Forecasting Non-stationary Time-Series Using Artificial Neural Networks

<p align="center"><strong>Annotation on a master&#39;s dissertation on &laquo;</strong><strong> Forecasting Non-stationary Time-Series Using Artificial Neural Networks</strong><strong>&raquo;</strong></p><p>My Master&#39;s paper is devoted to analysis of the effectiveness of the usage of artificial neural networks to predict the financial non-stationary time series. The issue of forecasting financial non-stationary time series is highly relevant today, because it is considered to be the basis for the financial and business planning as well as enterprise clear management and optimization processes. Effective forecasting tool allows predicting future values ​​of the interest parameters and acting with respect to the obtained results.</p><p>The purpose of my master dissertation is to show how neural networks are able to cope with the problem of forecasting in the various types of non-stationary time series and how effective they are.</p><p>According with the purpose of my dissertation, I have put forward the following tasks:</p><p>1) To show the theoretical basis of structure formation of the neural network;</p><p>2) To provide theory on machine learning and neural network optimization;</p><p>3) To determine are these methods are effective to build a network to predict different types of non-stationary time series;</p><p>4) To show how is the choice of the method of network developing and optimization influence on outcome prediction;</p><p>5) To make conclusion about the effectiveness of using neural networks in terms of prediction non-stationary time-series.</p><p>All real data is used in my paper is available in the internet. To complete practical part of the dissertation, I used mathematical high-level software.</p><p>&nbsp;</p><p>While writing the paper, I used data from the official United States statistics service and from the trade base and analytical portal &ldquo;finam.ru&rdquo;.</p><p>The study concludes that the effectiveness of the usage of artificial neural networks for forecasting of non-stationary time series is highly depends on chosen machine learning algorithm and optimization method.</p><p>The structure of the paper includes an introduction, four chapters, fourteen paragraphs, a conclusion and bibliography. In this paper, we used thirty tables, thirty-four graphics and thirty-one source literature.</p>