【主 题】Regularized Dimension Reduction with Applications to Call Center Data
【时 间】 2017年04月13日（星期四）16:00－17:00
【地 点】 上海财经大学统计与管理学院大楼1208室
【摘 要】Massive data have brought forth many new and exciting areas of scientific inquiry, particularly in the field of statistics. In this talk, I will introduce a couple of novel dimension reduction methods for massive data analysis, including a supervised principal component analysis (PCA) framework and a nonparametric two-way hazards model.
The supervised PCA is motivated by applications where the low rank structure of the data of interest is potentially driven by auxiliary variables measured on the same set of samples. The proposed method can make use of the information in the additional data to accurately extract underlying structures that are more interpretable. We also extend the framework to accommodate special features of data such as high dimensionality and smoothness through regularization. We demonstrate the advantage of the method using call center arrival rate data.
The two-way hazards model is motivated by the business analytics application of modeling hazard functions of call center customer waiting times. Such waiting times reveal important patterns of customer patience (how long a customer is willing to wait) and offered wait (how long a customer is required to wait) which are closely related to customer satisfaction and service quality. We develop a nonparametric model that can recover a low-rank and smooth hazard surface. The estimated surface is highly interpretable and can capture continuous patterns over waiting time across different times of day. We demonstrate the advantage of the method using call center waiting time data.