CASPIAN JOURNAL
MANAGEMENT AND HIGH TECHNOLOGIES
Single- and multi-dimensional time series: an analysis of possible methodsfor optimizing readout and characteristic estimation
 |
Read | Brumshteyn Yuriy M., Ivanova Mariya V. Single- and multi-dimensional time series: an analysis of possible methodsfor optimizing readout and characteristic estimation // Caspian journal : management and high technologies. — 2012. — №4. — pp. 35-45. |
Brumshteyn Yuriy M. - Ph.D. (Engineering), Astrakhan State University, 20а Tatishchev St., Astrakhan, 414056, Russian Federation, brum2003@mail.ru
Ivanova Mariya V. - post-graduate student, Astrakhan State University, 20а Tatishchev St., Astrakhan, 414056, Russian Federation, rum2003@mail.ru
The article provides an analysis of possible methods for optimizing the readout and estimation of characteristics for the single- and multi-dimensional time series. It presents a series of possible classifications for the one-dimensional time series, paying special attention to models portraying an optimal choice of step-type behaviour for readout time. For the multi-dimensional series, the paper considers various possible indicators. Questions of data accuracy are analyzed for the single-dimensional time series, with models provided offering an optimum choice of accuracy at a fixed time step between readouts. The critique also analyzes joint choice models offering a step-on time and accuracy for the single-dimensional time series. Subsequently, the document considers different approaches to the selection of a variable time step in case of a change in the parameter’s speed for the series. In so doing, the commentary describes traditional approaches to the analysis of the single-dimensional time series on the basis of allocation of a trend, periodic components and casual rest points. Moreover, it shows variants of generalization of these approaches in case of unequal readout accuracy. The analysis also covers modern approaches on the basis of spectral characteristics of time numbers, including methods concerning the veivlet-analysis. Variants of other methods, such as the ’sliding window-type’ are considered in order to smooth out the time-number data. The blueprint also gives examples from the multi-dimensional time series for various spheres of human activity. For such numbers, it considers approaches such as the interrelation between such numbers in order to estimate their ’coherence.’ The study then proposes some modifications to known methods. Special attention is given to questions of estimating ’delay-advancing’ for the time series. In conclusion, the research work suggests perspective analytical directions and methods related to forecasting procedures for the multi-dimensional time series.
Key words: time series,one-dimensional,multi-dimensional,analysis methods,forecast methods,optimization of readout,unequal accuracy measurements,dynamic management,numerical methods