Esti̇mati̇on of di̇recti̇on of exchange rate, gold pri̇ces and stock market returns wi̇th arti̇fi̇ci̇al neural network and hi̇gh order markov chai̇n models

In this study The Istanbul Stock exchange national 100 index, USD/TRY exchange rate and Gold price returns predicted with Artificial Neural Network and High Order Markov Chains models. Each unit of transition probability matrix of Markov chain was computed by estimated values of Artificial Neural Network Algorithm. Both Markov Chain Models and Artificial Neural Network Algorithm are efficient methods separately at the prediction of highly volatile and complicated financial time series such as stock market, exchange rate and gold price returns. Utilizing Markov Chain Models and Artificial Neural Network Algorithm together, both confidence level for transition probability matrix and powerful results was obtained for one-step ahead prediction. For each process the correct classification rate of Artificial Neural Network algorithms was used as confidence level of transition probability matrix. Although the satisfactory information relative to future prediction of series was obtained both by second order and third order Markov Chain Models, the results show that third order Markov chain models contain more information than second order Markov Chain Models. Put it in a different way, in general the Istanbul Stock exchange national 100 index, USD/TRY exchange rate and Gold price returns contain more information from three days before than two days before. Lastly, the results indicated that as the order of Markov Chain Models decreases the prediction power of Markov Chain models and Artificial Neural Network Algorithm goes down correspondingly.