Continuous Automated Machine Learning

Machine Learning product deployment is difficult and takes time. As business changes, redeploying working ML models becomes even more difficult.

HAKsys's Continuous Automated ML deployes fast.
It requires very little maintenance and intervention in production to produce high accuracy results continuously.

HAKsys does this by supporting the entire data and analytics pipeline continuously by its:
- its ability to integrate to different data sources easily
- automated data cleaning, transformation and selection procedures
- advanced modeling enabled by automated configuration of hundreds of different machine learning models
- continuous machine learning pipeline to support continuous business
- high performance and scalability on real time data streams
- augmented intelligence enabled by understandable models that can learn from data and humans, continuously

Continuous Automated ML for Banking

Payment Fraud Detection

Challenge:

Detection of fraud patterns in credit card payment streams and in particular Card Copying Fraud.

Solution:

We used past data on transactions such as frequency, amount, merchant, location, authentication type of purchases to create features to detect copy scores for the cards. We also created terminal fraud scores. Combining these two, we achieved better card copying fraud scores.

Result:

Up to 80% accuracy was achieved on fraud detection. In addition to card copying fraud, we were also able to detect other and new forms of payment fraud with high accuracy.

NPL

Challenge:

Customer has to spend a lot of call center and legal personnel time to figure out if a customer can pay any of his debt and if so how much and when. It is also costly to gather data on customers.

Solution:

Based on call center frequency and speech data, as well as past behavior of the customer and the history of his debt, machine learning classifiers are trained for different tasks, such as if the customer will agree to a protocol agreement and make a small down payment or if he will make any payments within the next months.

Result:

HAKsys achieved 15% improvement on the payment probability prediction. Classifiers were trained online based on ever changing customer data. ML result explanations helped with better understanding of the customer behavior. As well as value of each customer feature

Segmentation

Challenge:

Credit risk can change based on many complex factors, ranging from the client's behavior patterns and and economical factors. The bank wants to automate the credit risk detection and prediction process which is performed by its credit risk team using different data sources and their experience.

Solution:

Past data on transactions of each customer, payment frequency, amount, currency, increase and decrease on the last payment ratios, LCV is used to create features to detect and predict credit rating. Heterogeneous data sources are collected and processed using HAKsys's continuous ML technology.

Result:

Up to 75% for risk detection and 70% accuracy for predicting risk in the next quarter are achieved. The detection models help the bank employees understand what affects risk for each client micro-segment and hence help them design actions to be taken for the benefit of the client and the bank.