3 Presentations

• 
  09:00 AM - 09:25 AM

  Prediction Waiting Time in Multi-Skill Call Centres with Callback Option

  • Mamadou Thiongane, presenter, Université de Montréal
  • Pierre L'Ecuyer, Université de Montréal

  We present methods to estimate the waiting time of a client when he arrives to the multi-skills call center with the goal to announce this estimated delay time to the client and possibly to offer a callback option to the client. The customer can thus make a more informed choice to either wait, give up, or ask to be called back. We built predictor by regression with the method of splines, neural networks and kriging method.

• 
  09:25 AM - 09:50 AM

  Modélisation d'arrivées d'appels à un centre d'appel par un modèle de Poisson à taux d'arrivées lognormales// Lognormal Based Rate Model for Modeling of Call Center Arrivals

  • Nazim Regnard, presenter, Université de Montréal

  Nous introduisons un modèle de Poisson constant par morceaux, dont le taux d'arrivée stochastique et lognormal est spécifié sous une forme souple et adaptée à la modélisation des temps d'arrivées d'appels à un centre d'appel. Une application sur données réelles sera présentée // We introduce a piecewise constant Poisson model, with a lognormal stochastic rate specified in a very flexible way, especially designed for the modeling of arrival times of calls to a call center. A application on real data will be presented.

• 
  09:50 AM - 10:15 AM

  Hierarchical Bayesian State-Space Model for Call Center Arrival Rate Forecasting

  • Nicolas Chapados, presenter, ApSTAT Technologies
  • Pierre L'Ecuyer, Université de Montréal

  We address the problem of forecasting call center arrival rates with a two-level hierarchical state-space model, which can account for both inter-day and intra-day persistence patterns. The model is capable of accommodating seasonalities at monthly, weekly and daily horizons, and deal with special events through the use of explanatory variables. A Bayesian framework is retained throughout, allowing short histories and rare events to be considered.