Multifactor analyzer of arbitrary enterprise activity on the IEM platform

WANTED: talented mathematicians for an interesting and money contract.
Target specialization - statistics, mathematical modeling, neural networks.
The task description is below.


The second attempt to humanly formulate the problem from the previous posting.


goal


Application development for in-depth analysis of enterprise activity data accumulated in the IEM system . The output is supposed to receive an industrial commercial product, a universal solution for Big Data analysis, compatible with all IEM-solutions on the Ultimate Solid platform .


General statement of the problem


Development of a mechanism for searching for atypical deviations in the data of the performance results of standardized business processes. Initially, it is supposed to use the methods of statistics, possibly neural networks. And all that comes in handy too. The “atypicality” of deviations is a configurable parameter of the degree of paranoid system (it is also sensitivity, scalar).


A detailed attempt to describe Wishlist


The database of the IEM system collects complete structured information about the progress of the company's business processes in real time.



An example of visualizing the data structure of a real system operator in a financial projection


All history, transactions and other (including aggregated) attributes of processes and process events are saved. The progress of business processes is strictly standardized and guaranteed to be closed by the system circuit.


The output is data on the results of a large number of similar procedures (for example, “statement of account” - “receipt of money to the account” - “reservation of goods” - “shipment of goods”, and so a million times). The depth of detail has no fundamental limitations, and is determined by the depth of standardization of real business processes .


Inside the array of structured data, it is proposed to look for non-standard (relative to a given degree of paranoid) deviations.


Example: all sales managers have approximately the same turnover, profitability (profit), but one does not have atypically many warranty returns.


Continuous reliability, consistency and completeness of IEM database data is guaranteed by the platform. Among other things, they contain information about accounting objects in a variety of directories and about all the events and processes in documents, registers and other mechanisms. All data structures and their relationships and interactions are described by metadata stored in the same database in a structured, normalized way.


Ideally, the work of the future application should look like this: you configure access to the desired database, indicate the degree of paranoidity, and that’s all.


The application independently reads metadata, exhaustively describing the business logic of the enterprise, builds chains of business processes, groups by them the actual results of their development, and in each group searches for atypical ones.


Further, it performs certain actions with them, understanding the nature of which is included in the scope of the task (theoretical part), and at the output, risk factors spit out - the counterparty, employee, office, time for placing orders or other entities whose behavior information is stored in the system.


Finished about the ideal.


Performer Requirements


At the current stage, we need a person (a group of comrades) who a) deeply realizes what is at stake, b) build a mathematical methodology for solving the problem in the general case.


The methodology includes methods and heuristics for determining significant parameters (or determining indistinguishability for a given set of parameters), determining the process of constructing data analysis, and other technical details.


Given the vagueness and atypical nature of the task, any other appropriate proposals from people who can argue their own competence will be considered. The application in question has a high market capacity, therefore, various options for cooperation with a sane contractor are possible.


Technical points


The database is Oracle 12c EE.
If necessary, real-time translation can be implemented in Hadoop or similar repositories. But, following the IEM methodology, direct data collection from the application server is the preferred solution.


Suggestions to send to bigdata@ultimatebusinessware.ru


Also popular now: