კონფერენცია

∘ Merab Pkhovelisvili ∘ Mariam Giorgobiani ∘ Natela Archvadze ∘ Gaioz Pkhovelishvili ∘
MODERN FORECASTING MODELS IN ECONOMY

Summary

A new approach for business forecasting is discussed in the article. This means usage of parallel data paradigm of programming. Parallel data are different kind of former data, which give us chance to predict an event in dynamic mode. Also, functioning of forecasting process online is being discussed. This method helps us to use super computers not only for original purpose - calculation big amounts, but for processing parallel data online.

Supercomputers are computers with a high level of performance, which are used to work with those applications, which require more intense computations. Creation of supercomputers, widespread adoption of cluster systems, connection of computers with each other by means of local and global networks, caused attraction of users to computation process and made possible to perform various tasks. This includes a tasks of economic forecast, simulation of nuclear test, etc. It can be said that supercomputers bring new possibilities for automation of prediction process.

Business-forecasting tasks may include: demand, intermittent demand, time and space hierarchies, shares, macroeconomic indicators, commodity groups, new products and more. Consider two tasks: predicting demand and interruption request. That's the solution of these two tasks will be discussed in the example when "parallel data".

In this article we consider, at the one hand, the models of new type of predictable processes (called conditional, temporal, expandable matrix of vectors) and, on the other hand, an new paradigm of programming (called parading of parallel data) for processing of these models.

For the computer presentation of mathematical models of prediction processes, we will use so-called conditional temporal expandable matrices of vectors (hereinafter – the expandable matrix). For the computer presentation of mathematical models of prediction processes, we will use so-called conditional temporal expandable matrices of vectors (hereinafter – the expandable matrix).

Each event impacting an event to be predicted, is represented in the form of separate vector of data. The word “conditional” in the name of matrix means that it is now known in advance, how many events impact an event to be predicted (i.e. how many vectors of data are contained in the matrix). The matrix is dynamical, therefore, some events (and respective vectors) are erased from it, some are added and some are moved into another place. The word “temporal” is used in the name of matrix because the number of row depends on the time. These vectors are shown on the figure by directions of arrows, which are directed from the bottom upwards. In each vector, the data is arranged according to the time

The matrix, corresponding to the model of predictable processes, as it was noted, is dynamically variable, which means a change of its sizes. Thus, the word “expandable” is its name. The matrix constantly expanses upwards, new data is added to it and the number of columns varies, the data is added or taken away according to the event function, which will be considered below. It is also possible that a matrix is expanded downwards – the data is added, corresponding to old, already occurred events or archive data.

Each expandable matrix for each business forecasting task may describe the one territorial region, therefore, for the certain task of prediction, more than one expandable matrix can exist, which will be built for certain period of time.

Ivane Javakhishvili Tbilisi State University Paata Gugushvili Institute of Economics International Scientific C O N F E R E N C E S "ECONOMY – XXI CENTURY"

E ISSN 2587 - 4691	∘ Merab Pkhovelisvili ∘ Mariam Giorgobiani ∘ Natela Archvadze ∘ Gaioz Pkhovelishvili ∘ MODERN FORECASTING MODELS IN ECONOMY Summary A new approach for business forecasting is discussed in the article. This means usage of parallel data paradigm of programming. Parallel data are different kind of former data, which give us chance to predict an event in dynamic mode. Also, functioning of forecasting process online is being discussed. This method helps us to use super computers not only for original purpose - calculation big amounts, but for processing parallel data online. Supercomputers are computers with a high level of performance, which are used to work with those applications, which require more intense computations. Creation of supercomputers, widespread adoption of cluster systems, connection of computers with each other by means of local and global networks, caused attraction of users to computation process and made possible to perform various tasks. This includes a tasks of economic forecast, simulation of nuclear test, etc. It can be said that supercomputers bring new possibilities for automation of prediction process. Business-forecasting tasks may include: demand, intermittent demand, time and space hierarchies, shares, macroeconomic indicators, commodity groups, new products and more. Consider two tasks: predicting demand and interruption request. That's the solution of these two tasks will be discussed in the example when "parallel data". In this article we consider, at the one hand, the models of new type of predictable processes (called conditional, temporal, expandable matrix of vectors) and, on the other hand, an new paradigm of programming (called parading of parallel data) for processing of these models. For the computer presentation of mathematical models of prediction processes, we will use so-called conditional temporal expandable matrices of vectors (hereinafter – the expandable matrix). For the computer presentation of mathematical models of prediction processes, we will use so-called conditional temporal expandable matrices of vectors (hereinafter – the expandable matrix). Each event impacting an event to be predicted, is represented in the form of separate vector of data. The word “conditional” in the name of matrix means that it is now known in advance, how many events impact an event to be predicted (i.e. how many vectors of data are contained in the matrix). The matrix is dynamical, therefore, some events (and respective vectors) are erased from it, some are added and some are moved into another place. The word “temporal” is used in the name of matrix because the number of row depends on the time. These vectors are shown on the figure by directions of arrows, which are directed from the bottom upwards. In each vector, the data is arranged according to the time The matrix, corresponding to the model of predictable processes, as it was noted, is dynamically variable, which means a change of its sizes. Thus, the word “expandable” is its name. The matrix constantly expanses upwards, new data is added to it and the number of columns varies, the data is added or taken away according to the event function, which will be considered below. It is also possible that a matrix is expanded downwards – the data is added, corresponding to old, already occurred events or archive data. Each expandable matrix for each business forecasting task may describe the one territorial region, therefore, for the certain task of prediction, more than one expandable matrix can exist, which will be built for certain period of time.