RSS
Логотип
Баннер в шапке 1
Баннер в шапке 2

Jet Detective

Product
The name of the base system (platform): Apache Hadoop
Developers: Jet Infosystems
Date of the premiere of the system: 2017
Branches: Mechanical engineering and instrument making,  Trade,  Financial services, investments and audit
Technology: SaaS - The software as service,  cybersecurity - the Fraud detection system (fraud)

The solutions Jet Detective are the cornerstone of machine learning technology and artificial intelligence. The main advantages of Jet Detective – the speed and accuracy of identification of a fraud with the minimum indicators of false operations.

The solution from Jet Infosystems aggregates hundreds of thousands of operations per minute from a set of sources (from network channels to business systems), spending fractions of a second for the analysis of each event. Application of the general expert rules of the analysis along with methods of machine learning allow to prevent not only the known types of fraud, but also implementation of new schemes. Besides, the solution helps to reveal the preparing plunders and complex cases of fraud when illegal acts are committed by different persons, in the different systems and at different times.

The Jet Detective system is intended for the companies of any industries: banks, retail, industrial enterprises, etc. It can be unrolled both on the platform of the customer, and in a cloud according to the SaaS model. Today based on this solution the Jet Infosystems company already implemented several pilot projects in the organizations of the financial sector and retail.

As the director of the cybersecurity Center of the company Alexey Grishin reported, systems are the cornerstone of machine learning technology, artificial intelligence and processing with their help of Big Data. At the same time developers used a software platform of Apache Hadoop which can be implemented on any computer architecture and allows to scale linearly computing powers of systems constructed on it.

The Jet Infosystems company claims that (thanks to the Hadoop platform) from relational DBMS and indiscriminateness to the capacity of servers (thanks to it) total costs on the Jet Detective system approximately allow to reduce failure by 40% in comparison with the systems of an antifraud with traditional architecture.


"Emergence of Jet Detective became the logical answer to present requirements of the companies for which the fraud problem is most relevant. If earlier the created schemes of fraud existed for years, then today they change so quickly that the companies do not manage to study them and the more so to configure the mechanism of their identification in systems, – Grishin Alexey, the director of Information Security Center of Jet Infosystems company comments. – Communicating with customers, we understood that the majority of systems offered in the market do not meet completely their requirements. Some solutions are closed – difficult to make necessary changes to them, and sometimes and it is impossible. Others – are insufficiently exact in determination of a fraud. The third – demand continuous control and independent making changes in rules of the analysis that reduces the speed of reaction to illegal acts. Our long-term examination in the field of an antifraud and software development and also understanding of requirements of the market gave us the chance to create essentially new solution which does not have the listed shortcomings".


Big Data machine learning The head "and" of the center of software solutions of Jet Infosystems company Kolesnikov Evgeny emphasized that use during creation of the Jet Detective system of a markup language for forecast modeling (Predictive Model Markup Language, PMML) is the innovation for the on-line systems antifraud. PMML as it noted, along with integration flexibility concerning data sources allows the user to adjust independently initial option of model according to changes in and out of customer company.

According to Evgeny Kolesnikov, specialists of the company constructed mathematical factory which depending on the data arriving on an input selects those algorithms and methods of processing which set for this specific case yields the best result.

Jet Detective in the least loaded hours with available frequency (for example, once a day) is retrained and every time creates new model taking into account the data obtained for last period. Self-training does a system effective when using and in those companies in which there are no analysis and data processing specialists in a digital format (Data Scientist). Address the help of integrator concerning system operation to customers, according to in Jet Infosystems, it is necessary seldom — only in case of essential deviations in work from the expected results.

Evgeny Kolesnikov considers difference of this development from the systems of counteraction to a fraud presented at the Russian market availability of mechanisms of analytics to their modification, a possibility of operational change of rules of decision making, flexible scaling and the user-friendly interface facilitating to the operator process of adoption of the correct solutions. Besides Jet Detective used as the platform which is able to import data from others, other than PMML, models, for example, constructed on the IBM SPSS Modeler platform.