engineering, model tuning, model selection and deployment, machine learning interpretability. Driverless AI also offers automatic visualization and MLI (Machine Learning Interpretability).  

Driverless AI’s extensible customizable data science platform addresses the needs of a variety of use cases for every enterprise in every industry. It runs on commodity hardware. Driverless AI is developed to leverage graphical processing units (GPUs), such as multi-GPU workstations and services.  

Driverless AI for the financial industry

Financial institutions are one of the earliest adopters of digital technologies like Machine Learning (ML), Artificial Intelligence (AL), Data Science, etc. They operate quickly, and with scale, to detect fraud, monitor transactions, and extend customer-friendly services. Financial portals must be ready to react as customers transact at any time.

Driverless AI enables financial portals to speed up the development and deployment of models efficiently. Driverless AI provides the following advantages over the rule-based system:

  • Underwrite the credit
  • Detect and reduce fraud
  • Determine risk and insight
  • Knowing the customer and cross-selling products

Driverless AI in detecting money laundering

Money laundering is a serious and prevailing problem for the financial services sector. The United Nations Office on Drugs and Crime estimates that around $2 trillion is washed annually through the banking system. For banks that fail to stop money laundering, fines have increased by 500 times in the last decade, crossing $10 billion each year. Due to which banks have employed large teams of people to find and investigate suspicious transactions.

Driverless AI’s advanced features engineering and model creation capabilities enable financial service providers to build anti-money laundering (AML) models quickly.   

AML system offers the following benefits over the rule-based system:

  • It reduces the false positives
  • It can ingest solutions customized for money laundering.
  • The solutions are strategically placed between the AML system and the investigator
  • It uses an out-of-loop ML approach to classify alerts as false positives, or true positive
  • A curated set of alerts are given to the investigator

Demo with Driverless AI

Dataset

The AML system works with pre-marked anti-money laundering data, transactional banking data, and banking KYC data. The following demo contains a synthetic dataset that has the same distribution as the financial dataset. The dataset is made of both numeric and categorical columns with fields like person’s account number, date, kind of business, typology, etc.


The target column helps to understand whether an alert is suspicious enough to be sent for further investigation. It helps in reducing false positives. The important point to note is that the model learns the behavior based on data (transactions), and not specific to any individual.

How Driverless AI works

The data is integrated into a driverless AI instance and treated as a supervised ML problem. Once the dataset has been ingested, the recipe specifically customized for identifying false positives in money laundering alerts will be brought in from the expert settings.

Finally, you can select F1 as the scorer since the aim is to reduce the false negatives as much as possible. Now, you can adjust the accuracy, and time and interpretability settings to suit our needs.

Driverless AI Screen with AML Recipe in Action

The H2O Driverless AI Experiment AML Demo screen appears as follows:

Driverless AI Experiment

Apart from the F1 scores, the driverless AI enables users to see the variable importance of the features used for the model building purpose, at each iteration.

Here are the steps performed by the driverless AI to find the optimal final model:

Driverless AI improves profitability of Visión Banco

Visión Banco is a financial services provider based in Asunción, Paraguay. It provides loans and other financial services to small and micro-sized companies in Paraguay. Financial services offered by the bank include credit card services, remittances, utility and tax collection services, pension plan contribution plans, and payment transfer services.

Data scientists at Visión Banco were required to expand its services and offers to customers, and easily determine credit risks with accuracy and speed. The team also wanted to enhance its practices by implementing predictive analytics to predict customer churn, but could not do it efficiently without a new tool.

First, they hired an external consultant to design a model using IBM SPSS Software. It took a considerable time of a year to complete the process. They also used some open source tools like R, H2O, and Openscoring.io. Yet the predictive analytics were still taking considerable time and effort.

Engaging with H2O Driverless AI

Ruben Diaz, a data scientist at Visión Banco, joined a beta test of H2O Driverless AI to simply see how Driverless AI might perform in developing models for the bank. In a contest to start a proof of concept for Driverless AI, Diaz used beta software to develop models to assess churn prediction, and he secured a position in the top 10.

Implementation of H2O Driverless AI

Visión Banco implemented H2O Driverless AI on IBM Power System. As a result, data scientists saved time and increased revenue by building and deploying models that improved the accuracy of credit risk model and doubled customer propensity to buy a credit product.

Here are the improvements that Visión Banco saw:

  • Time savings: According to Diaz, in the past, the process of building the models would take 6 months or more.  Now it is less than a week.
  • Accuracy improvements: With tighter models, the bank estimates that it can earn millions in additional revenue by targeting the right customers for offers.
  • Increased results: The bank increased its customer propensity twofold to buy rate, using H2O Driverless AI.  

“The deployment part for data scientists is sometimes forgotten,” explains Diaz, “but it’s very important. It’s the way the model comes to life.” Diaz and Visión Banco got impressed with the ease of H2O deployment.

The process of data science heavily depends on building models – which takes time and expertise to become perfect. Driverless AI enables data scientists to work on projects faster and more efficiently by using automation to build key machine learning models in just minutes or hours. Efficient models can ultimately facilitate real business improvements, like enhanced projections.

To watch a quick demo connect with us at – neelesh.dugar@act21softwares.com

Subscribe to our Blog Posts
Thank you!