A pattern decomposition algorithm for data mining of frequent patterns
Knowledge and Information Systems
Engineering Applications of Artificial Intelligence
| Hi-index | 0.01 |
The real-life customer servicing databases share a useful characteristic which can be properly exploited to solve a common problem with the databases themselves. This useful characteristic is that either remark or memo fields are always included in the databases; customer service representatives can use these fields to write down specific things about the service records. This design helps alleviate the following common difficulty with the categorization of customer-service-related problems: customer requested service records are often misclassified owing to human ignorance or bad design of problem categorization. In this paper we propose an ontology-supported technique to preprocess the remark fields, trying to discover meaningful information to help re-categorize misclassified service records. This process restores the database into one with more meaningful data in each record, which facilitates the mining of better association rules. The technique was applied to a real-life trouble shooting database obtained from a telecommunication company. The results show a substantial improvement in the quality of mined trouble shooting rules can be obtained.