Research And Application Of Query Dispatch Algorithm Based On Multi-objective Optimization

With the rapid development of China's economy, the product sales have been in the increasingly fierce competition, the enterprise's marketing emphasis is shifting from product, channel, price and brand to service gradually, which providing customers with exact answer timely is the key of the enterprise to provide service. Query process and dispatch technologies are the key factors that affect the efficiency of query processing, at present, the query processing technology has been basically mature, but the customer's satisfaction of query service still has a large space to improve, the main reason is that the technology of query dispatch needs to be improved.At present, query dispatch technology mostly combines VIP priority with queuing,namely in the case of VIP is preferred, queries line up and wait to be answered according to their arriving time. Obviously, the strategy can not meet the needs of enterprises to provide personalized service and adjust skill groups. In view of the above questions, this article embarks from the characteristics of customer query service industry and studies the existing query dispatch policies and procedures systematically. Then base on genetic algorithm, neural network and markov decision making process, this paper proposes a new query dispatch model, and implements the model in the customer query processing system eventually. In this paper, the research content and innovation points are as follows:1.The new model of query diapatch is established, which includes the definition of terms and the model, some decision making algorithms in the process of query dispatch are listed.2.The optimization system of dispatch strategy in query dispatch system is given. Firstly,objectives of multi?object optimization are selected. According to the different characteristics of each optimization objective, the objective function is established. Finally,the multi?objective optimization algorithm based on genetic algorithm and neural network is proposed.3.The Markov decision?making process of query dispatch is realized, the query dispatch process is mapped into Markov decision?making process creatively, and the incentive mechanism of Markov decision?making process is used to complete the reinforcement learning of query dispatch decision making.4.The query dispatch system is developed, and then applied to a Mobile Corporation's customer query processing system, the main function is dispatching customer queriesaccording to optimize results. The application results show that, the query dispatch system effectively improves the customer satisfaction, the operator satisfaction and the efficiency of the customer query processing system.5.On the basis of the above technology, this paper designs and develops a set of new customer query processing system, including the message receiving and sending module,the natural language understanding module, the telephone traffic dispatching module and so on.The new customer query processing system test results show that the model can not only meet the needs of enterprises to provide personalized service and adjust skill groups,but also effectively improve satisfaction of customer and operator, reduce the cost of customer service department. The results of its application have been highly recognized by the mobile company.