基于网络聊天的服务中心的建模与运作管理

Many companies have been building a new type of service centers, where customers receive real-time service by chatting with agents via instant messaging over the Internet. Compared with the traditional phone call centers, an agent in web-chat based service centers can serve multiple customers simultaneously, and customers may need to wait for an available agent before entering the server pool or wait for the response from the agent after starting their conversations. In that situation, customers may abandon from the queue as out of patience or renege even after receiving their service as feeling a slow response. How to accurately model such a service center by providing an optimal staffing level and control policy in order to achieve certain service qualities is the key issue we would like to address in this project. More specifically, this project focuses on the following three problems: (1) To offer an optimal staffing and control policy, we first need to quantitatively describe a web-chat service system with customer abandonment using a many-server queue model with games. (2) As different service channels, from phone-call to web-chat even to email, it is promising to build a multifunctional service center with lower operating costs compared with separate service centers with different service channels. (3) Because it could be difficult to model such a complicated stochastic system, then simulation tools would be helpful to analyze a multifunctional service center without understanding the underlying stochastic dynamics. It is also an interesting research direction to find the optimal policy using optimization via simulation techniques, especially in parallel computing environments.

随着互联网的快速发展，一种以网络聊天为基础的新型服务系统正逐渐被各大公司所使用。与传统的电话服务相比，网聊系统中的客服代表可以同时与多个顾客聊天；而顾客所面临的等待不仅可能发生在接受服务之前，还可能在开始服务之后。这就意味着，顾客可能在等待进入系统时失去耐心而离开系统，或者在接受服务的过程中得不到及时的回复而放弃完成服务。对这样的服务系统如何准确建模、并在保证顾客满意度的前提下合理安排客服人员以及设计有效的服务机制，是本项目要研究的主要问题。具体来讲，本项目着重研究：(1)在考虑顾客离开行为的情况下，如何用排队理论和博弈模型来量化描述此网聊系统、并给出最优的排班和服务策略；(2)如何建立一个多服务渠道的综合性系统，包括电话客服、网聊客服甚至电邮客服，以节约运营成本并保证一定的服务质量；(3)在随机模型无法求解的情况下，如何用仿真优化的方法来刻画此大型服务系统，并在并行计算的环境下模拟运算。

本项目研究了互联网高速发展背景下所兴起的以网络聊天为基础的新型客服系统。与传统的电话服务相比，网聊系统中的客服代表可以同时与多个顾客聊天；而顾客所面临的等待不仅可能发生在接受服务之前，还可能在开始服务之后。对这样的随机服务系统如何准确建模、并在保证顾客满意度的前提下如何合理安排客服人员以及设计有效的服务机制，是本项目的主要研究内容。具体来讲，本项目通过Markov矩阵模型和仿真优化方法，分别从微观角度对在线客服系统进行建模，以及在仿真建模之后从优化算法角度去寻求系统设计运作的最优策略。相关的研究成果发表在《Operations Research》、《INFORMS Journal on Computing》和《Naval Research Logistics》等国际国内学术期刊上。研究成果揭示了在线客服系统的特性和理论分析难点，为企业在建立在线客服系统时提供了理论指导。研究所建立的随机模型和优化算法，不仅仅适用于在线客服系统，还能够轻易拓展到其他研究问题，比如医疗系统中的医生问诊病人、护士管理病床问题；比如基于仿真建模的其他大型复杂随机系统的优化决策问题。