The paper describes the design of interactive inductive learning-based classification system. The architecture of machine learning systems can be viewed from two perspectives, namely, (1) the stages of system design and (2) model of system’s functioning and components. Both of these design issues of different existing classification systems are discussed in the related work. A general architecture for the interactive classification system is proposed. Domain-dependent parts of the system are specified in the more detailed architecture of the interactive multi-label classification system for study course comparison. Interactive inductive learning-based classification system in uncertain conditions could ask a human for decision, and it is has been proven that applying this approach can reduce the number of misclassified instances, especially, when the initial classifier performs poor.