Comarch Anti-Money Laundering (CAML) is a fraud detection software dedicated to financial institutions obligated to monitor, investigate and report transactions of a suspicious nature to financial regulatory authorities. It optimizes the existing anti-money laundering processes by significantly enhancing the effectiveness of most commonly used – and inefficient – rule-based approaches, characterized by high false-positive rates and unable to consider complex interdependencies between various activities carried out to launder money.
One of the most important features of Comarch Anti-Money Laundering software is the ability to learn from historical examples, or else: discover hidden patterns allowing to spot and understand relationships and similarities between data and, further down the road, learn to detect anomalies or categorize and predict specific events. Our transaction monitoring software makes investigating suspicious activities simpler, faster and more effective by reducing the number of false alarms and minimizing risk of false negatives through the use of advanced anomaly detection algorithms designed to uncover new money laundering patterns. All these features translate into greater productivity and efficiency of AML departments in financial institutions, which in turn allows the institutions to reduce their operating costs.
At the heart of Comarch’s AML software is an analytical AI-based engine that processes data streams and detects money laundering activities. The scope of operations being monitored is broad and covers deposits, withdrawals, purchases, fund transfers, merchant credits, payments, trading activities, investments and social media activity. The engine leverages most recent supervised and unsupervised learning techniques to discover anomalies and improve the detection breadth.
Additionally, it categorizes and scores suspicious transactions according to their probability of being a real money laundering attempt. Calculated score output serves as a reference to rank new attempts. The incidents going beyond certain score value are sent to analysts for further review while the ones remaining below the value can be discarded based on historical cases. Such alert prioritization greatly speeds up the work of AML teams.
Before the AI engine can correctly analyze bits of information and draw conclusions, it needs properly prepared data. The process of getting data ready for an anti-money laundering program requires several preparatory activities. The key step of data preparation is a transformation which typically involves scaling, decomposition and aggregation. This step is also referred to as feature engineering, and, if properly carried out, can be very beneficial to the performance of a final solution.
Data preparation is a broad subject that can involve a lot of iterations, explorations and analyses. The data pre-processor module is responsible for integration with data sources existing in financial institutions’ databases and the correct introductory processing and transformation of data which will be subsequently analyzed in the AI engine.
The front office module provides tools for anti-money laundering analysts to manually review and analyze suspicious activity alerts. It supports users in analytical tasks allowing to manage the alert flow, which is important in the case of a multi-stage analysis process. The tool also comes with a visualization of statistics, detected anomalies and suspicious cases arranged by categories, in a clear and understandable way for the analyst. The front office user interface is designed based on the latest UX standards and can be delivered as a standalone application or easily integrated with the existing solutions, acting as an extension to them.
The money laundering process evolves over time, and new advanced fraud patterns appear, which makes it necessary to constantly monitor the solution’s performance. As new data arrives, the algorithms that were trained based on historical data may require periodical re-training (e.g. in the case of sudden increase in the number of false alarms). The monitoring module is responsible for gathering statistics, analyzing results and warning about any unusual performance loss.
In response to market trends and real business and technology issues in the compliance area, Comarch has created an AI-based anti-money laundering software (Comarch Anti-Money Laundering) that significantly increases the efficiency of suspicious transaction detection compared to rule-based methods.
CAML is based on machine learning which is a state-of-the-art approach where Artificial Intelligence (AI) is used for creating models that, based on historical data, can determine the occurrence of certain events with high accuracy.
The ability to learn from past incidents is one of the most important features of Comarch’s fraud detection software. It can spot and understand relationships and similarities between data and, further down the road, learn to detect anomalies or classify and predict specific events. CAML’s algorithms are able to learn with or without supervision.
A rapidly growing number of new techniques and methods of money laundering is one of the reasons of higher false-negative rates. That is where Comarch Anti-Money Laundering can help by applying anomaly detection, which has the potential to identify events which do not conform to an expected pattern in a dataset, and improve the breadth of the said detection by uncovering new money laundering patterns. CAML leverages most recent unsupervised learning techniques and uses unlabelled data to detect bank customers’ abusive activities.
Once an anomaly is detected, it gets categorized and scored, which illustrates the probability of a real money laundering case. A binary classification is used for categorizing anomalies. As this is a supervised learning technique, the algorithm is given a training set containing input and output data, labelled with a real categorization. After it had been shown a number of historical cases, it is enough to provide it with data inflow, indicate the purpose, and let it do the rest. The incidents going beyond certain score value are sent to analysts for further review while the ones remaining below the value can be discarded based on historical cases. Such alert prioritization greatly speeds up the work of AML teams.
Once the best models are built and taught, they can be applied to the validation cases–a sample from the modelling dataset that was not used to train the models, to confirm its robustness. Validated model is then deployed into bank’s IT environment to detect and score anomalies.
Applying machine learning to Comarch Anti-Money Laundering helps reduce the false-positive rate for cases needing manual review, which translates into greater efficiency for AML analysts through the use of advanced classification and anomaly detection algorithms.
Tell us about your business needs. We will find the perfect solution.