Prediction of queuing behaviour through the use of artificial intelligence

Sammanfattning: Companies are constantly trying to minimize queues and optimize staff schedules, therecent computerization of queue systems provide huge amounts of data and thus newpossibilities for this. The classic way of analyzing queues is through queue theory which ispurely mathematical but with the new large datasets it is possible to explore predictionsthrough Artificial neural networks.This study investigates if it is possible to use the data from queue systems to provide aprediction with Artificial neural networks that can be used to plan a schedule, and comparethe results with schedule optimization through queue theory.The tests are made with a nonlinear Autoregressive neural network with external input,trained with Levenberg-Marquardt backpropagation and compared with calculations of aM/M/s model queue system.The results from the Artificial neural network were positive and indicate that it could bepossible to use it for predicting the right amount of servers each day but further tests mustbe made since the amount of data provided for the training was to small.When comparing the results from the neural network from that of the queue theoryalgorithms it seemed that they complemented rather than competed with each other.