Concept of Observer to Detect Special Cases in a Multidimensional Dataset

In a dataset, the special cases are data that appears to be inconsistent with the neighborhoods of data. The special cases are warnings that usually require specific processing in social networks, medical applications or complex processes. This paper proposes to use the observer’s paradigm to detect special cases in a multidimensional data space. Thus the observations allow to define the neighborhoods of data. Then we propose a rareness index for each data. The special cases have the highest values of rareness index. Experimental results show the ability of the method to detect these special cases. We conclude this paper with a brief discussion.