A Model of User-Oriented Reduct Construction for Machine Learning

An implicit assumption of many machine learning algorithms is that all attributes are of the same importance. An algorithm typically selects attributes based solely on their statistical characteristics, without considering their semantic interpretations. In order to resolve difficulties associated with this unrealistic assumption, many researchers attempted to introduce user judgements of the importance of attributes into machine learning. However, there is still a lack of formal framework. Based on decision theory and measurement theory, a model of user-oriented reduct construction is proposed for machine learning by considering the user preference of attributes. It seamlessly combines internal information and external information. User preferences of attributes are extended to user preferences of attribute sets. Accordingly, user preferred reducts can be constructed.