Recent financial crises, with an increased volatility and, hence, uncertainty factors, have introduced a high "noise" into the data taken from the financial sectors and overall from any data related to the financial markets, so that the known statistical models do not give trustworthy results. As we know the solutions of the optimization problem can show a remarkable sensitivity to perturbations, coming from the data, in the parameters of the problem. To overcome this kind of difficulties, the model identification problem has been generalized by including the existence of uncertainty with respect to future scenarios through Conic Multivariate Adaptive Regression Splines (CMARS), whose data are assumed to contain certain information with respect to input variables. Then, with the help of robust optimization which can deal with a wider data uncertainty, CMARS has been robustified and named as Robust CMARS (RCMARS). We decrease the estimation variance by using robustification in CMARS. In contrast to early studies, where RCMARS was presented in theory and method and illustrated with a numerical example, in this study, we present RCMARS results for real-world data from financial markets, particularly, from the Istanbul Stock Exchange, Turkish and US economy, showing that RCMARS can generate more accurate models with a smaller variance.