**2. Outlier detection**

Outliers are measurements that are different from other values of the same dataset and can be due to measurement errors or to the variability of the phenomenon under consideration. Hawkins (Hawkins, 1980) defined an outlier as *"an observation that deviates so much from other observations as to arouse suspicion that it was generated by a different mechanism"*.

The detection of outlier is an important step of data mining because it improves the quality of data and it represents a useful pre-processing phase in many applications, such as financial analysis, network intrusion detection and fraud detection (Hodge, 2004).

Classical outlier detection methods can be classified into four main groups: distance-based, density-based, clustering-based and statistical-based approaches. All these approaches have several advantages or limitations and in the last years a lot of contributions have been proposed on this subject. Artificial intelligence techniques have been widely applied to overcome the traditional methods and improve the cleanness of data; in particular some fuzzy logic-based approaches proved to outperform classical methodologies.
