New Approaches for the School Timetabling Problem

School timetabling is a hard task that educational centers have to perform regularly and which implies a large waste of time and human efforts. For such reason designing techniques for the automatic generation of timetables is still of interest. Even though many contributions exist, the characteristics of the problem vary depending on the school policies, the country (laws), and other particular variables.The complexity of this problem makes it difficult to find anoptimal solution, so approximated techniques are traditionally used in practice. In this paper, we focus in the Spanish school timetabling problem and present several approaches to deal with it. The first technique proposed is based on the random non ascendent method (RNA). Then we provide several genetic algorithms which differ on the  policies used for selecting how the next generation is created (including  elitism) as well as on the levels of mutation considered. Finally, we study how to combine the two previous approaches. We run experiments both on synthetic and real scenarios in order to compare all the proposals. Even though the RNA and some of the pure genetic algorithms obtain good results in practice, we show that by joining RNA with genetic algorithms we gain stability in the results.