The Data Preprocessing of Behavior Pattern Discovering in Collaboration Environment

Collaboration environment supports organizations and individuals sharing resource, taking advantage of their knowledge, and cooperating to accelerate their work especially for large scale problem. To find out the law of collaboration works through an automated approach is beneficial to improve collaboration environment, and is known as behavior pattern discovering. Traditionally, preprocessing is considered as data preparation, and mainly covers four kinds of methods as variable selection, sample selection, variable construction, and variable transformation. The preprocessing in behavior pattern discovering focuses on sample selection and variable transformation, because of the four challenges in behavior pattern discovering that the lack of reference knowledge, the cutting short of continuous recorded data, the transformation of resource data according to meanings, and the utilization of various recorded information. This paper emphasizes the indicators that information reservation and data reduction, and points out the major challenge in preprocessing is how to balance the two. Further more, the paper gives out effective algorithms on variable transformation and event sequence dividing. Such algorithms assure that the whole behavior pattern discovering approach works under an efficient and automatic way.