There are many possible techniques to classify the data , DA and MLR techniques are Mostly important, they are often used to handle categorical dependent variable problems. This study aims to find the best statistical model for graduates data in Palestine, a comparison between discriminant analysis and multinomial logistic regression model on set of real data " Circumstances of Graduates From High Education and Vocational Training Survey Main Findings (2005- 2006)" where conducting by Palestinian Center Bureau of Statistics (PCBS). In this study we compared the two statistical models using different assessment techniques (leave-one-out cross-validation and method ROC curves) and obtained the best estimate of accuracy and error rate in order to achieve the best model for the data. Two method applied at Graduates data(2005-2006) which has 9 independent variables and one dependent variable (Employment Status of graduates) with three categories (Employment , Unemployment and Outside of LF). As result of classification of the two techniques, MLR can predict better than DA. This is justified by correct classification of 67.2% by the MLR model and 65.2% by the DA in the analysis. The ROC curves difference in the area under the curve (AUC), MLR 91.42% versus DA 53.52% . Thus the results we have from Multinomial Logistic Regression model are better than the results of Discriminant Analysis model for this data set .