This study aims to compare the effect of 5 different feature selection algorithms on support vector regression (SVR) to predict eye angular displacements corresponding to electrooculography (EOG) data. Feature extraction was done from the vertical and horizontal channels in the EOG signals in the data set. Eye angular displacements were estimated by processing these features with the SVR method. The performance of the feature selection methods was compared according to the RMSE evaluation criteria. The lowest error rates were obtained by using the minimum redundancy maximum relevance (fsrmrmr) feature selection method, 1,97E+00 in the vertical channel and 7,11E+00 in the horizontal channel.