Many universities and higher institutions around the world pay much attention on the admission of students to their different colleges. For this reason, predicting the academic performance was the main purpose in order to help students to be success ful in their study life. This can be done by investigating the performance of students according to their cohort, gender and major. The performance is evaluated through the cumulative GPA. The main objective of our study was to predict the performance of students in col lege of science. To this end, seven predictors have been used; including the high secondary school scores in the major subjects, the competitive score (CS) and the location (governorate). In statistics field, many methods can be used to achieve our goal. In this study, we mainly focused on the following multivariate statistical techniques: Discriminant Analysis (DA), k-Nearest Neighbours (K-NN) and Or dinal Logistic Regression (with three distinct models). Then, we have compared them by computing the Apparent Correct Classification Rate (ACCR). The pre diction purpose have been illustrated through some explicit examples. In addition, observations made from the descriptive statistics part (Chapter 2) motivated us for suggesting some adjustments in the admission process. To sum up, we found that Quadratic Discriminant Analysis (QDA) is the most appropriate model for predicting the performance of female students. On the other hand, (k-NN) was found to be the most suitable statistical method for predicting the performance of male students. Furthermore, future research on new ordinal models, built on fat-tailed and skewed distribution, may lead to a substantial improvement of the ACCR for male stu dents.