At recent years, estimating the churners before they leave has gained importance in environment of increased competition in company strategy. In this paper, churners are tried to detect by using data mining classification techniques. Attribute reductions are tried for decreasing the runtime and increasing achievement of models and performance was measured by using different classification method. In addition, outlier analysis is applied to dataset and then effects on classification results are examined. This classification methods are tested in two datasets which are taken from Telecommunication Companies. Recall and Precision Rates are used as performance criteria.