A Neural Network and Genetic Algorithm Hybrid Model for Modeling Exchange Rates: the Case of the US Dollar/ Kuwaiti Dinar
Total Page：16
File Type：pdf, Size：1020Kb
Topics in Middle Eastern and African Economies Vol. 12, September 2010 A Neural Network and Genetic Algorithm Hybrid Model for Modeling Exchange Rates: The case of the US Dollar/ Kuwaiti Dinar By: Meriem DJENNAS*, Mohamed BENBOUZIANE** and Mustapha DJENNAS** *Faculty of economics, University of Amiens, France ** Faculty of economics, University of Tlemcen, Algeria [email protected], [email protected], [email protected] Abstract Predicting exchange rates is one of the most leading financial problems because of its intrinsic difficulty and practical applications. In recent years, many nonlinear models have been proposed in the literature to modify the results of prediction in order to improve the forecasting performance of high frequency exchange rates. Neural networks and chaotic models are among the models that have been exploited and have shown promising results. The main objective of our research is to conduct a comparative evaluation of nonlinear models on a set of data and variables and to verify the predictive power of neural models under the same experimental conditions. This study uses a criterion to evaluate the model performance: the square root of the mean squared error. Our research study will be applied to the case of the US Dollar – Kuwaiti Dinar exchange rate. Key Words: Exchange Rate –Neural Networks – Genetic Algorithm - Kuwaiti Dinar. INTRODUCTION Usually, the study of the financial market evolution, and more particularly exchange markets, is considered as one of the most important research fields in international finance as long as exchange markets are considered as open systems that react on the base of collected information during the time periods. The behavior modeling process of the exchange markets requires beforehand a deepened knowledge of the exchange system and factors that influence it. Hence, the inherent objective is to extract the founding rules of a relevant modeling process. It goes without saying that in these markets, the exchange values are difficult to predict. In this fact, the development of an efficient strategy and a prediction tool that resists in facing the financial shocks is the subject of several research and studies. The major risk which an intervening party is dealing with is the non-linearity of exchange rate data sets. Thus, this study is articulated with the following problematic: the contribution of the application of artificial intelligence tools in the modeling of the agents’ behavior intervening in a financial market while trying to improve their training processes for the prediction. It consists in applying a hybridization of two fundamental tools of artificial intelligence, the artificial neural networks (ANN) and the genetic algorithms (GA). First, we construct a neural network model for the exchange rate forecasting. Agents-based modeling in an artificial market by neural networks is perfectly adapted for the resolution of complex and non-structured problems which are difficulty accessible by the mathematical or statistical approaches (non-linearity 1 Topics in Middle Eastern and African Economies Vol. 12, September 2010 conditions or inexistence of the adequate mathematical model). We will show that neural network models offer the advantage to operate in the changing, hostile and unpredictable environments. Secondly, we apply the agents-based model that provides a convenient structure to test the learning capacity and reasoning concerning interactions between traders in the exchange market where the agents’ training will be studied by a genetic algorithm. Since agents operate and interact in the environment, the issue is to take decisions that offer a higher level of flexibility and performance. Thus, we suppose that agents can take the best decision (optimizing their performance) if they possess a better knowledge (availability of information) about the market environment. Therefore, the objective is to implement a rigorous artificial market that permits to understand some real world market mechanisms and to foresee some others. As far as possible, most fundamental elements and properties of the macro environment of the exchange market as well as those of the micro environment will be preserved. This research reasoning will be applied in the Dollar – Kuwaiti Dinar exchange rate. Without calling into question the econometric methods utility for the prediction, we will demonstrate, on the basis of a comparative survey, that artificial intelligence tools are more effective under some market conditions like the non-linearity. Thus, we will try to demonstrate that they permit to predict with a higher level of precision the tendency and variations of the exchange rate in a financial market. The rest of the paper will be articulated as follows: first we will vive a description of the market of exchange rates in the GULF countries, Then we proceed to give more insights on technical methodology and how to deal with neural networks and genetic algorithms. In the third section, we will apply these techniques to forecast the US Dollar Kuwaiti Dinar exchange rate, and finally we will give some concluding remarks. Section I. The Exchange rate Market in the GCC Countries I.1. Introduction During the Bretton Woods institutions, the Arab countries have shown some willingness to establish cooperation managing their exchange rate policies. In 1945, 22 countries have planned to launch a single currency called "Arab Dinar". Sixty years later, only countries of the Cooperation Council for the Arab States of the Gulf CCASG1 (Saudi Arabia, Kuwait, Oman, Qatar, United Arab Emirates and Bahrain) still ongoing efforts to create a common currency in 2010. The introduction of the single European currency in 19992, which experienced a sharp appreciation against the dollar on the foreign exchange market due to the financial crisis, seems to have been emulated since the Gulf countries have confirmed during a summit in Doha in March 2009, the introduction in 2010 of a single currency. Although Oman has decided to withdraw from the project and that Kuwait has decided to link its currency to a basket currencies rather than the dollar in order to face inflation, the determination of Gulf monarchies did not been initiated so far. The Kuwait decision is explained by the sharp rise in inflation in the Arabian Peninsula. 1 Or Gulf cooperation Council GCC. 2 For a detailed discussion on the Euro introduction, see Neaime and Paschakis (2002). 2 Topics in Middle Eastern and African Economies Vol. 12, September 2010 The Gulf countries have been forced to reduce their interest rates to follow the Fed’s (Federal Reserve System) downturn decisions to fight against speculation on their currencies, although rate increases are deemed necessary to curb the inflation. The central banks of the GCC countries are thus in the same uncomfortable position as the European Central Bank (ECB). But the dollar weakness and inflation rising (two economic factors that erode their oil revenues) have undermined the single currency project even though some divisions persist between countries. Thus, if Saudi Arabia, at the initiative of the GCC in 1981, the largest producer and exporter of crude oil in the world, make lobbying to keep the deadline, others countries like the UAE, bring up logistical challenges for the project establishment. I.2. Economic integration in the Gulf countries In 1981, the United Arab Emirates, Bahrain, Saudi Arabia, Oman, Qatar and Kuwait have put together the Gulf Cooperation Council (GCC) to create an economic and financial competitive integration. The six countries also share many common cultural, historical and social values. These factors, in addition to geographical proximity, have facilitated the interaction and business transactions, and have created a homogeneous exchange zone. The objective of GCC is to strengthen alliances and relationships between member’s countries. It aims to standardize various areas of activity such as economics, finance, trade, tourism, legislation, administration, agriculture, etc. To accomplish the above objectives, the council outlined several steps. The most important are: 1. Member countries should allow free movement of imports and exports of natural resources, agricultural and industrial products. 2. Trade policy with other regional economic conglomerates should be unified to create well-balanced trade and exchange relationships. 3. Creating a free movement area for people and goods. 4. The investment rules should be harmonized in all member countries. 5. Member countries should coordinate their fiscal and monetary policies and establish cooperation between financial institutions and central banks. The ultimate goal of the GCC countries is to jump from a cooperation and coordination strategy to a more advanced economic integration. The union has already succeeded in creating a common market called "The Gulf Common Market" in January 20083. Such that goal requires disengaged movement of people and physical capital. Finally, the GCC aim to establish a competitive monetary and financial integration by adopting a common currency in 2010. Efforts were also intensified to unify the foreign trade policies in the GCC countries. Establishing a single currency should expect to promote trade and financial integration, facilitate foreign direct investment, and support development of the Gulf countries in an optimum currency area (OCA). 3An interesting description of the CCG is contained in a recent study realized by the European Central Bank; see Sturm et al.