4 Presentations

• 
  10:30 AM - 10:55 AM

  Asymmetry in the Complexity of the Multi-Commodity Network Pricing Problem

  • Quang Minh Bui, presenter, Université de Montréal
  • Margarida Carvalho, Université de Montréal
  • José Neto, Télécom SudParis

  The network pricing problem (NPP) is a bilevel problem, where the leader optimizes its revenue by deciding on the prices of certain arcs in a graph, while expecting the followers (also known as the commodities) to choose a shortest path based on those prices. We investigate the complexity of the NPP with respect to two parameters: the number of tolled arcs, and the number of commodities. We devise a simple algorithm showing that if the number of tolled arcs is fixed, then the problem can be solved in polynomial time with respect to the number of commodities. In contrast, even if there is only one commodity, once the number of tolled arcs is not fixed, the problem becomes NP-hard. We characterize this asymmetry in the complexity with a novel property named strong bilevel feasibility. Finally, we describe an algorithm to generate valid inequalities to the NPP based on this property, accommodated with numerical results to demonstrate its effectiveness in solving the NPP with a high number of commodities.
• 
  10:55 AM - 11:20 AM

  Diverse Assortments in Online Recommendations

  • Mahsa Hosseini, presenter,
  • Opher Baron, Rotman School of Management, University of Toronto
  • azarakhsh malekian,
  • Shreyas Sekar,

  This paper explores the business benefits of incorporating diverse content into recommendation systems, particularly in online marketplaces and platforms that rely on ranking algorithms. While promoting popular content in the short term may improve user engagement, it can also lead to a lack of diversity and ultimately undermine the long-term health of the marketplace. In this paper, we make a connection between customer engagement and satiation. Satiation examines how past consumption patterns determine consumers’ willingness to pay for goods/services. We find conditions under which a more diverse policy is optimal in this setting. Moreover, we show that optimal policy favors diversity in earlier periods, and after some time, the optimal policy keeps offering the product with the largest value. This implies that a more diverse recommendation set is preferred over longer time horizons. Also, we show that we have a satiation threshold policy, determining the optimal policy for any satiation value at every period.
• 
  11:20 AM - 11:45 AM

  Omnichannel Assortment Optimization with Two-stage Decisions

  • Venus Lo, presenter, City University of Hong Kong

  Many retailers operate both physical and online stores. Since retail space is expensive, a retailer may offer a full assortment of products in his online store and a subset in his physical store. Consider the assortment optimization problem faced by an omnichannel retailer when customers may purchase from either stores and make decisions over two stages. In the first stage, customers arrive to the physical store and purchase from the smaller in-store assortment. Customers who do not purchase will examine the in-store assortment before they proceed to the online store in the second stage. When these customers browse the online assortment, their preferences depend on their examination of similar products in the physical store. Our goal is to select an in-store assortment that maximizes the retailer’s total expected revenue over both channels and stages. Suppose customers choose according to the independent demand model in the first stage and an extended multinomial logit model in the second stage, such that the preference weights depend on the in-store assortment. The assortment optimization problem is NP-hard, and I present a fully polynomial-time approximation scheme (FPTAS). Numerical experiments show that the FPTAS has very good performance and can be used to identity good assortments.
• 
  11:45 AM - 12:10 PM

  A comparative analysis of learning-based algorithms for dynamic pricing in smart freight platform

  • Shiri Ali, Polytechnique Montréal
  • Samira Keivanpour, presenter, Polytechnique Montréal

  Smart transportation platforms are becoming increasingly popular in the current market due to their convenience, reliability, and efficient transportation services. The platforms may utilize various technologies such as AI, and big data analytics to optimize transportation operations and reduce carbon emissions. The platform can help with fleet management, route optimization, and driver behavior monitoring to enhance fuel efficiency and reduce environmental impact.
  Well-organized planning through pricing can significantly cut costs and increase profitability in transportation platforms. However, there are several uncertainties associated with pricing. Shipping demand(supply) based on shippers' (carriers') preferences may change swiftly. Another uncertainty is competitor pricing, if that is changed regularly, it might require the platform to adjust prices accordingly. Therefore, the platforms set flexible pricing based on values of uncertainties.
  Many researchers have developed dynamic pricing models for ride-sharing and ride-hailing platforms. However, few studies are focused on the freight transportation sector. An effective dynamic pricing model for freight transportation platforms requires considering different uncertainties through interaction with the environment and learning methods. So, It is helpful to analyze the existing learning-based algorithms and models for dynamic pricing in ride-sharing and to identify the opportunities and challenges of their application in smart freight platforms.