A Novel Approach for Outlier Detection using Rough Entropy

Outlier detection is an important task in data mining and its applications. It is defined as a data point which is very much different from the rest of the data based on some measures. Such a data often contains useful information on abnormal behavior of the system described by patterns. In this paper, a novel method for outlier detection is proposed among inconsistent dataset. This method exploits the framework of rough set theory. The rough set is defined as a pair of lower approximation and upper approximation. The difference between upper and lower approximation is defined as boundary. Some of the objects in the boundary region have more possibility of becoming outlier than objects in lower approximations. Hence, it is established that the rough entropy measure as a uniform framework to understand and implement outlier detection separately on class wise consistent (lower) and inconsistent (boundary) objects. An example shows that the Novel Rough Entropy Outlier Detection (NREOD) algorithm is effective and suitable for evaluating the outliers. Further, experimental studies show that NREOD based technique outperformed, compared with the existing techniques. Key-Words: Data Mining, Outlier, Rough Set, Classification, Pattern recognition