Thesis (BS Applied Mathematics) -- University of the Philippines Mindanao, 2009

Nurse scheduling problem (NSP) involves producing daily schedules for nurses over a given time horizon, considering hospital policies which must be satisfied to obtain feasible schedules. Shuffled frog leaping algorithm (SFLA) is a population-based search algorithm where a set of frogs is partitioned to memeplexes wherein local searches are performed. There is no found literature which applied SFLA to NSP. This study explored the applicability of SFLA to a NSP where the PSO-based local search was modified by using the genetic algorithm operators: uniform crossover and violation-directed mutation. A modified shuffled frog leaping algorithm (MSFLA solution representation was formulated that fits the nurse scheduling problem in Davao Medical Center, a government tertiary hospital in Southern Mindanao. Nurse aid and nurse schedules were separately represented. A fitness function was developed which minimizes the penalties obtained by a schedule. Parameters were set to 30 individuals in a population, 6 memeplexes, 10 memeplex iterations, 1000 shuffling iterations, 100% crossover and mutation occurrence rates and 2% mutation rate. The MSLA produced feasible schedules but it failed to give the required number of day-off, did not distribute shifts fairly to nurses and violated the allowable conservative shift types. The schedules generated by the MSFLA were compared to the schedules done manually and by the Global Programming (GP) method of Sebastian (2007). The schedules generated by the MSFLA are better than the manual method. The GP method outperformed it but the nurse schedule of this method is not feasible.