Outlier detection is one most important issue in recent years. Outlier detection is the process of detecting errors in data. The recent methods are mostly based on the Numerical data, but these methods are not suitable in real time data such as web pages, business transactions etc., which are known as Categorical data. It is difficult to find outliers in categorical data. In this paper, we propose an approach to find outliers that is Comparison of Deviations. In Comparison of deviation method, hyper graph is used to calculate the deviations of each object in the database. After comparing all deviations which are having most negative term that objects are treated as outlier