Современные информационные технологии/3. Программное обеспечение

Жамбаева А.К.

Костанайский государственный университет им. А. Байтурсынова, Казахстан

Informational Technologies predicting of market economy

Today for a modern trader opened up broad prospects, due to the development of information technology. So, if the first investors had to calculate each formula manually, and each graph draw on graph paper. Now, with a development of computer technology in the hands of investors have an opportunity to facilitate trade, and make it more efficient and profitable. Scientific researches to improve methods of forecasting and analysis of data are conducted through the integration of fuzzy sets, chaos theory, and as a tool for implementation forecasts - neural network.

Use of neural networks is one of the most famous tools on the market today. Neural networks are very diverse in their architecture, but they all have one common base element - an artificial neuron, which mimics the properties of a biological neuron. One of the most important properties of a neural network - its ability to self-organization, self-adaptation for improving the quality of functioning. This is achieved through the network training - an algorithm given by a set of rules. Educational rules determine how relations changed in response to the input action. There are two classes of teaching methods: deterministic and stochastic.

Deterministic method of teaching performs correction procedure scales network step by step, based on the use of their current values, as well as input values, actual outputs and desired outputs.

Stochastic methods of teaching perform pseudorandom changes values scales, keeping the changes that lead to improvements. Education Network is one of the most significant problems in this area.

Majority of tasks of forecasting using neural networks boil down to the prediction of time series values. Pluses neural network models is their ability to learn, regardless of the nature of the input information.

Disadvantages of neural networks include the following: generally need around 100 observations to provide a suitable model. Another disadvantage of neural models - a significant cost in time and other resources needed to construct satisfactory model, it is known that teaching networks can consume a lot of time.

Neural networks are apparently non-deterministic, and after of training have a "black box" that somehow works, but the logic of decision-making neural network completely hidden from the expert. However, there are algorithms "Knowledge Extraction from neural networks", which formalizes the trained neural network to a list of logic rules, thereby creating the network on the basis of an expert system. The duration of an information processing network depends on its architecture.

The main cause emergence of the theory of fuzzy logic was the presence of fuzzy and approximate reasoning in describing man-made processes, systems, facilities. Fuzzy expert systems for decision making support are widely used in the economy. In business and finance, fuzzy logic has been recognized in 1988, after only the expert system based on fuzzy rules to predict financial performance predicted the stock market crash.

Characteristic of a fuzzy set is a membership function. Whereas in classical formal logic membership function could take only two values 0 and 1 (respectively, no and yes), then in fuzzy logic based on the concept of fuzzy sets, in which membership is expressed in varying probabilities or degrees of truththat is, as a continuum of values ranging from 0 (does not occur) to 1 (definitely occurs) (respectively, no, yes and maybe). Basis for the operation of a fuzzy inference is the rule base comprising of fuzzy statements in the form of "if-then" and membership functions for the corresponding linguistic variables. In general, the inference engine includes four phases; introduction of fuzziness, fuzzy inference, composition and leading to definition or defuzzification.

Due to the merger of several technologies of artificial intelligence appeared a special term - "soft computing", which introduced L. Zadeh in 1994. Currently, soft computing combined areas such as fuzzy logic, artificial neural networks, probabilistic reasoning, evolutionary and immune algorithms. They complement each other and are used in various combinations to create hybrid intelligent decision making system. Disadvantages fuzzy logic may include: lack of a standard methodology for constructing fuzzy systems; impossibility of mathematical analysis of fuzzy systems existing methods.

Литература:

1.Chennakesava R. Alavala Fuzzy logic and neural networks. Basic concepts & applications, New Age Publication, 2008 - 276 pages.

2.Christos H. Skiadas, Ioannis Dimotikalis, Charilaos Skiadas. Chaos Theory: Modeling, Simulation and Applications, W[ORL]D SC[IENTI]FIC PU\\BLISHING COM\\PANY, 2011 - 468 pages