An Algorithm for Outliers Mining Based on Rule

ZHANG Lu-lu,JIA Rui-yu,LI Jie
DOI: https://doi.org/10.3969/j.issn.1673-629x.2006.12.024
2006-01-01
Abstract:Outlier mining refers to the mining of the data with obvious departure,and with no general behaviors patterns from large amounts of data.Most existing algorithms don't have good performance when dealing with data-intensive transaction databases.This paper presents a highly efficient rule-based outlier mining algorithm.This algorithm uses multi-layer maximum outlier support and minimum outlier interest,calculates the power of 1-outlier set,and uses linked list to store the transaction identifier in data structure.All these make the time of scanning database only once.So the mining speed and efficiency are enhanced,the result are more useful to decision-making.In this paper,an experiment using this algorithm was carried out on part of a shopping center's scales database.The result presents that this algorithm can mine data-intensive transaction database more effectively and rapidly.
What problem does this paper attempt to address?