Outlier detection is considered as a critical and crucial stage in data pre-processing. What adds to the importance of this stage is that the results of data mining depend on it. In outlier detection, data that are totally different from other data and do not show the normal data’s behavior are detected and deleted. Data mining will be continued by using normal data. If the outliers will not be deleted, and be processed accompanying with the normal data, it may lead to getting incorrect model that does not show the normal behavior of data. These data are deleted while either their numbers are high in some cases or they have special meanings and values. This study investigates the usage of outlier in these cases in order to examine whether the separate processing of these data is advantageous or not. For this purpose, the data processing approaches will be separately done on outliers and normal data in order to achieve model/models of expressing each group’s behavior. In analyzing the results, will use accuracy, and the accuracy of suggested approach, which is the separate analysis of normal data and outlier is compared with the normal approach, which is outlier deletion. In this regard, four standard data set, which are usually applied in outlier detection, that are German credit, WDBC, Pima Indian Diabetes and BCW and one set of dummy data will be utilized. Based on the accomplished implementations, processing of these data, at least according to the approach that was used in this study, is not suitable as a common method. Key words: outlier, outlier detection, using outlier