Outlier Analysis of Categorical Data using NAVF

Outlier mining is an important task to discover the data records which have an exceptional behavior comparing with other records in the remaining dataset. Outliers do not follow with other data objects in the dataset. There are many effective approaches to detect outliers in numerical data. But for categorical dataset there are limited approaches. We propose an al-gorithm NAVF (Normally distributed attribute value frequency) to detect outliers in categorical data. This algorithm utilizes the frequent pattern data mining method. It avoids problem of giving k-outliers to get optimal accuracy in any classification models in previous work like Greedy, AVF, FPOF, and FDOD while finding outliers. The algorithm is applied on UCI ML Repository datasets like Nursery, Breast cancer mushroom and bank dataset by excluding numerical attributes. The experimental results show that it is efficient for outlier detection in categorical dataset.