RiskShield Machine Learning powered by BigML
FIGHTING FINANCIAL CRIME WITH HYBRID AI
The Hybrid Artificial Intelligence Approach takes the best of the knowledge-based and machine learning worlds to create a powerful financial crime fighting strategy.
The 4 Step RiskShield ML Process
Before jumping into a machine learning project, it is of utmost importance to define the goal and determine what question is being addressed. What do we want to predict or find out and do we have enough data to feed to the machine?
1. Data Preparation
In a typical machine learning project, approximately 80% of resources will be invested into this first step, which encompasses data transformations and feature engineering. The good news is, much of this can already be accomplished within RiskShield. Data from RiskShield can easily be transferred to the machine learning environment. With one click, a data set can be generated and field types are automatically identified (numeric, categorical, text, date-time, or items). The data set can then be split into training and testing data.
2. Model Building and Inspection
After the data set has been created and split, a supervised or unsupervised learning model can be selected and also executed with a single click using the training data. Model selection will heavily depend on the goals that were set at the beginning of the project. For example, supervised learning models can be used to predict whether or not a specific transaction is fraudulent. It uses past labeled data, either "fraud" or "not fraud", to predict the outcome of future transactions or claims. In this case, a tree was created and the selected path shows a number of transactions classified as fraudulent.
This is the stage during which the trained supervised model is evaluated using the previously separated testing data. These evaluations are also executed with one click. Predictions can be made using the test set and the quality of results can be seen through the use of simple ratios and advanced statistical measures based on numbers in the confusion matrix. This step answers the question, "how did the predictions generated from my model perform based on the unseen data." If the results are satisfactory, the model can be exported to real-time production.
Many machine learning projects fail because the trained models are never put into production. The seamless integration of the RiskShield ML learning environment allows the evaluated financial crime fighting models to be implemented into RiskShield's real-time decision engine. Simultaneously, transaction data as well as investigation results are fed back into the machine learning environment. This data is used to create new models and further train existing models, in the end creating a smooth feedback loop.
Supervised vs. Unsupervised Learning
With supervised learning, the data we are providing to the machine for learning purposes has clear labels. For example, transaction data can be labeled either as fraudulent or legitimate based on feedback from customers and investigation results. This information is passed on to the machine which is tasked with finding commonalities amongst the respective classifications so predictions can be made for future transactions as to which class the transaction belongs to: fraudulent or legitimate. This is very helpful in terms of pattern recognition, identification of new modus operandi as well as identifying the probability of a payment default.
Unsupervised learning is used when the input data does not have labeled responses. For example, in the case of insurance fraud or application fraud, criminals are not reporting back to financial institutions that they successfully committed a crime. With these types of data sets, clustering and anomaly detection algorithms are used to identify suspicious activity. For example, with application fraud, a merchant applies for an account as a bakery. The bank can take data from bakeries and input it into the machine to search for unusual activities or behavioral patterns that one bakery exhibits compared to all other bakeries. Anomaly detection and clustering can also prove useful within AML Compliance topics such as Ongoing Customer Due Diligence.
RiskShield Processing and Machine Learning Environments
The RiskShield Machine Learning environment is used to create and test models that can be implemented in real-time within the RiskShield decision engine. The models are transported to RiskShield using a PMML interface. These models can be used to supplement the scorecards, dynamic profiling, fuzzy logic and other methods associated with the traditional knowledge-based approach.
RiskShield ML offers you the following benefits:
Intuitive platform for developing supervised and unsupervised predictive models
Full integration with RiskShield’s decision engine in order to execute the models in real-time
Beautifully simple interactive visualizations that make models interpretable
Combination of predictive models and knowledge-based decision algorithms enabling precise financial crime detection
The Hybrid AI Story
Our colleague Roy Prayikulam introduces the concept of Hybrid AI for the financial crime fighting sector as seen in this clip from his presentation at the Merchant Payments Ecosystem (MPE) conference in Berlin.
Our InfoPaper includes more about the benefits associated with our Hybrid AI approach to financial crime and risk management. You will also find out more about our RiskShield ML offering.
The BigML Partnership
RiskShield ML, powered by BigML, serves as an enhancement to the current RiskShield machine learning offering, making model creation easy, accessible and interpretable for all organizational functions. Both INFORM and BigML have recognized great synergies between their respective technologies and are bringing a unique approach to the fight against fraud, money laundering and other financial crimes. BigML’s promise to bring machine learning to everyone in a beautifully simple environment combined with RiskShield’s powerful decision engine further enables financial institutions to take a Hybrid Artificial Intelligence approach to their fraud and money laundering prevention efforts.
Working with INFORM
INFORM GmbH is a global company in advanced optimization software systems and a leader in providing intelligent, customer-centric fraud prevention and AML compliance solutions. With RiskShield we offer a multi-channel platform that detects and manages suspicious activities, minimizing losses and optimizing efficiencies using advanced analytics, machine learning and intuitive rule management controls. The RiskShield Machine Learning package provides everything needed for a smoothly functioning Hybrid Artificial Intelligence approach encompassing both knowledge-based and machine learning methods. Our expert team of data scientists has a deep understanding of what it takes to develop and manage optimized decision models using machine learning.