|
Article information
2020 , Volume 25, ¹ 5, p.80-90
Chirikhin K.S., Ryabko B.Y.
The application of artificial intelligence and data compression techniques for forecasting of social, economic and demographic indicators of the Novosibirsk region
This paper describes and experimentally investigates a time series forecasting method based on data compression and artificial intelligence techniques. Its basic idea is to combine various algorithms that can estimate the compressed size of a sequence of discrete values into a single method of forecasting. Generalizations of the method to continuous and multivariate cases are described. We use several popular data compression libraries (zlib, bzip2, ppmd), as well as applications of relatively lesser-known algorithms based on formal grammars (re-pair) along with our implementation of an algorithm based on finite automata with multiple heads. All these are employed to make forecasts for some social, economic and demographic indicators of the Novosibirsk region. The article elaborates both our methodology and the methods used for data preprocessing. Confidence intervals with a confidence level of 0.95 are plotted for all predicted values; the average relative errors calculated from the results of predicting the already fixed values are given. Cases in which multivariate forecasting turned out to be more accurate than univariate forecasting are: 1. Average annual population and natural population growth in the Novosibirsk region. Figure 1 shows their graphs along with multivariate predictions, while confidence intervals and mean relative errors are presented in tables 1, 2. 2. Life expectancy and number of deaths in the Novosibirsk region. Figure 2 shows their graphs along with multivariate predictions, confidence intervals and mean relative errors are presented in tables 3, 4. Table 5 gives univariate forecasts of some other indicators of the Novosibirsk region. We think that the results of computations show that the proposed method is capable of finding non-trivial patterns in the data and can be used in practice.
[full text]
Keywords: universal coding, multivariate time series, artificial intelligence
doi: 10.25743/ICT.2020.25.5.007
Author(s):
Chirikhin Konstantin Sergeevich
Position: Student
Office: Federal Research Center for Information and Computational Technologies, Novosibirsk State University
Address: 630090, Russia, Novosibirsk, Prospekt Akademika Lavrent'yeva, 6
E-mail: chirihin@gmail.com
Ryabko Boris Yakovlevich
Dr. , Professor
Position: Head of Laboratory
Office: Federal Research Center for Information and Computational Technologies, Novosibirsk State University
Address: 630090, Russia, Novosibirsk, Academician M.A. Lavrentiev avenue, 6
Phone Office: (383) 334-91-24
E-mail: boris@ryabko.net
SPIN-code: 5580-5794 References:
1. Reinsel G.C. Elements of multivariate time series analysis. New York: Springer Science & Business Media; 2003: 380.
2. Box G.E., Jenkins G.M., Reinsel G.C., Ljung G.M. Time series analysis: Forecasting and control. John Wiley & Sons; 2015: 712.
3. Kaastra I., Boyd M. Designing a neural network for forecasting financial and economic time series. Neurocomputing. 1996; (10):215–236.
4. Winters P.R. Forecasting sales by exponentially weighted moving averages. Management Science. 1960; 6(3):324–342.
5. Engle R.F., Kroner K.F. Multivariate simultaneous generalized ARCH. Econometric Theory. 1995; 11(1):122–150.
6. Ryabko B., Astola J., Malyutov M. Compression-based methods of statistical analysis and prediction of time series. Switzerland: Springer International Publishing; 2016: 144.
7. Ryabko B.Ya. Prediction of random sequences and universal coding. Problems of Information Transmission. 1988; 24(2):87–96.
8. Chirikhin K.S., Ryabko B.Ya. Experimental study of the accuracy of compression-based forecasting methods. Vestnik NSU. Series: Information Technologies. 2018; 16(3):145–158. (In Russ.)
9. Tim S. Prediction of infinite words with automata. Theory of Computing Systems. 2018; (62):653–681.
10. Chirikhin K.S., Ryabko B.Ya. Application of artificial intelligence and data compression methods to time series forecasting. Proceedings of the International Workshop “Applied Methods of Statistical Analysis. Statistical Computation and Simulation — AMSA’2019”. Novosibirsk. 2019: 553–560.
Bibliography link:
Chirikhin K.S., Ryabko B.Y. The application of artificial intelligence and data compression techniques for forecasting of social, economic and demographic indicators of the Novosibirsk region // Computational technologies. 2020. V. 25. ¹ 5. P. 80-90
|
