呼叫中心智能排班系统关键技术

摘要

The effects of learn ratio and training precision on the maximum training number and convergence time of BP neural network,as well as the accuracy of telephone traffic prediction were analyzed.According to the daily periodic characteristics of history telephone traffic,the idea of using several BP neural network models separately in different time sections of the day im-proves the precision of telephone traffic significantly was proposed and testified.Based on the predicted telephone traffic and Er-lang-C formula,the agent number was calculated.The agent curve fitting was carried out by using typical classes and the as-sumed service level,which calculated the number of telephone representative.The call center intelligent scheduling system was developed based on J2EE,the rational allocation of call center resources was achieved through reasonable scheduling.%分析学习率和训练精度对BP神经网络训练最大次数、收敛时间和话务量预测精度的影响；根据呼叫中心历史话务量数据的日变化特点，提出并验证采用分时间段多次调用BP神经网络模型的方法比整体预测所得话务量预测结果精度更高；基于话务量预测结果，使用Erlang-C公式进行坐席数预测，结合呼叫中心的典型班次、设定的服务水平等参数进行坐席数曲线拟合，得到每个典型班次对应的话务员数量；开发呼叫中心智能排班系统，通过合理的排班实现大型呼叫中心资源的合理配置。