AN APPROACH TO THE METEOROLOGICAL DATA ANALYSIS AND PREDICTION FOR THE TRAINED MODEL FOR THE METEOROLOGICAL STATION: HUMAN FAILURES IN SYSTEMS OF SYSTEMS DESIGNING

Abstract

Well-known approaches to solve the applied problems in Big Data System are the Deep Learning Machines. The specific Big Data Systems are the parts of real weapon systems in artillery. To improve the efficiency of such systems is an important task for Ukrainian Defence forces. The meteorological station operates with Big Data determining the parameters of the atmosphere for some artillery mission. How to use the DLM and special models that are capable to predict some meteorological information for artillery mission is our task for research and system designing.

The designing of the Systems of Systems (SoS) is a complicated process, so some kind of the decomposition is a typical approach with next control and adaptation. Obviously, the control and analysis that are qualified and done in time are the measures that could prevent from some incorrect or non-optimal solutions in the SoS, the time and some sources would be reduced too in the process of the solving the problem. It is shown in the paper the first stage of designing some part of the SoS concerning to the information processing system in the concrete sample of a meteorological station. The meteorological data were used as the reference data for the trained model that could calculate the meaning of horizontal and vertical angles of the turning antenna. This ‘feed-back’ task will allow to estimate the possibilities of the different trained models on the actual values of the angles. The dataset was obtained from "Radiotheodolite-UL" meteorological station. A regression model was created on Python with using such libraries as Pandas and Scikit-learn. The final result was unexpected, it allows to make an incorrect evaluation of the DLM for similar tasks. The reasons for failure were analyzed. The researches for the designing of the SoS are continued.