Abstract: Outlier analysis is becoming an important technique in data mining whose task is to identifying the data objects that are completely different from the majority of all objects. Outlier detection is necessary and useful with numerous applications in many fields like medical, fraud detection, fault diagnosis in machines, etc. In this study, we tend to propose a cluster based outlier detection algorithm which can be fulfilled in two stages. In the first stage we construct cluster using mutual nearest neighbor graph clustering algorithm. In the second stage we find the cluster outlier factor based on size of each cluster. The concept is to find outlier value of object and outlier clusters are extended to the formation of cluster. This algorithm is used to identify objects must confided as outlier object and outlier clusters in a database. This algorithm is based on the thought of mutual nearest neighbor graph clustering. The proposed algorithm can be used to identify the outlier value factor in the database and to detect the outliers and outlier clusters efficiently. The simulation result shows that the proposed method yields better results in outlier detection.