Brief scientific review

AR-CONTROLtm overview

Moscow 2015

Introduction

AR-CONTROL in brief

Autoregression used extensively in the finance industry to produce forecasts, all types

of AR- and ARMA (autoregression and moving average) methods.

These are usually based on fixed model’s order and usually not optimal.

"AR-CONTROL",at first, is unique because it guarantees to optimality of the forecast across these four dimensions:

Ø Consistency - ensuring good behaviour of algorithms from n-infinity

Ø Unbiased estimation - across any sample size averaging all possible realisations of the stochastic process, which leads to an estimation of the true parameters of the simulated process

Ø Effectiveness - Out of all possible forecasts, the model guarantees that this will be the most theoretically efficient

Ø Reaches the Rao-Cramer boundary

Additionally

· "AR-CONTROL" uses exclusive technique for the most difficult practical problem: choice of the best model.

· Moreover, "AR-CONTROL" runs automatically in full with regard to mathematics.

Brief scientific review

In many branches of science, engineering, economy and the finance, practically in all sections of natural sciences and the humanities there are phenomena, which are necessary for studying in time and space.

And almost always the case, by way of unexpected phenomena, random impulses, casual tests and random errors interferes.

At the description of a phenomenon with an uncertain outcome it is supposed that for the choice of result in such situation the nature as if would throw dice (the classical object in the probability theory).

The question of how such approach to the phenomena of the surrounding world is applicable is solved not by its logic substantiation, but by the results of practical application.

The concept of casual, its origin and ratio of casual and regular is often discussed. Regular phenomena are caused by the reasons that can be named, indicated and studied. If something changes, it happens because something else changed and the second is the reason for the first. And when changes occur at an invariance of environmental conditions, we explain it a contingency. There is a temptation to take into account all or nearly all the factors, influencing the process and thus to present a problem in a determined kind.

Let's consider further on that in the majority of practical problems it is hardly expedient and even impossible to take into account all the factors influencing the process. Therefore we shall consider the available data as the realization of a random process.

The examples of such time series are: in science - the results of supervising and experiments; in engineering (technology) - the results of the observation of the parameters of a technological process; in economics - daily and internal stock quotations, currency exchange rate, monthly volumes of purchases and sales of goods, annual volumes of production.

For economical and social data there are difficulties caused by the fact that usually there is only one short sample of data available, and they are stationary only during a small period of time, as they exist in the conditions of the fast changing economy and social sphere. Then the problem of forecasting is troubled by the fact that knowing a lot of far away significance of the time series depreciates and is of little help in predicting the future.

Further we shall not divide the data on seasonal, cyclic and random components, considering such approach a separable class of problems.

Дата добавления: 2015-10-26