Sosina K. O., Biloborodko O. I., Bondar O. E.

Oles Honchar Dnipropetrovsk National University

ALGORITHMIC AND TIME SERIES PREDICTION SOFTWARE

Time series is an ordered sequence which is generated as a result of observing a phenomenon that is changing in nature. These types of time series are widely used in various areas of human activity. In economics, they are daily stock prices, exchange rates, weekly and monthly sales, annual production, etc. Time series are not limited solely to economic values. Their use is known in the analysis of processes in electric power and nuclear industry, chemical and oil industries. In these cases, smaller time discreteness units are more commonly used than in economics, such as minutes and even seconds in data processing in the nuclear industry or the study of transient processes in chemical kinetics. Time series analysis in business also plays a very important role. Predicting monthly sales of goods is the basis of the policy inventory control, predicting future corporate earnings is the basis for decision-making in the investment policy [1].

The classification of time series is performed on time, the form of presentation levels and observing intervals. Any time series consists of two components: deterministic and random ones. There are two approaches to the analysis of the deterministic component: regressive and adaptive. When using adaptive methods of predicting, time series is represented as

(1).

In this case, the process of smoothing time series is a consistent calculation of the predicted value one step forward and the calculation of the deviations of the predicted values ​​from the values ​​of the original series. In order to predict time series, the models of linear growth (Theil-Veydzh, Box-Jenkins) and seasonal adaptive models are used. Seasonal variations may be presented in two forms:

multiplicative  , (2)

 additive  , (3)

where   – is a smooth trend component,   – are seasonal ingredients,   – is an error [2].

The above-mentioned algorithms have formed the basis of the developed software. The development environment is Delphi 7.0. The analysis of the results obtained from the software application has led us to the conclusion about the feasibility of using the adaptive techniques to build short-term predictions and its advantages in terms of implementation complexity and execution time.

The list of references:

1. Lukashin Y. P. Adaptive methods of short-term forecasting time series / Y. P. Lukashin. – M.: Finance and statistics, 2003. – 416 p.

2. Pristavka A. P. Statistical analysis in the ASTD: Time series / A. P. Pristavka, P. A. Pristavka, S. A. Smirnov. – D.: RIO DSU, 2000. – 112 p.