Атты III халықаралық ғылыми-тәжірибелік конференция жинағЫ


"SCIENCE AND EDUCATION IN THE MODERN WORLD



Pdf көрінісі
бет40/418
Дата24.09.2022
өлшемі8,11 Mb.
#40095
1   ...   36   37   38   39   40   41   42   43   ...   418
Байланысты:
III TOM

"SCIENCE AND EDUCATION IN THE MODERN WORLD:
CHALLENGES OF THE XXI CENTURY" 
NUR-SULTAN, KAZAKHSTAN, JULY 2019
 
37 
processing data streams will be required. The ability to transfer the model and the integration of 
decision-making logic into operating systems is of paramount importance for preventing fraud 
on the widest scale - given the scope of fraudulent transactions. 
It is very important to be able to explain what exactly the machine learning system does. 
The system or solution, the principles of which are clear to the user, is usually called "white 
boxes". As a rule, methods and models of machine learning are impenetrable "black boxes" - the 
user does not know exactly how they work. It is very difficult (almost impossible) to explain to 
analysts why they got one or another result or solution. There are many approaches to adding 
evaluation cards, taking into account local linear approximation, as well as creating textual parts 
and graphic visualizations. All these are only approximate values, but they give users an idea of 
the machine learning model, as well as useful recommendations on the study of fraudulent 
activities. 
Everything is changing, and we need to be able to adapt to change. Continuous 
monitoring of machine learning-based fraud detection systems is an indisputable key to success. 
As the models and the underlying data change, the quality of the input data deteriorates and the 
overall system performance decreases. This problem is not only peculiar to machine learning 
systems, but also to rule-based systems. However, new methods of machine learning are able to 
effectively adapt to new, as yet unknown, patterns. This makes it possible to reduce the number 
of necessary measures (although not to exclude them all) in retraining and evaluating the 
operation of the machine learning system. 
An effective monitoring system actively explores the data that enters the system, 
evaluates the forecasts and explanations generated by the machine learning model, and also 
notifies administrators of changes in data and statistics trends before radical changes affect the 
performance of the entire company. 
For one of the financial institutions, combating fraudulent transactions was an intractable 
problem. It was necessary not only to identify illegal operations, but also to provide a high level 
of customer service. An effective fraud detection system should not block legitimate customer 
transactions. 
The financial institution sought to modernize the existing rule-based system and achieve an 
optimal balance between control functions and customer service. For this, his representatives 
turned to SAS. Their goal was to use the capabilities of neural networks to create two separate 
systems for evaluating fraudulent actions: 
A system for calculating the likelihood that a customer's account is under the control of 
fraudsters. 
A system for calculating the likelihood that a single transaction is fraudulent. Thanks to this 
approach, the financial institution was able to identify transactions in the amount of almost $ 1 
million per month, which were erroneously identified as fraudulent, as well as identify 
transactions in the amount of $ 1.5 million per month, which were fraudulent, but were not 
detected by the previous system. Our solution not only helped the company more effectively 
detect fraud, it also made it possible to significantly increase customer satisfaction by easing 
tensions in their relationship with the company. How? Thanks to a significant improvement in 
transaction confirmation procedures and increased fraud detection. 
Successful machine learning programs always involve varying degrees of continuous 
experimentation. It is not enough just to create a machine learning model and send it ―to float 
freely‖. Scammers are smart, and technology is constantly changing. The presence of an isolated 
environment, a sandbox, in which data scientists can experiment with various methods, data and 
technologies to combat fraud, in this case becomes a critical condition for the implementation of 
the most important programs. Investments that are aimed at optimizing the work and increasing 
the productivity of data scientists involved in the detection and prevention of fraud will pay for 
themselves almost instantly. 




Достарыңызбен бөлісу:
1   ...   36   37   38   39   40   41   42   43   ...   418




©emirsaba.org 2024
әкімшілігінің қараңыз

    Басты бет