No abstract

In this paper, a new concept of Trusted Knowledge (TK) is introduced. Trusted Knowledge are data from trusted organizations such as ministries, statistical offices and so on which can replace a domain expert in the evaluation phase of the data mining task. Two approaches to applying Trusted Knowledge are introduced. The first one called "Explanation system" offers additional information relevant to the resulting patterns which can help the user to better understand results of the task. The second one called "A/TKformulas" filters out the resulting patterns which are consequences of Trusted Knowledge and thus enables the user to concentrate on the interesting patterns. Conversely, the user can request to be shown only the resulting patterns which are consequences of TK to see which of them are in line with TK. Feasibility of the newly proposed framework is demonstrated in a case study.