CC BY
49
YflK 519.95
TECHNOLOGY OF INTELLECTUAL INFORMATION SYSTEM DESIGN FOR ESTIMATION OF SOCIAL OBJECTS
© A.A. Arzamastsev, N.A. Zenkova, K.G. Troitzsch, A.V. Neudakhin
Key words: automatic expert system; developing intellectual core; neuron-net models; mathematical formalization of tasks of information accumulation.
The automatic technology of information system design of expert systems containing the functional advantages of the developing intellectual core and network technologies of interaction is discussed. The results of mathematical task formalization of information accumulation and design of developing intellectual core on its base are given.
The estimation of various social objects is necessary in different fields of social life, science, education, etc. They are usually used for forecasting, evaluation of different properties and other goals in complex man-machine systems.
At present this estimation is possible by means of computer and mathematical simulation methods which is connected with significant difficulties, such as:
- time-distributed process of receiving information about the object;
- determination of a corresponding mathematical device and structure identification of the mathematical model;
- approximation of the mathematical model to real data, generalization and parametric identification of the mathematical model;
- identification of the structure of the links of the real social object.
The solution of these problems is impossible without a special intellectual information system which combines different processes and allows to predict the behaviour of such an object. However, most part of existing information systems lead to the solution of only one special problem. From this point of view the development of the general technology of designing such systems is very important.
The technology of intellectual information system development for estimation and forecasting the professional ability of respondents in the sphere of education can be a concrete example of such a technology.
Job orientation is necessary and topical in present economic conditions. It helps to solve the problem of expediency of investments to a certain sphere of education.
Scientifically validated combined diagnostic methods of job orientation are necessary to carry out professional selection in higher education establishments. The requirements of a modern society are growing, with the earlier developed techniques being unable to correspond to them sufficiently.
All these techniques lack an opportunity to account all necessary professional and personal characteristics. Therefore, it is necessary to use a system of various tests.
Thus, the development of new methods of job orientation for entrants is necessary. The information model of the process of job orientation is necessary for this purpose.
Therefore, it would be desirable to have an information system capable of giving recommendations concerning the choice of a trade on the basis of complex personal characteristics of entrants.
Concept of intellectual information systems building
A concept of building intellectual information systems includes the following main components: an information part and a module which works with artificial neural networks (ANN).
The purpose of the information part consists in the accumulation, storage and provision of the information about the object, and also maintenance of the interface for the end user.
The purpose of the module working with ANN is the construction of the artificial neural network and its training on the sets of input variables of the object with their corresponding output conditions. An ANN is capable of classifying new information and, on its basis, of being trained further. The result of the work of this component is the intellectual model of the object.
Information part of the intellectual information system (IIS)
The information part of the system represents a set of PHP-scripts co-operating with a MySQL database. This component functions according to the scheme presented on fig. 1.
Figure 2 contains a more detailed description of the processes in the information part of the IIS.
Meaning of the links presented in fig. 2:
1) The user sends the request to the HTTP-server, containing some data. Any Internet browser (Internet Explorer, Mozilla Firefox, Opera, etc.) can be used for connecting to the HTTP-server as the client of the system;
2) HTTP-server translates PHP-scripts, acting in accordance with the user's request;
3) PHP-scripts refer to a database with request;
4) request results are sent to the PHP-scripts to process;
5) the result of processing is a generated HTML-page;
6) the HTTP-server sends the user the response which is a generated HTML-page which is displayed in the user’s Web-browser.
This component supports three levels of access - the user, an expert and an administrator. A specific set of authorities and functional capabilities corresponds to each level:
a. administrator-level authorities control the user management;
b. the expert level is empowered to manage the object;
c. the user level allows to enter information about the parameters of the object and to get a result on the basis of entered data.
The specific problem under consideration in a given subject area is called an object in the intellectual information system.
The object of the system is defined by a set of input variables with the corresponding output conditions. Creating an object is possible for the expert and associated with the determination of such characteristics as:
- the number of input parameters;
- input parameters that will be used for later analysis;
- proposed output conditions.
The accumulation of information about the object can be done in two ways.
The first way (with user and expert). In this case, the user is registered in the system, introduces the variable values of the object. These data are stored in the database. Then, the expert determines the appropriate output condition of the object by analysing a set of input data that is previously entered by the user. Thus, the knowledge is stored in the database and forms the base of the knowledge about the object. The information which is stored in the knowledge base generates the classes defined in accordance with the output conditions of the object.
The second way (without user). In this case, the expert, already having sets of input data about the object, can load them into the system. The set can already be analysed or evaluated by the expert subsequently.
Ґ
Expert
Internet browser (Internet Explorer, Mozilla Firefox, Opera,...)________________
HTTP-server
PHP-scripts
Fig. 1. The scheme of functioning of the information part of the system
Internet browser (Internet Explorer, Mozilla Firefox, Opera, ...)
m
[Re
Request |
The server response
HTML
W 1 HTTP Server Apache
[2l J 151
Data request html-page
i
PHP- scripts
Fig. 2. The mechanism of functioning of the information part of the IIS
Intellectual subsystem of the intellectual information system
The intellectual subsystem is intended to implement the processes of constructing, training, modification and operation of ANN models. ANN models are autonomous systems of knowledge representation in the IIS (Fig. 3).
This subsystem includes basic components:
1) block of data monitoring, which provides transfer of input and output data. The monitoring block tracks the incoming flow of new information into the database;
2) converter, preparing the training set for the network;
3) constructor of the ANN, providing automated creation and training of the intellectual model (ANN) of the object.
This intellectual information system includes the basic distinctive components, such as: web-interface, the database used for long-term storage of the technical information, the intellectual core including the developing, ANN model based system of the experts' knowledge representation. The intellectual core of the information system will consist of ANN models, i.e. models constructed on the basis of artificial neural networks.
Figure 5 shows the functioning of the intellectual information system (IIS).
The mechanism of interaction between the components of the integrated system is presented in fig. 4.
The top part of the fig. 5 presents the dependence of the primary information accumulation over time (in arbitrary time units), R is the number of records in the database received from the terminals. The bottom part of fig. 5 presents the process of training the ANN model, E is the reduced mean-square error of the ANN model, and L shows the levels of training of the ANN model.
E(t) = E* (i - e-kt)+ E0e-kt,
where
- E(t), E0 and E* are the values of the reduced mean-square error, its initial value and the level at which it will be fixed at the end of each training cycle of the ANN model;
- k is the parameter of the specific training rate which depends on n, the capacity size of the training set, using the method of residual minimisation;
- t - time in arbitrary units.
The learning process of the ANN model (the intellectual core of the IIS) depends on the number and quality of data in the training set.
Database
f " Block of data
monitoring Converter
X
Intellectual model (ANN)
Constructor of the ANN
Fig. 3. The scheme of functioning of the intellectual part of the system
Information subsystem
Fig. 4. Structural diagram of the intellectual information system
t\ h h >a h h, t
Fig. 5. Functioning of the intellectual information system (IIS)
It is necessary to have a lot of empirical data for the identification of the complex structure of the object. While new data come in, the expert will complicate the structure of the ANN model to achieve an acceptable level of adequacy. Therefore, this is a multistage training process. When level L5 (an acceptable level of the model training) will be reached, then the expert will finish the training of the model and switch the IIS into the operation mode.
The expert has the following rights: to determine the acceptable level of the error of the model, to complete the process of training the model, to determine the output conditions of the object corresponding to the input data (if the training set does not contain output parameters).
The mathematical formalization of the processes of information accumulation and building the developing intellectual core on its base was described earlier in [1]. It allows to solve the following problems:
- the estimation of the primary period of information accumulation, which is necessary for building ANN model of the research object;
- the determination of the primary structure of the ANN model on the basis of input characteristics of the research object;
- the estimation of the necessary number of training cycles of the intellectual core by the expert.
The technology of the design of intellectual information systems on the basis of an ANN model and with a distributed input of the data includes the following steps:
- study of the research object (for developing IIS); determination of the input variables and the possible output states of the object;
- accumulation of primary information about the object in the database through a distributed system of data collection;
- generation of the primary structure of the ANN model;
- planning of the step-by-step process of the developing ANN model; training of the ANN model;
- development of the ANN model on the basis of new empirical data coming in; this step needs the control of an expert who can determine when the formation of the knowledge representation will be finished;
- operation of the intellectual information system (without expert).
Application the technology of intellectual information system design for the estimation of social objects
The social object can be represented by investigation results of students of Tambov State University named after G.R. Derzhavin (TSU). For the construction of the intellectual core of the information system we can use the empirical data received earlier by the staff of the Computer and Mathematical Simulation Department at TSU. These empirical data come from questioning high school students in Tambov [2].
These students answered questions of a questionnaire which consisted of the following sections: statistical information; progress; professional potential; public activity; creativity; leisure; health; family.
The results of the questionnaire will be both input and output variables (which correspond to database fields) on the basis of which the training of the artificial neural network will be done.
For the artificial neural network we can use the software which we developed during our previous activity [3]. This program is a universal neural network simulator using an original algorithm of self-organization of neural network structure which allows to achieve good flexibility and adaptability of the model to empirical data. The algorithm is described in [3].
This technology of intellectual information system design can be used for simulation of other various social, medical and other objects as well.
The advantages of this technology are: universality for a variety of subject areas, multi-user access to the same object of research, developing an intellectual core on the basis of ANN models, and support of the networking technology of interaction.
REFERENCES
1. Arzamastsev A.A., Zenkova N.A., Neudakhin A. V. The formalization of the problem of expert information system building with the developing intellectual core on the bases of ANN models // TSU Reports. 2010. V. 15. Issue 1. P. 287-290 (In Russian).
2. Arzamastsev A.A., Zenkova N.A. Sociological investigation of the students of the Tambov State University named after G.R. Derzhavin. Tambov State University Publisher House. 2005. 65 p. (In Russian).
3. Arzamastsev A.A., Zenkova N.A. etc. Realization of TEMPUS TACIS project “System modernization of University Management ” at Tambov State University // Information Technologies in University Management. Int. Conf. Tambov, October 17-20, 2006. Tambov State University Publisher House, 2006. Р. 9-44. (In Russian).
ACKNOWLEDGMENTS. This work was supported by the DAAD program “Michail Lomonosov” (B) (grant # А0974042, 2010/2011).
Поступила в редакцию 22 ноября 2010 г.
Арзамасцев А.А., Зенкова Н.А., Троич К.Г., Неудахин А.В. Технология дизайна интеллектуальной информационной системы для вычисления социальных объектов
Обсуждается автоматизированная технология построения экспертных систем, сочетающих в себе функциональные преимущества развивающегося интеллектуального ядра и сетевых технологий взаимодействия. Приведены результаты математической формализации задач накопления информации и построения на ее основе развивающегося интеллектуального ядра.
Ключевые слова: автоматизированная экспертная система, развивающееся интеллектуальное ядро, нейросетевые модели, математическая формализация задач накопления информации.
