The Curriculum-Based Course Timetabling (CB-CTT) is a University Timetabling problem considered as NP-hard. It is widely studied in the literature because of both scientific and practical interest. In CB-CTT, lectures have to be assigned to time slots and rooms with respect to resource constraints, i.e., the teachers' availability. Heuristics and hyper-heuristics are known to give a good compromise between performance and rapidity to solve university timetabling problems. These methods have many parameters and can benefit from information about instance problems to improve their performance. Many features are proposed in the literature to help the parameterization of heuristics. In this work, we focus on landscape metrics and particularly we study local optima networks. We compute many features in order to identify the relevant ones. Then, we build a model to predict the performance of a local search based on the selected features.