An Optimization Method On Service Composition For Personalized Requirements

With the globalization of economy, the total international trade of china grows steadily. As the most important transportation means, the marine logistics have been greatly promoted and developed. The marine logistics is a complex business which is made up of multiple basic functions. To meet the different demands for each customer, it should offer a feasible scheme by integrating multiple resources.This article analyzed and researched the marine logistics process by serviceoriented theory. The service will be described in two aspects, which is service function and Qo S. In reality, the customers can only make macro, vague and uncertain personalized requirements, but they don’t provide specific service function features and Qo S. It has became a difficult problem that how to find service function features from customers’ personalized requirements. In addition, finding a suitable Qo S-based service selection method to reduce the cost of service delivery has been the hot issue of current research. This thesis has studied above problems from the following two respects:Firstly, according to the service feature and customers personalized requirements feature, the marine logistics service and corresponding customers are described. The initial Bayesian network model is established by using the both sides’ features as the node. Bayesian estimation is used to learn network parameters from historical data. Using the Bayesian scoring combined with K2 algorithm or artificial bee colony algorithm to learn network structure, the sophisticated Bayesian Network will be got. It has been verified the effectiveness of Bayesian network in finding the service features from customers personalized requirements feature in marine logistics.Secondly, a multi-objective optimization problem mathematical model for Qo S-oriented service selection is established. In dealing with multiple objectives, if there is a target component weight, the optimization problem is transformed into a single objective problem based on the aggregate functions. An improved single objective artificial bee colony algorithm is designed to find the optimal solution. On the contrary, the improved multi-objective artificial bee colony algorithm based on Pareto dominant mechanism is designed to find all of the non-dominated solutions.Finally, the effectiveness of the improved algorithm is verified under maritime logistics.