4 Presentations

• 
  10:30 AM - 10:55 AM

  An ALNS to solve the time-dependent home health care problem

  • Florian Grenouilleau, presenter, CIRRELT
  • Louis-Martin Rousseau, Polytechnique Montréal
  • Nadia Lahrichi, Polytechnique Montréal

  Due to the increasing number of elderly, optimization of health resources becomes primordial. In our problem we study the Home Health Care problem and we implement a method using ALNS (Adaptative Large Neighborhood Search). This study, in collaboration with the AlayaCare company, attempts to take into account all business constraints such as: the time window for patients and nurses, time-dependent travel times or patient preferences.

• 
  10:55 AM - 11:20 AM

  Maxout networks to predict homecare patients hospital readmissions

  • Jonathan Vallée, presenter, AlayaCare

  We use maxout neural networks to predict when a patient is heading toward a hospital readmission early enough for care workers to take action and try to prevent it. Our results show that maxout networks are a promising alternative to manually engineered alerts based on vital sign thresholds.

• 
  11:20 AM - 11:45 AM

  Data-driven ALNS-based algorithm for home care routing problem

  • Antoine Legrain, presenter, Polytechnique Montréal
  • Nadia Lahrichi, Polytechnique Montréal
  • Louis-Martin Rousseau, Polytechnique Montréal
  • Jonathan Vallée, AlayaCare

  With the rise of the big data era, more information is available thus allowing to model precisely the reality. In home care delivery services, nurses roads need to be updated at least once a day to take into account any events such as new patients arrival or emergency patient visits. We propose a dynamic ALNS-based algorithm to adapt the current schedule to changes. The data-driven procedure will be detailed in this presentation.

• 
  11:45 AM - 12:10 PM

  Two new history-based delay predictors for call centers

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

  We propose two predictors to estimate the delay of customers in call centers. The first method predicts the delay by extrapolating the wait history of customers currently in queue. The second predictor uses an exponential smoothing of the delays of past customers conditional to the queue length.