Speaker identification is a field of which usage grows faster in security systems and forensic sciences. Depending on the tasks, online or offline applications are presented. It is an important problem that how much they are accurate, how much they are fast or how hard is its computation. In this study, the accuracy and the speed of the classifiers that can be used on speaker identification and the effect of the number of Mel Frequency Cepstrum Coefficients (MFCC) are examined. A dataset containing eighty speakers are classified with "Support Vector Machines", Linear Discriminant Analysis", "k-Nearest Neighbor" and "Naive Bayes" classifiers. The results are compared based on the accuracy, the speed and the number of MFCC. The highest classification accuracy is computed as %98,3. The applicability of the presented systems on real time is discussed.