• A
  • A
  • A
  • ABC
  • ABC
  • ABC
  • А
  • А
  • А
  • А
  • А
Regular version of the site

Forecasting inter-day ruble exchange rate dynamics using forecast combinations

Student: Avramenko Ol`ga

Supervisor: Vladimir Pyrlik

Faculty: Faculty of Economics

Educational Programme: Master

Year of Graduation: 2014

<p>There are a lot of medium-term forecasting models for time series nowadays.<br />Method of combining individual forecasts was chosen in this work. One of the main advantages of the combination is reduction of the forecast error that increases forecast accuracy. As far as the idea of ​​combining forecasts is relatively new, so there are some gaps in its presentation and application. One of the major issues that are still unresolved is a method for estimating the weights combined with individual forecasts.</p><p>In this research we illustrated this method by the forecasting currency. We suggest a special case of weighted least squares to combine individual forecasts of a family of SARIMA models. The resulting predictions were combined with some weights, given in several different ways. &nbsp;In addition to daily data on the dynamics of the ruble against other major world currencies - the euro and the dollar, this paper also addressed intraday data, namely hourly and 4-hourly data.</p><p>We propose a new technique combining short-term forecasts of the time series based on a family of nested models. Proposed technique is a special case for combining the weighted regression, in which the outliers to the dynamics of individual forecasts have zero weight, and other observations gain weight in proportion to the square of the observation serial number. Varying the cutoff level of outliers we can find a compromise between noise pollution forecasts as a result of outliers included in the sample and reduction in the quality of the combining forecast due use of too small a sample. In this research we also propose weights obtained by the above methods, set in proportion to the forecasted values.</p><p>As a family of models considered for the forecast model specification SARIMA is used. The coefficient MAPE is used to assess the results (forecasts), since it is more resistant to scatter projected data on the true values ​​and the scatter relative to the trend of the time series. Application of the proposed combining technique will provide better predictions in terms of mean absolute percentage error of the forecast than the individual forecasts or combining technique based on the OLS.</p>

Student Theses at HSE must be completed in accordance with the University Rules and regulations specified by each educational programme.

Summaries of all theses must be published and made freely available on the HSE website.

The full text of a thesis can be published in open access on the HSE website only if the authoring student (copyright holder) agrees, or, if the thesis was written by a team of students, if all the co-authors (copyright holders) agree. After a thesis is published on the HSE website, it obtains the status of an online publication.

Student theses are objects of copyright and their use is subject to limitations in accordance with the Russian Federation’s law on intellectual property.

In the event that a thesis is quoted or otherwise used, reference to the author’s name and the source of quotation is required.

Search all student theses