Outlier mining is concerned with the data objects that do not comply with the general behavior or model of the data, such data Objects, which are either different from or inconsistent with the remaining set of data. Studying the extra ordinary behavior of outliers helps uncovering the knowledge hidden behind them and providing an approach to the decision makers to make profit or improve the service quality. Hence, mining for outliers is an important data mining research with numerous applications, including credit card fraud detection, criminal activities in E-commerce, unusual usages of telecommunication services, Weather Forecasting etc. Moreover, it is useful in digital and customized marketing for identifying the spending behavior of customers with extremely low or extremely high incomes, or in medical diagnose for finding unusual results to various medical treatments.
Some data mining techniques discard outliers as noise or exceptions. While in some applications, these exceptions are considered more interesting than regularly occurring ones like in terrorism attack. Challenges in outlier detection include finding appropriate data models, the dependence of outlier detection systems on the application involved, finding techniques to distinguish outliers from error or exception, and providing justification for identification outliers. Outliers can be detected through N-gram technique but this technique is using a large storage space to store metadata and data dictionary. There are a number of compression models e.g. Content tree weighting method, LZ77, LZ78, LZW that are used in compressing text & image. Burrows