CC BY
0
Секция «Математические методы моделирования, управления и анализа данных»
УДК 519.246.78
ОБРАБОТКА ЭКОНОМИЧЕСКИХ ПОКАЗАТЕЛЕЙ ДЛЯ ЗАДАЧ ПРОГНОЗИРОВАНИЯ ВРЕМЕННЫХ РЯДОВ
П. А. Панфилов, Т. В. Пен
Сибирский государственный университет науки и технологий имени академика М. Ф. Решетнева Российская Федерация, 660037, г. Красноярск, просп. им. газеты «Красноярский рабочий», 31
E-mail: pen_tatiana@mail.ru
В представленной работе рассматривается прогнозирование индекса производства в промышленной сфере при помощи нейронной сети.
Ключевые слова: статистический анализ, нейронные сети, индекс промышленного производства, многослойный персептрон.
PROCESSING ECONOMIC INDICATORS FOR PROBLEMS OF FORECASTING TIME SERIES
I. A. Panfilov, T. V. Pen
Reshetnev Siberian State University of Science and Technology 31, Krasnoyarskii rabochii prospekt, Krasnoyarsk, 660037, Russian Federation
E-mail: pen_tatiana@mail.ru
In the present paper, we consider the prediction of an industrial index using a neural network.
Keywords: statistical analysis, neural networks, industrial production index, multilayer perceptron.
Analysis and processing of statistical indicators is the main criterion for making managerial decisions in the planning of economic tasks. This article will consider the task of forecasting production in the industrial sector.
To create a forecast of future indicators, it is necessary to build a mathematical model. To build such a model, an analysis of the statistical base was carried out and indicators were used that reflect the indicators of the industrial sphere of Russia from 2006 to 2019 inclusive [1].
The natural way to process and analyze the data and the dependencies of indicators is the time series apparatus.
To predict time series, we use a structural model, namely, in the future, the creation of a model is carried out using the mathematical apparatus of neural networks.
The forecast construction algorithm is as follows: first you need to conduct a preliminary analysis of the time series; then form a set of training examples; the next step is to build a neural network model; then it is necessary to train the neural network by the method of back propagation of error; The final step is to build a forecast using a previously trained neural network model.
By definition, a neural network model contains an input layer, hidden layers and an output layer. For processing and analyzing data, and in the future to build the model, we will use a multilayer perceptron.
When analyzing statistical indicators, it was revealed that the variables will take only positive values, which means that a differentiable nonlinear logistic function will be used as the activation function.
Актуальные проблемы авиации и космонавтики - 2020. Том 2
For the learning process of the neural network, the backpropagation method will be used. The main essence of this method is to supply training data to the network inputs, back propagate the error, and then adjust the weights for more accurate output indicators.
The learning process itself begins with an untrained and unprepared network. In the inner layer, the signals are transmitted through the installed network and determine the output indicators in the output layer. After that, all output values are compared with the target values, the occurrence of differences indicates an error. The considered error is a scalar function of the balance and should be minimized when the outputs correspond to the set results [2].
Upon completion of training, it becomes possible to optimize synoptic weighted compounds, and also establish connections between inputs and outputs.
Using mathematical software, which is an integral assistant in the analysis of statistical indicators, which includes analysis and forecasting of time series, a model for forecasting the dynamics of the industrial production index was built.
The results obtained using a neural network demonstrate the accuracy of the model. Based on the data obtained, we can conclude that the predicted values of the neural network have relatively close values with the initial indicators of the series. So, the constructed model is adequate and can be used for further analysis of a number, and then management decisions in the consideration of economic problems [3].
Based on the results obtained, a forecast of the industrial production index for 2020 was built.
The error of the constructed model is permissible, does not exceed 15%, the quality of the obtained model can be estimated as an adequate model, which means using it in further work with statistical economic indicators, on the basis of which you can make managerial decisions in the future.
Using the mathematical model, we can conclude that the trend is positive, in 2020 production is expected to increase in the industrial sector.
References
1. State statistics [Electronic resource] - 2020. - URL: http: // gks.ru/ (accessed date: 03/01/2020)
2. Khaikin S. Neural networks: full course, 2nd edition; trans. from English - M .: Publishing Williams House, 2006.
3. Kremer N.Sh., Prutko B.A. Econometrics / Textbook for high schools. - M .: UNITY-DANA, 2005.
© Panfilov I. A., Pen T. V., 2020
