Abstract
The problem of classification in information systems has been studied by many authors, and different methods have been developed. When information is diffuse and the number of obtained values for each attribute is large, the problem becomes complicated. The first step is to create a model that define, based on historical data, the accepted interval of values for each attribute, moving from the minimum to the maximum obtained values in the database. This is what is defined as interval-valued information systems. Using these models, it is possible to find the more important attributes in the classification. This can be done applying decision trees, neural networks, rough sets, fuzzy logic, information theory, among others, or a combination of these methods. After the model is created, it is possible in many cases to identify the new received data and assign to it a class. The main goal of this work is to present three approaches for the solution of this task. In the authors opinion none of them is always better than the other, and the application of one of them depends on the author expertise and the concrete application. Examples are included for clarification purposes.