Data mining is one of the most exciting fields of research for a researcher. In data mining, outlier detection is one of the important area where similar kind of data objects are grouped together and the objects that does not belong to the group are termed as outliers. This helps in finding objects that have different behavior with respect to other objects. Due to the presence of outliers overall nature of the data may be compromised. So it is a challenging task to find outliers present in the data. Every day huge amount data is flowing around us which belong to different streams, so our main is to find the objects that does not belong to the particular stream. In this paper, different outlier detection algorithms are described and implemented and the best algorithm among them is found based on their performance with the help of MOA tool. Performance issues like memory consumption, domain queries, time are shown. MOA tool contains prescribed algorithms where one can be used as a base algorithm to compare remaining algorithms. Each algorithm is an increasing and adaptive to concept extension. Finally the performance of each algorithm is tabled.
展开▼