Scientific and technical results of work at the first stage of ASR (applied scientific research)

By | 29.12.2017

While completing the work at the first stage of applied scientific research on the topic “Development of neural network system for forecasting air accidents and for safety risks management based on retrospective data including variety of parameters and text descriptions of events” the following results were obtained:

  • There was carried out an analytical review of modern scientific and technical, normative, methodological literature, which touch upon scientific\technical problem investigated during the ASR;
  • Theoretical studies of existing neural network methods were carried out;
  • Selection and substantiation of the methods solving the tasks of ASR were made;
  • Patent investigations were made in accordance with State Standard 15.011-96;
  • Mathematical models of dynamic function of event flux density and event trends were developed;
  • Software implementation of developed mathematical model was carried out;
  • Software documentation for software implementation of developed mathematical models was developed;
  • Requirements to data sets for forming training and test samples of machine learning algorithms were developed;
  • An application for a security document (patent, certificate) was prepared and state fee was paid;
  • All necessary materials were prepared for participating in the events aimed at highlighting and promoting intermediate results of ASR (conferences, seminars, symposiums, exhibitions, etc., including international ones);
  • Preliminary marketing researches with the aim of studying marketing requirements, product requirements and commercialization perspectives for the results of intellectual activity obtained within completing ASR were carried out;
  • Project page containing developed content was created;
  • The equipment, components and software needed for the project was purchased;
  • Depersonalized data sets for formation of training and test samples of machine learning algorithms for the event forecasting system were developed,
  • Design and development of integration solution of machine learning modules in C++ into standard Customer’s IT system using J2EE and JNI technologies was carried out;
  • Standard methodology for implementing event forecasting system into Customer’s IT system;
  • Technical requirements and suggestions for the development, production and operation of product were developed, considering technological capabilities and features of the industrial partner — company of the real economy segment.