Abstract:

This paper focused on the analyzing the impact of the different stock market indices up to 7 delay on the increasing or decreasing in the Czech stock market (describing by PX index) between two trading days. Firstly , the influence of the stock market indices of the European region is analyzed, then the data set was enriched by stock markets representing the G-20 Economies .The model building step was applied separately for two data sets : for simple and logarithmic returns. The impact of the different stock market indices on the direction of PX index’s changes was analyzed by applying Stochastic Gradient Boosting approach. In case of the European region analysis only, the changes of PX index (increasing or decreasing ) are most influenced by CAC40, DAX, AEX Amsterdam, IBEX 35, BY enlarging the data set by the stock markets of G-20 Economies the list of the most important variables has been changed and the models ‘ quality measured by area under the ROC curve has been improved by the changes in time t of the following stock market indices : S&P 500 Index , NASDAQ composite index, Russell 1000 Index , Dow Jones Industrial Average . For both steps : before and after the data set enlarging, one of the similar features is that the top four most important variable come from time t,i.e. without lagging.