Estimation of load traffic and income in Mersin Port by artificial neural networks and Monte Carlo simulation

The new feasibility analysis model L MF Z developed for coastalprojects consists of three interrelated sub-models: 1) Artificial NeuralNetwork (ANN) to determine the rates and capacity of cargo byconsidering the economical development of hinterland 2) Queuingmodel to determine the waiting to service time and the berth occupancyratios by waiting time modeling of ships using discrete queuingsimulation 3) Importance Sampling Monte Carlo (ISMC) to simulate shiparrivals/departures from the quays and to estimate income/expenditureparameters of the coastal project. In this work as a case study, theproposed model was applied to the Mersin Port in Turkey and the futureloading/unloading cargo rates of piers were predicted by ANN?s. Theload traffic of Mersin Port, which is placed on 36o46?20?? north latitude,and 36o39?00?? east longitude, was estimated by using artificial neuralnetworks. The harbour?s management rights were transferred byprivatization for 36 years for a price 775 million dollars. The annualrevenue expected from the loading-unloading operations wasdetermined. Mersin Port and its surroundings were examined, and alsoinformation about its hinterland was presented. Historical load data,gross national product, population data of harbour were utilized. The2number of days for which Mersin Port is out of work were estimated bylong term wave statistical analysis. Load volume of Mersin Port up tothe year of 2030 was predicted by feed forward artificial neuralnetworks (FFANN). Capacity rate of Mersin port was also calculatedwith that estimated loads. A new model that uses both FFANN andMonte Carlo simulations was developed. Using Monte Carlo method, bymultiple recurring similarities, arbitrary values from probabilitydistributions were appointed to financial variables and then the randomincome estimations were made from probability distributions ofloading-unloading activity of the management. During management ofthe port, results obtain from frequency distribution of 30 000 tests of allvariables were used to confirm the expected income of port and theexpected annual income was calculated.Science Code : 624.02.03Key Words : Artificial intelligence, Monte Carlo Simulation,Feasibility Studies, Traffic of Mersin PortPage Number : 124Adviser : Assoc. Prof. Dr. Can Elmar BALAS