This study investigates an approach of knowledge discovery and data mining in insufficient databases. An application of Computational Taxonomy analysis demonstrates that the approach is effective in such a data mining process. The approach is characterized by the use of both the second type of domain knowledge and visualization. This type of knowledge is newly defined in this study and deduced from supposition about background situations of the domain. The supposition is triggered by strong intuition about the extracted features in a recurrent process of data mining. This type of domain knowledge is useful not only for discovering interesting knowledge but also for guiding the subsequent search for more explicit and interesting knowledge. The visualization is very useful for triggering the supposition.Área: Ingeniería de Software - Bases de DatosRed de Universidades con Carreras en Informática (RedUNCI)