Conference Title: International Conference in Business,Technology and Innovation 2013
Conference Title: UBT Publications
Contributing Editos: Prof. Albert Qarri, Prof. Ibrahim Krasniqi, Krenare Pireva, Evelina Bazini, Nita Abrashi
Co-Organizers: Ardian Emini, Vlora Aliu, Betim Gashi, Xhemajl Mehmeti, Kushtrim Dragusha, Murat Retkoceri, Kaltrina Bunjaku, Leonita Braha
Start Date: 2013-11-01
End Date: 2013-11-02
Venue, City, Country: : Hotel Bleart , Durres , Albania

Paper Title

A Neuro-Genetic Model In GDP Forecasting: Albania Case


Dezdemona Gjylapi, Vladimir Kasemi


ANN is the emulation of a biological neural network, which is composed of many interconnected neurons. Artificial neural networks have an essential characteristic that distinguishes them from traditional algorithms, because they are defined as a model that learns to perform a task rather than being directly programmed. A genetic algorithm consists in applying the principles of natural biological evolution in artificial systems. Genetic algorithms belong to the class of optimization procedures they are mainly applied in spaces that are too large and complex to be searched in a finite way. Genetic algorithms have been used for neural networks in two main ways: to optimize the network architecture and to train the weights of a fixed architecture. A particular type of evolving systems, namely neuro-genetic systems, has become a very important topic of study in neural network design. The aim of this paper is to present a neuro-genetic model that we used in GDP forecasting. GDP is an economic indicator that represents the value at market prices of all goods and services produced within a country in a given period (usually a year). Its forecasting is of particular importance for the fiscal and monetary policy makers in the preparation of timetables which have as a target macroeconomic stability and sustainable economic growth. We use 10 factors that affect the determination of the DGP for the Albania’s GDP forecasting. The genetic algorithm is used to train the weights of different architecture MLPs. We compare the output of these NN and find the best NN architecture which archives the high accuracy for GDP forecasting. As a result we can say that the neuro-genetic model achieves a better GDP forecasting than the traditional methods do.


artificial neural networks, genetic algorithm, forecasting, GDP