Mining fuzzy rules in a donor database for direct marketing by a charitable organization

Abstract

Given a donor database by a charitable organization in Hong Kong, we propose to use a new data mining technique to discover fuzzy rules for direct marketing. The discovered fuzzy rules employ linguistic terms, which are natural for human users to understand because of the affinity with the human knowledge representations, to represent the association relationships revealed in the data. The proposed approach utilizes an objective measure to distinguish interesting associations from uninteresting ones. Furthermore, it allows the ranking of discovered rules according to an uncertainty measure and allows quantitative values to be inferred by the discovered fuzzy rules. The domain expert from the organization is interested at finding how the response of a donor is affected by his demographics (e.g., age, education, occupation, salary, etc.) and his donation histories (e.g., the average yearly donation frequency, the average monthly donation amount, etc.). We applied the proposed approach to the donor database in order-to mine a set of fuzzy rules. The experimental results showed that our approach is able to achieve accurate prediction of donor's response. By examining the discovered rules, the domain expert has found some unexpected patterns and formulated some direct mail strategies for future use.