Abstract:

In this study, the BIST 100 index, which is traded in the Stock Exchange in Istanbul, is intended to be predicted through the 10-year index bond interest rates of the American Treasury. 258 data obtained in literature were analyzed by two mathematical methods used in recent years. The random forest (RF) model and the multi-variable adjustable regression currents (MARS) model used in the time series field are tree-based regression models used in this study. In the models used BIST 100 index closing prices depend on the variable American Treasury 10 year indicator bond interest rates are determined as an independent variable. In the analysis phase, 206 data models were trained and 52 data were used to test the models. The statistical success of the models has been tested by the failure square average (HKO) and the Nash-Sutcliffe model efficiency ratio (NSE) success criteria. When the results were examined, it was found that the MARS model had the highest NSE value and that the American Treasury could predict the 10 year indicator bond interest rates of the BIST 100 index. In our country, the newly used methods in the financial field estimates are believed that more successful investment decisions can be made. It is also believed that the models created by the study will later be developed and keep the light to other researchers.

Abstract:

In this study, it is aimed to estimate the BIST 100 index traded in Borsa Istanbul using the US Treasury 10-year benchmark bond interest rates. The 258 data obtained were analyzed with two mathematical methods used in the literature in recent years. Random Forest (RF) Model and Multivariate Adaptive Regression Spline (MARS) Model used in the time series field are tree-based regression models used in this study. In the models used, the BIST 100 index closing prices are the dependent variable; The US Treasury 10-year benchmark bond interest rates were determined as the independent variable. During the analysis phase, 206 data were used in training the models and 52 data were used in testing the models. Whether the models were statistically successful or not was tested with the success criteria of mean squares of error (MSE) and Nash–Sutcliffe model efficiency coefficient (NSE). When the results are analyzed, it is seen that the MARS model has the highest NSE value and the US Treasury 10-year benchmark bond interest rates can predict the BIST 100 index. It is thought that more successful investment decisions will be made thanks to these new methods used in the estimations made in the field of finance in our country. In addition, it is believed that the models created by the study will be developed later and shed light on other researchers.

Keywords: