Growing amount of information in the world has increased the need for computerized classification of different objects. Therefore, more important have become automatic data processing techniques which make use of machine learning. Classification is one of the machine learning tasks where the program learns to classify new instances from the provided facts. In the thesis the term automatic classification is used to denote a computerized classification process which excludes the user or expert involvement starting from the classifier’s training with provided data set till applying it for new instance classification. Application domains and data are getting more complex leading to inability for automatic classification approaches to always reach the desired result. Thereof, the thesis is devoted to the development of automated or semiautomatic classification solution which incorporates both machine learning facilities and interactive involvement of a domain expert in the classifier’s applying stage for improving its results if the classifier makes uncertain classification. One of the classification method groups used in machine learning is inductive learning. Results obtained from the inductive learning methods are interpretable not only for machines, but also for their users. This is a fundamental advantage over other classification methods. To fully utilize interactivity with an expert, the classification system is based on inductive learning for extracting human-readable classification rules. The doctoral thesis provides interactive inductive learning based classification system’s (InClaS) model which gathers algorithms, architectures and guidelines for developing an interactive classification system. The aim of developing this system is to decrease the number of misclassified instances regarding to the “traditional” automatic classification system. This model is particularly intended to be applied in domains with a multi-label class membership. The InClaS model has been approbated in two problem domains – education and medicine – which demonstrate the ability to reduce the number of misclassified instances supposing that instances with uncertain classification (unclassified and classified with low confidence) are detected and transferred to the expert for assessment. The results of the thesis are published in 13 international scientific publications and presented in 12 international conferences. The doctoral thesis includes introduction, 6 sections, main results and conclusions section. It consists of 160 pages, 47 figures and 34 tables in the main text, 11 appendices. The bibliography contains 139 references.