Graphs are a powerful concept useful for various tasks in science and engineering. In applications such as pattern recognition and information retrieval, object similarity is an important issue. If graphs are used for object representation, then the problem of determining the similarity of objects turns into the problem of graph matching.

In this thesis the comparison of input graphs with databases of graphs is studied. Graph matching in general is a computationally expensive procedure. If an input graph is matched against a database of graphs, the size of the database is introduced as an additional factor into the overall complexity of the matching process. In this work, an approach based on machine learning techniques is pursued to reduce this factor. The graphs are represented using feature vectors. In a preprocessing step, these vectors are used to build a decision tree indexing the database. At runtime, the decision tree is used to eliminate a number of graphs from the database as possible matching candidates.