Content

1. Introduction
2. Theme's actuality
3. Work communication with scientific programs, plans and themes
4. Aim and research tasks
5. Probable scientific novelty
6. Practical importance of work
7. Review of research and development
   • The regression cognitive modelin
   • The simulation modeling method
   • The neural networks
   • Fuzzy logic application
   • Genetic algorithms application
8. The basic content of master’s work
9. Conclusions
10. List of used literature

 

1. Introduction

2. Theme’s actuality

5. Probable scientific novelty

6. Practical importance of work

• we investigated and systematized theoretical information about statistics;
• we chose and analyzed classical prediction methods;
• we investigated modern prediction methods based on neural networks;
• we investigated existing program packet Matlab Simulink;
• the practical importance of work consists in program development of telecommunication business profit prediction based on billing system data

7. Review of research and development

The regression cognitive modeling
The aim of cognitive modeling consists of hypothesis generation and checking about functional structure of under review situation which can explain system’s behavior.

Taking in account these factors we create a model of the researched situation as the hypothesis aggregate which is able to explain the system development. Besides hypothesis that have been created which are able to explain the influence mechanism between different system’s factors or set the cause-and-effect relation between them.

For the communication factors interference analysis and its forecasting it is offered to use the regression cognitive modeling methodology which completes cognitive situation graph with regression analysis mechanisms [8].

The general target setting of venture management in the context of RCM related to special top-factor apportionment and its value is desirable to increase, decrease or “insert” in the definite values range. But the RCG structure is not changed and the search task consists of coefficients (graph arches) assortment when the wishful aim is achieved.

The search solution methodology in combination with the RCM first of all is useful for decide variant modeling, management tactics or strategy. A constructed model of search solution with RCM models allows finding an optimal solution of transferring from a real value of any factor to wishful one, which can be used during the reaching an appropriate decision.

 

The simulation modeling method

Today the control modeling of business process becomes not only process of collection, analysis and enormous information file handling to discover an effective management script, but also forming process and problem analysis.

The simulation modeling is a rational decision making method which helps to form the statistics of possible outlet activity parameters.

Application of this method lets use the billing system for proving management decisions in Communication Company, for elaborating tariff scale, for planning and fixing a budget; gives an opportunity to analyze the situation of subscribers’ enrollment, services introduction, and changes of the call routing scheme [9].

Before the introduction such a tariff scale it is necessary to investigate different tariffs, select the right ones and achieve the optimal load distribution. The fusion of a simulation model and the billing system allows realizing the payment task between operator and clients.

Such a model is used for elaborating tariff scale planning module. Because of provided services investigation it is possible to form tariff scales for subscribers and existing operators. The outlet data is used for proving management decisions about the determination of optimal tariff politics and the business’ development strategy [10].

 

The neural networks

An artificial neural network is a set of neurons connected to each other. An artificial neuron is similar to biological neuron because of its properties.

An aim of prediction is risk decreasing during the decision-making. In most cases forecast is erroneous and an error depends on forecasting system and prediction methods. To decrease an error there must be prognosis recourses increased. Under such a level of error we can achieve minimum of prognosis recourses level [11].

The prognosis result using the neural network is a class, which consists a variable not its concrete value. The class forming depends on prediction aims. The general method consists in the class fragmentation of determination area with necessary accuracy. Classes can represent qualitative or numeral view on variable change.

The neural networks prediction is possessed of several disadvantages. As a rule we need about 100 values for acceptable model creation. It is quite a big data number and there are a lot of cases when such a historical data number is inaccessible.
But it is necessary to say that it is possible to make satisfactory neural networks model even if there are no data enough. Model can be detailed while new data become available [12].

The simple neural network is perceptron, which is on the figure 7.1

Figure 7.1 - Perceptron – is the simple neural network

We use multi-ply net to solve more complicated problem. This net is on the figure 7.2

Figure 7.2 - The scheme of multi-ply net
(animation: volume – 27,4 KB; size – 791x571; amount of shots – 4; frequency of shots changing – 2000 ms; amount of repeated cycles – a continuous cycle of reiteration) 

 

 

Fuzzy logic application

The material flow prediction of industrial undertaking is proposed to realize using fuzzy logic, fuzzy knowledge and genetic algorithms.

Fuzzy quantities allow formalizing of values with qualitative basis, revealing cause-effect relations between operated parameters and variables influencing on it, and formulate fuzzy forecast in conditions of prediction parameters uncertainty.

The accurate material flow prognosis of modern undertakings is difficult because of objective and subjective causes. It is appropriate to make material flow prognosis using fuzzy logic methods and fuzzy knowledge. We set a problem, structure a process and elaborate mathematics model of the material flow prediction of industrial undertaking using fuzzy knowledge base.

Fuzzy numbers getting as a result of “not quite accurate dimensions” in many cases is similar to the theory of probabilities distribution, but in comparison with probabilistic methods fuzzy logic methods allow abruptly reduce calculations size that in its turn leads to quick-action fuzzy systems increasing.

 

Genetic algorithms application

Genetic algorithms allows accomplishing tasks of prediction, classification, optimal variants searching and is completely indispensable in that cases when in usual conditions a problem solving based not on strict description (in mathematics meaning), but on intuition or experience.

It is naturally using genetic evolution mechanisms for neural networks studying because neural networks models are elaborated in much the same way as brain and realize some of its particular qualities, which appear as a result of evolution.
An advantage is effectiveness of global minimum searching of adaptive relieves since large regions of acceptability of neural networks parameters are investigated in it [13].

Genetic algorithms allow operating by discrete values of neural networks parameters. It simplifies elaborating of numerical hardware-controlled neural networks realization. During the neural networks studying on computer not oriented on hardware-controlled realization the possibility of discrete values using in some cases can lead to total training time reduce [14].

8. Basic content of master’s work

First part. A review of problem: business profit prediction based on billing system data. In this part are described methods and mathematical organization of task? and also the prediction parameters.

Mathematical organization of task
Profit is the fundamental factor for economic and social business development. Therefore the valid income forecasting is of great importance for business.

For the solution of this problem we take so-called gross revenue, i.e. without tax takeout, wage costs and constant or non-permanent costs. In our case we sum up all the resources matriculated on account current of the enterprise for the consumed connection services. Further this amount we call “profit”. Thus the profit calculation formula for the Communication Company is:

(8.1)

Where ki – a number of services in i–tariff;
n – a number of tariffs;
mi – a number of subscribers, using  tariff i;
xijl – a number of services received by a subscriber j in a tariff i for a service l;
til – a service cost l in tariff i.

The task of the profit prediction consists mainly of forecasting some parameters with help of which the profit is calculated. As long as ki – a number of services in i–tariff, n – a number of tariffs and til – a service cost l in tariff i we determine as we wish, so we consider these parameters are known. So our task comes to mi and xijl determining (predicting).

From the one hand we have a number of subscribers mi using tariff i. Offering new services or new tariffs of existing services we have to predict how many new subscribers will be attracted, how many subscribers change the tariff and how many clients will use new services with new tariff.    

On the other hand        we have a number of services xijl received by a subscriber j in a tariff i for a service l. In fact we have to predict how much connection services (existing services with new tariff or service innovation) will be used by every subscriber.

It is obvious that the small bulk of data much easier to process analytically. Hence we can make a conclusion that it is desirable to divide an enormous data file contained in the billing system for example into groups of subscribers inside the tariff (for example by the amount of consumed services).

Depending on the billing system data we can determine the loyalty of different subscribers groups in the first place; we can predict their behavior, services and tariff preference, using duration of provided services, possibility to change tariff scale and so on.

We need several methods to accomplish this task. In order to predict changes of the number of subscribers we require investigating of causative models (which use connection between interesting for us time sequence and one or more other time sequences) using the theory of fuzzy quantities.

And in order to forecast the amount of services consumed by subscribers it is necessary to analyze time sequences (it is a sequence of variables’ value ordered in time) using neural networks.

So we need such a model which can solve the problem of business profit rising and the net loading alignment, we have to elaborate the tariff scale planning modules.  The introduction of such forecasting system allows supplying connection with new possibilities, formed a new services list that influents on getting bigger profit.

Second part. The prediction method development.

Third part. The development of model of checking prediction methods.

Fourth part. The development of informational and program supplying for business profit prediction system.

 

9. Conclusions

There is an inner corporative statistic system, which includes financial and marketing accounts. That comprehensive statistics formed by particular parameters regularly or as may be necessary. Proposals clients are formed on the base of this statistics, as there are a lot can be modeling and predicting with that accounts’ help. This is more likely not owned by a specified person approach, but in the same time there is a possibility to work more personal with clients.

Using billing data collaborators of special marketing department can choose one particular client or a group of subscribers united by some parameter for offering them less popular product. This allows company to solve two problems. First of all, in order to forecast consume profile of new service by the more numerous audiences, limited number of persons is offered to use it during some period of time. In this case they are simply observed to understand how the new proposal works. (They watch how many subscribers agree to connect, how service consuming changes)

Besides subscribers base processing enables to check marketing suppositions about dependence between connection consuming profile and additional services, which could be interesting to clients. Thus a company can offer the service outright to subscriber directed particularly toward him or quite narrow segment of users.

10. List of used literature

1. Биллинг неголосовых услуг [Электронный ресурс] / А. Голышко. – Режим доступа: www.connect.ru/article.asp?id=6170
2. Современные CRM/PRM-системы и web-технологии в бизнесе телекоммуникационных компаний [Электронный ресурс] / А. Гургенидзе. – Режим доступа: www.connect.ru/article.asp?id=6617  
3. Naumen Telecom: Комплекс OSS/BSS для российских компаний (ТССС). [Электронный ресурс] – Режим доступа: www.naumen.ru/go/company/press/TCCC_08_Telecom
4. Трубникова, Е.И. Способы управления социально-экономическими системами / Т.В. Полулях, Е.И. Трубникова // Инфокоммуникационные технологии, 2007, том 5, № 1, с. 72-77.
5. Трубникова, Е.И., Биллинг как средство принятия управленческих решений при формировании тарифных планов телекоммуникационной компании. / Е.И. Трубникова, А.В. Добрянин // Проблемы Материалы XIII Юбилейной Российской научной конференции профессорско-преподавательского состава, научных работников и аспирантов, 30 янв.– 4 фев. 2006г. – Самара: Поволжская Государственная Академия Телекоммуникаций и Информатики, 2006, с. 208-209.
6. Zachman J. A Framework for Information System Architecture // IBM System Journal, 1987, vol. 26, № 3, pp. 276-292.
7. Zachman J. Enterprise Architecture: The Past and The Future // DM Direct, April 2000.
8. Разработка моделей выявления взаимозависимых факторов в телекоммуникационном графике на основе регрессионно-когнитивных графов диссертация кандидата технических наук [Электронный ресурс] / А.В. Мелик-Шахназаров. – Режим доступа: http://www.dissforall.com/_catalog/t8/_science/39/210263.html
9. Инновационное управление телекоммуникационной организацией на основе метода имитационного моделирования [Электронный ресурс] / Е.И. Трубникова. – Режим доступа:   http://www.nimb.nnov.ru/_data/files/ARD_Trubnikova.pdf   
10. Трубникова, Е.И. Применение имитационного моделирования в процессе бюджетирования. / Е.И. Трубникова // Роль высших учебных заведений в инновационном развитии экономики регионов: Международная науч.-практ. конф., 10-12 окт. 2006г. – Самара: Самар. гос. экон. акад., 2006. – с. 217-220.
11. Бутенко А.А. и др. Обучение нейронной сети при помощи алгоритма фильтра Калмана. // Труды VIII Всероссийской конференции «Нейрокомпьютеры и их применение »: Сб. докл., 2002. – с. 1120-1125.
12. Решение задач прогнозирования с помощью нейронных сетей [Электронный ресурс] / Акулов П. В. – Режим доступа:  http://www.masters.donntu.ru/2006/fvti/akulov/diss/index.htm 
13. Вороновский Г.К., и др. Генетические алгоритмы, нейронные сети и проблемы виртуальной реальности. – Х.: ОСНОВА, 1997. – 112 с.
14. Батищев Д.И. Генетические алгоритмы решения экстремальных задач. – Воронеж: ВГУ, 1994. – 135 с.

 

Important remark: during the writing this abstract the master's work wasn’t finished yet. The final ending is planning on December, 2009. The complete text of work and all the information about work you can receive from the author or her scientific adviser after the mentioned date.

Top