ADAPTIVE INFORMATION TECHNOLOGY OF DECISION-MAKING SUPPORT FOR

TIME SERIES PREDICTION

Kvetniy R.N., Kotsubinskiy V.Y., Kislitsa L.N., Kazimirova N.V.

Vinnitsa National Technical University

Actuality

At the modern technique development level, when even a domestic technique is equipped with microprocessor devices, the intellectual adaptive control systems, able to adapt to the very wide range of external conditions, are very up-to-date [1]. A lot of the automatically guided technical systems, developed in ХХ-th century, are based on the theory of management, built on the deep analytical understanding of mechanics and physics laws. In practice attempts to get the exact model that describes its behavior are not possible due to lack of knowledge about an object and its environment. Also attempts to describe their properties analytically assume usage of very complex mathematical models [2].

Today there is a great number of the adaptive decision support systems (DSS), which are used in different areas. Among them it is possible to point out the next ones:

- «Karkas» is the computer system, based on knowledge of experts in main scientific area, that carries out and controls the process of students studying.

- «Delta» is the adaptive system of analysis and prediction, which can be applied in different areas, that solves tasks of authentication, monitoring and behavior prediction in the complex dynamic multipartite systems and manages them.

- "Eydos" is the universal cognitive analytical system to be used in any areas, that require tasks of authentication and prognostication of situations or states of complex objects according to external signs [3].

All these systems implement some combined algorithm:

1) automatic evaluation of environment state (situations);

2) viewing of possible alternatives in knowledge base;

3) choice of the "best" alternative according to some principle;

4) making of its realization command (or making of recommendations for its realization) [3].

Issues in creation of such systems are known; among them there is a lot of situations (dozens and hundreds) and necessity of forming of adequate great number of alternatives for them. And, as a result, empiric data processing with the use of hierarchical and networks structures requires creation of the proper algorithmic and mathematical instruments [3].

Meantime, authors represent simpler in creation, flexible DSS ‘Trade Keeper’ which provides more high-quality "prompt" while making a decision. The algorithm of managing decision making is changed depending on a current situation, so it is adaptive. Such system is considered in the given article.

Task purpose.

The work purpose is development of information technology of decision-making with adaptation to the user requests, that can be used for solving the tasks of authentication, monitoring and prognostication of the complex dynamic systems with a lot of parameters and managing them.

Basic part.

Usually DSS is created for certain tasks class and helps a person to accept a decision in the problem analysis. Person inquires necessary information, studies problems, gets advices from an expert system, and applies different mathematical methods, as well as knowledge of experts. DSS is developed as follows:

- uniting DSS with automated informative systems and communication networks;

- rapprochement of DSS with expert systems and development of «intellectual» SSPR;

- improvement of technological DSS base [4].

Basic preconditions to DSS creation are:

1) supposition, that finding of the best decision for some task is replaced by a sequence of the best decisions for a sequence of partial tasks that can be united in a general task;

2) supposition, that the initial order of single tasks performance can be changed depending on a situation while managing decision for current partial task is found;

3) supposition, that on the stage of decision of every partial task the number of alternatives is comparatively small and these alternatives are known;

4) supposition, that situations characteristics and their number is beforehand set while consideration of every partial task [5].

If these pre-conditions are correct, the process of making recommended managing decision by DSS assumes to be a set of the stages, which determines a procedure of finding common managing decision [4].

The expert decision support system may contain the blocks as follows:

1) an interaction block is an interface «user-system»;

2) a block of problems analysis and decision-making (logical conclusion) for letting user to apply expert knowledge while a user enters description of concrete situation in the system, and the mechanism of logical conclusion is provided by the search of expert knowledge, related to this situation;

3) database containing information about an object, problem structure, known cause-and-effect relations and so on.;

4) knowledge base containing expert knowledge, stored in a computer.

Using well-known rules and stages authors developed the expert system «Trade Keeper», which enables user to analyze the financial behavior of assets and, thus, facilitates a decision-making process so its efficiency is increased [6].

The algorithm of the system «Trade Keeper» functionality is presented at the figure 1. Now let consider every stage of system work in details.

On the first stage user needs to be registered and to work using a private login . With the own knowledge base user get a choice to work with individual decisions tree generated by system. At the figure 2 a new user registration form is presented.

Figure 2. A new user registration form

On the second stage user creates the required rules of decision-making and indicators which can be used for the building of decisions tree. In the special sections user can create new or edit existing rules and indicators.

On the next stage decision-making strategy is actually created. As a result of implementation of required steps user gets a strategy that is built on the basis of the created decisions tree. The given strategy helps to make decisions for further successful trading [6]. At the figure 3 an example of strategy, built by the strategies wizard is represented.

Figure 3. Example of strategy, built by the strategies wizard

User can get familiar with the details of the created strategy. This window (fig. 3) represents information about the created strategy and used rules, indicators and decisions tree.

On the fourth stage user selects financial asset and operation. System functionality allows to begin a trading. For this purpose user has to do the following actions:

- choose strategy among offered ones;

- choose type operation (Buy/Sell/Short) and enter the brief name of financial asset;

- choose a classifying rule;

- set the required indicators parameters;

- enter the asset name and its rate.

The result of this stage is information about the parameters of current trade.

On the fifth stage there a generation of decisions is done by the system. The system creates a message that predicts user probability of success of a trade. On this stage user needs to accept or cancel a transaction. If user decision is negative, user can begin all process from the second stage. If user decision is positive, a financial operation is executed and its results are added in to the knowledge base for future usage.

Advantages of developed DSS «Trade Keeper» are:

- system can reconstruct the process of the best decision search automatically;

- flexibility so system easily adapts to the user;

- simplicity in the usage and modification;

- increasing of efficiency of decision-making process;

- possibility to use knowledge of external sources and user ones.

Conclusions

Information technology of decision-making support was developed in the given work. It differs from the similar existing ones by original structure, simplicity in the usage and possibility of adaptation to the user and using of his knowledge in the subsequent decision-making process. Functional possibilities of technology are considerably increased in different areas, where decision-making is required.

Literature

1. Э.В. Попов, И.Б. Фоминых, Е.Б. Кисель, Статические и динамические экспертные системы. – М.: Финансы и статистика, 1996. – 176с.

2. B.G. Buchanan, D. Bobrow, R. Davis, E.M. Shorlife Knowledge – based systems. // Annu. Rep. Computer Science – 1990, №4, P.395–416.

3. Е.В. Луценко Универсальная автоматизированная система распознавания образов «Эйдос» (версия 4.1). – Краснодар: КЮИ МВД РФ, 1995. –76с.

4. О.И. Ларичев, Е.М. Мошкович Качественные методы принятия решений. Вербальный анализ решений. – М.: Наука. Физматлит, 1996. – 208с.

5. О. В. Балашов,  Е. М. Грубник, В. В. Круглов Cистема поддержки принятия решений с адаптацией алгоритма вывода // Электронный математический и медико-биологический журнал «Математическая морфология» – 2006, №1, С.12–18

6. Р.Н.Квєтний, В.Ю. Коцюбинський, Л.М. Кислиця, Н.В. Казимірова Адаптивна експертна система підтримки прийняття рішень в internet-трейдингу // Електронний науковий журнал «Інформаційні технології та комп`ютерна інженерія» – 2009, №2, С.1–6

7. Р.Н.Квєтний, В.Ю. Коцюбинський, О.М. Козачко, Л.М. Кислиця, Н.В. Казимірова Разработка способной к самообучению системы принятия решений // Електронний науковий журнал "Науковi працi ВНТУ" – 2008, №2, С.45– 52