A Re-solving Heuristic for Dynamic Assortment Optimization with Knapsack Constraints

Xi Chen,Mo Liu,Yining Wang,Yuan Zhou

2024-07-08

Abstract:In this paper, we consider a multi-stage dynamic assortment optimization problem with multi-nomial choice modeling (MNL) under resource knapsack constraints. Given the current resource inventory levels, the retailer makes an assortment decision at each period, and the goal of the retailer is to maximize the total profit from purchases. With the exact optimal dynamic assortment solution being computationally intractable, a practical strategy is to adopt the re-solving technique that periodically re-optimizes deterministic linear programs (LP) arising from fluid approximation. However, the fractional structure of MNL makes the fluid approximation in assortment optimization highly non-linear, which brings new technical challenges. To address this challenge, we propose a new epoch-based re-solving algorithm that effectively transforms the denominator of the objective into the constraint. Theoretically, we prove that the regret (i.e., the gap between the resolving policy and the optimal objective of the fluid approximation) scales logarithmically with the length of time horizon and resource capacities.

Optimization and Control,Machine Learning

What problem does this paper attempt to address?

The paper primarily focuses on addressing the multi-stage dynamic assortment optimization problem, particularly in the context of the multinomial logit (MNL) choice model under resource knapsack constraints. Specifically, the goal of the paper is to select a subset (i.e., "assortment") from a given set of products to offer to customers, in order to maximize the total profit that the retailer can obtain from these sales activities.

Since precisely solving the optimal dynamic assortment is computationally challenging, the paper proposes a re-solving strategy that periodically re-optimizes a deterministic linear program (LP) derived from a fluid approximation. However, the fractional structure in the MNL choice model makes the assortment optimization in the fluid approximation highly nonlinear, presenting new technical challenges.

To address this challenge, the authors propose a new cycle-based re-solving algorithm that effectively transforms the denominator of the objective function into part of the constraints. Theoretical analysis shows that the regret (i.e., the gap between the proposed strategy and the fluid approximation optimal solution) grows logarithmically with the length of the time horizon and the resource capacity.

The key contributions of the paper can be summarized as follows:

1. **Design of a New Algorithm**: A new cycle-based re-solving algorithm is designed, which simplifies the problem by transforming the denominator of the fluid approximation's objective function into constraints and uses a cycle-based strategy to overcome the nonlinearity of the objective function.

2. **Theoretical Results**: It is proven that the regret between the proposed strategy and the fluid approximation optimal solution grows logarithmically with the length of the time horizon and the resource capacity. This is significantly better than the \(O(\sqrt{T})\) regret achieved in previous work.

3. **Practical Applications**: The work not only makes theoretical advancements but also proposes practical methods to handle the dynamic assortment optimization problem, which has broad application prospects in fields such as online retail and advertising.

In summary, this paper addresses the problem of dynamic assortment optimization using the MNL model under resource constraints and proposes an effective re-solving strategy that significantly reduces the gap with the fluid approximation optimal solution.