Publication details
Graph Mining and Outlier Detection Meet Logic Proof Tutoring
| Authors | |
|---|---|
| Year of publication | 2014 |
| Type | Article in Proceedings |
| Conference | Proceedings of EDM 2014 Ws Graph-based Educational Data Mining (G-EDM) |
| MU Faculty or unit | |
| Citation | |
| Field | Informatics |
| Keywords | logic proofs; resolution; educational data mining; graph mining; outlier detection |
| Description | We introduce a new method for analysis and evaluation of logic proofs constructed by undergraduate students, e.g. resolution or tableaux proofs. This method employs graph mining and outlier detection. The data has been obtained from a web-based system for input of logic proofs built at FI MU. The data contains a tree structure of the proof and also temporal information about all actions that a student performed, e.g. a node insertion into a proof, or its deletion, drawing or deletion of an edge, or text manipulations. We introduce a new method for multi-level generalization of subgraphs that is useful for characterization of logic proofs. We use this method for feature construction and perform class-based outlier detection on logic proofs represented by these new features. We show that this method helps to find unusual students' solutions and to improve semi-automatic evaluation of the solutions. |