The Online Examination System (SUO) conducted by the Universitas Terbuka (UT) is now one of the exams that is increasingly in demand by Universitas Terbuka students. The Online Examination System is implemented in all branches of the Open University in major cities in Indonesia. The Amount of Interest Students make UT need to predict how much interest is the registration of SUO, so that UT can prepare facilities and infrastructure so that the exam can run smoothly. Regression is a method that can be used in analyzing predictions, several regression models such as Multiple Linear Regression, Decision Tree, Support Vector Regression and Artificial Neural Network are used in this study. In addition to obtaining accurate predictions, in machine learning a feature selection is added to eliminate unrelated data and also duplicate data. Cities were grouped using the K-Means method. Decision tree model without feature selection has good performance with RMSE = 169.047 +/-0.000 and R2 = 0.749. And for the ANN model combined with optimize selection (evolutionary) it has a good performance value of RMSE = 115.023 +/-38,183 (micro average: 120,592 +/-0.000) and R2 = 0.841 + / -0.082 (micro average: 0.824) with several attributes related to the prediction process.