Faculty: The Faculty of Computer Information Technology and Automation

Speciality: Automatic Control Systems

Theme of master's work: "The infectious incidence forecasting in the region"

Supervisor of work:Orlov Juriy Konstantinovich

Forecasting of a level of infectious is necessary for a substantiation of system of preventive actions.Depending on a level of scientific development of a problem and opportunities of realization of recommendations in concrete conditions of the purpose of carrying out of preventive actions may have over a wide range – from stabilization infections even at rather high level to its steady decrease and liquidation of separate infections in observed territory or among served contingents of the population.

The decision of a problem of development of preventive actions is especially important in our region, considering its specificity: high concentration of the industrial enterprises, the complex ecological conditions negatively affecting on health of the population.

The state sanitary-and-epidemiologic stations at all levels are engaged in development of a plan of infections preventive measures.

It is operative information the registration and processing, processing of the statistical data received during the certain period, decision-making on methods of preventive maintenance for the future period.

The significance of dataware increased in the most different medical technologies. It becomes the critical factor of development practically in all areas of knowledge. Therefore development and implement of information systems is for today of one of the most actual problems.

The theme of work is very important, as first of all it’s related to the health of the person.

The effective systems based on the mathematical methods of infections forecasting are not introduced today in our region.

Factors which influence on character of the tendency of infections are: quality and efficiency of preventive maintenance, activity of the activator circulation, long-term periodic fluctuation of infections level.

Periodicity is a natural form of epidemic process phase development. It is influenced with a set of natural and social conditions. Up to the middle of 70th years of the last century the epidemic cycles durations were 4, 6, 9, 13 and 19 years. But during the subsequent period there were significant changes of epidemics periodicity and since 1975 a two-year-old cycle was established. Two-year periodicity of infections corresponds to alternation of the "hot" and "cold" summer periods, i.e. it reflects a rhythm of climatic conditions. The analysis of infections periodic fluctuations is necessary for the several problems solving. Periodic components stability and fluctuations amplitude are additional criteria of epidemic process controllability. In a combination with a theoretical line of the tendency the periodic components estimation may be used for infections forecasting for forthcoming years.

The input data for the analysis of long-term dynamics are formed of the operative accounting information of infections after their correction taking into account final diagnoses.

The incidental outbursts of infections with established reasons are excluded from data. The minimal duration of the studied period in 10 years is taken into account at the analysis of long-term dynamics.

It is possible to allocate two ways in the decision of a problem of forecasting: the forecasting based on the models of time series and on the neural networks.

A number of supervision of the analyzed random variable made during the consecutive moments of time , make a time series. Elements of time series are not statistically independent and not equally distributed. A time series is influenced by following factors:

- long-term, forming the general tendency;

- seasonal, forming periodically repeating in the certain season of fluctuation of an analyzed attribute;

- cyclic: demographic, economic, astrophysical and other cycles;

- casual which cannot be registrated.

It is necessary to consider following conditions at forecasting on the basis of the statistical analysis of time series:

- demanded horizon l of forecasting, i.e. on how many time steps (l) forward we are going to build our forecast;

- length analyzed time series;

- presence at an analyzed time series of a seasonal component or any "non-standards" .

The second way – use of neural networks. Development neural methods enables their uses as tool of scientific researches by means of which it is possible to study objects and the phenomena.

The basis of the work of self-trained neural programs is the neural network representing the set of neurons - the simple elements connected with each other. The structure of connections between neurons is similar in biological objects.

Neural network must have channels for its connection with an external world. Some channels provide the receipt of the information from the external world, others deduce the information from the neural network. Some of the neurons can not connected with an external world. It is obvious, that there is a huge quantity of the neurons connection ways. The most widespread architecture is the layered one.

The neural networks using at forecasting is the effective and modern decision of the problem. This method has following advantages:

1. Neural networks make decision on the basis of the experiment which was received by them independently. The developer of system does not need to establish connection between the entrance data and the necessary decision. It is not required to spend time on various statistical processing, selection of the mathematical methods,the creation and the checking of mathematical models.

2. The decision which is accepted by the neural network, is not categorical one.

3. The neural network allows to model a decision-making situation of.

4. Neural network find the answer very quickly (shares of second). It allows to use them in various dynamic systems which demand immediate decision-making.

5. Neural networks allow to simplify the expert systems creation process, to define scientific search directions.

However, the neural network represents a single whole, and addition of new neuron to a network leads to the necessity to retrain a network completely .

Considering merits and demerits of the problem decision ways, the most effective way is to use the neural networks.

The further research purpose - gathering the information that is necessary for formation of computerized system, which basis in the neural network.

As a result it is necessary to receive system which could solve the roblem effectively. Thus it should have practical realization at the enterprise which solves a problem of infections forecasting.