Groupe d’études et de recherche en analyse des décisions

G-2016-92

Reduced variable neighborhood programming for the preventive maintenance planning of railway infrastructure

, et

In this paper we present our solution for the Challenge problem publicly announced by Railway Application Section (RAS), which operates within INFORMS. Variables in the problem represent decisions should the each particular railway truck segment be preventively maintained or replaced. The huge training data set is provided containing several attributes of each track segment, including the right decision regarding maintenance. Correct decisions are removed from the testing set. Research teams were competed who would have less wrong decisions on testing data. In order to solve this problem, we divide it into two phases: prediction and classification. Both phases are solved by using rules of recently proposed automatic programming method called Variable neighborhood programming. It appeared that our team was among awarded.

, 15 pages