Universität Rostock, 2016
Abstract: The aim of this thesis is to develop methods that suggest the users suitable subnetworks for integration during modelling. To this end, techniques from the field of recommender systems are used, which aim to predict the users’ interest in certain objects in order to filter and recommend the most suitable ones. Especially association rule mining is of particular relevance in this thesis. Its algorithms offer the opportunity to find patterns of joint appearance in a large set of items. For this purpose, biological networks are considered, which are represented as graphs and annotated with standardised ontology terms. Association rule mining then is applied with respect to structural and also to semantic similarity. For a partly modelled biological network the elements are found that may extend it. The obtained results form a solid basis for the development of a recommender system that facilitates the efficient reuse of networks and decreases the manual effort to find and integrate relevant structures.
master thesis free access