Innovate2Transform: Axis Bank Addressing Financial Crime Management With AI & ML

India’s banking industry has been one of the most resilient and transformative industries in recent times.

One cannot overlook the tremendous impact of technology on banks in the recent decade. India’s digital payments ecosystem is among the most advanced in the world, with its Immediate Payment Service (IMPS) being the only system at level 5 in the Faster Payments Innovation Index (FPII).

With 27 public sector banks, 21 private sector banks, 49 foreign banks, 56 regional rural banks, 1,562 urban cooperative banks and 94,384 rural cooperative banks and numerous cooperative credit institutions, India’s banking system is expansive. Majority of these banks have undergone a massive digital transformation over the years.

This transformation involved migration from decade-old systems and practices, to more modern architectures and number-driven models. Understandably, there is a certain amount of risk involved, making the industry quite vulnerable to external malicious forces.

Financial Crime Management is one such aspect of the industry, whose relevance has grown with the industry’s rapid move to digital. Banks now recognise that a ‘tick-box’ approach to financial crime compliance isn’t enough; basically looking into internal data such as transactions of customers and filing STRs. Another challenging risk a bank faces is foreign correspondent banking. In essence, a foreign correspondent banking relationship is built on the effectiveness of a foreign bank’s AML compliance program and ongoing monitoring capabilities.  To improvise the above processes, banks need to rely on secondary data coming from unstructured text. This could be items flagging any adverse news about the existing customers, correspondent banks (foreign & local cooperative banks) & potential customers.

It is critical to solve these issues as damages arising from financial crime can have multiple implications for a financial institution – financial loss, damage to reputation, shareholder value destruction or regulatory sanction.

By establishing a robust infrastructure for fraud case management, bank can benefit from more effective fraud risk governance through the provision of timely, accurate information on changes to fraud risks and the effectiveness of fraud controls.

One of the country’s leading private sector banks Axis Bank utilised advanced technologies to address some challenges in the ongoing security environment and ensure high value impact to the business.

Challenges to be Resolved to Ensure High-Value Business Impact:

  • Capturing secondary information in the form of unstructured data (news) pertaining to Financial Crime, AML & Correspondent Banking to compliment the current STR (Suspicious Transaction Reporting) filing process and disseminating as Threat Alerts to Business Units
  • Specific targeted threat alerts with minimal spams (given the spam ratio is 0.4%)
  • Standard storage of news & retrieval system for future references & analysis

These challenges were being manually executed – by running Google search results, which could mean that vital information could get missed out. There was no standard process in place to ensure storage and retrieval of information.

Instead, a strategic, technology-driven strategic approach was adopted.

A News API allows partners to send a search query to Bing and receive a list of relevant news articles. A user can request a query with keywords related to AML/Financial Crime/Correspondent Banking keywords and receive the JSON file for the top 10 results in one go. Consequently, search queries are grouped AML keywords representing similar themes such as:

  • PMLA Investigation Money Laundering
  • Hawala
  • Financial Scam Fraud
  • Ponzi Scheme
  • IT Income Tax Department raid case

Implementation of AI Modules:

Unstructured Data Pre-Processing: The cleaning consists of getting rid of the less useful parts of text through stop-word removal, dealing with capitalization and characters and other details. Annotation consists of the application of a scheme to texts. Annotations may include structural markup and tagging of part-of-speech. Normalization consists of the translation (mapping) of terms in the scheme or linguistic reductions through Stemming, Lemmazation and other forms of standardization.

Similarity Analysis: When documents are represented as term vectors, the similarity of two documents corresponds to the correlation between the vectors. This is quantified as the cosine of the angle between vectors, that is, the so-called cosine similarity. Cosine similarity is one of the most popular similarity measure applied to text documents, such as in numerous information retrieval applications and clustering too.

Named Entity Recognition: This is a process where an algorithm takes a string of text (sentence or paragraph) as input and identifies relevant nouns (people, places, and organizations) that are mentioned in that string. Named Entity Recognition can automatically scan entire articles and reveal which are the major people, organizations, and places discussed in them. Knowing the relevant tags for each article help in automatically categorizing the articles in defined hierarchies and enable smooth content discovery.

How was this solution different from traditional approaches?

To an extent, sophisticated analytics programs can help Fraud team utilize their data by searching for and revealing patterns hidden in structured data. But these sources only account for 20 percent of all available data. The real challenge for enterprises is getting value from a sea of news, documents, PDFs and other sources that make up the other 80 percent of data that can’t be understood by computers—information otherwise known as unstructured data. NLP enables computer programs to understand unstructured text by using machine learning and artificial intelligence to make inferences and provide context to language, just as human brains do. It is a tool for uncovering and analyzing the “signals” buried in unstructured data.

Key Observations:

Adverse news themes for targeted search were provided by business as one of the parameters to fine-tune overall solution. Some of the themes like “Bitcoin” “Maoist funding” etc had high weightage but did not add value in terms of actionable alerts.  Additionally, while working on the Bing API, tuning parameters for better search results in global markets, it was observed that local market with local language translated themes did not yield better results than global parameters with themes in English language.

Impact of Solutions:

  • There was a 50% increment of trigger reviews with Critical nature of AML violations in Q4 FY 2017-18.
  • This system served as ready reckoner for regulatory submissions
  • Robust Infrastructure for storage & retrieval leads to better analysis & due diligence

Axis Bank was recognised as a NASSCOM Game Changer for solving this challenge.

This is part of our ongoing series Innovate2Transform, where we bring to you the latest trends and innovation stories from the industry. 

For more updates, follow NASSCOM CoE IoT-DSAI on LinkedIn, Twitter, Facebook, and Youtube

Share This Post

Leave a Reply