Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Learning policies for resource allocation in business processes

J. Middelhuis, R. Lo Bianco, E. Scherzer, Z. A. Bukhsh, I. J. B. F. Adan, R. M. Dijkman

TL;DR

Two learning-based methods for resource allocation in business processes to minimize the average cycle time of cases are proposed, one of which leverages Deep Reinforcement Learning (DRL) to learn policies by allocating resources to activities and another is a score-based value function approximation approach.

Abstract

Efficient allocation of resources to activities is pivotal in executing business processes but remains challenging. While resource allocation methodologies are well-established in domains like manufacturing, their application within business process management remains limited. Existing methods often do not scale well to large processes with numerous activities or optimize across multiple cases. This paper aims to address this gap by proposing two learning-based methods for resource allocation in business processes to minimize the average cycle time of cases. The first method leverages Deep Reinforcement Learning (DRL) to learn policies by allocating resources to activities. The second method is a score-based value function approximation approach, which learns the weights of a set of curated features to prioritize resource assignments. We evaluated the proposed approaches on six distinct business processes with archetypal process flows, referred to as scenarios, and three realistically sized business processes, referred to as composite business processes, which are a combination of the scenarios. We benchmarked our methods against traditional heuristics and existing resource allocation methods. The results show that our methods learn adaptive resource allocation policies that outperform or are competitive with the benchmarks in five out of six scenarios. The DRL approach outperforms all benchmarks in all three composite business processes and finds a policy that is, on average, 12.7% better than the best-performing benchmark.

Learning policies for resource allocation in business processes

TL;DR

Two learning-based methods for resource allocation in business processes to minimize the average cycle time of cases are proposed, one of which leverages Deep Reinforcement Learning (DRL) to learn policies by allocating resources to activities and another is a score-based value function approximation approach.

Abstract

Efficient allocation of resources to activities is pivotal in executing business processes but remains challenging. While resource allocation methodologies are well-established in domains like manufacturing, their application within business process management remains limited. Existing methods often do not scale well to large processes with numerous activities or optimize across multiple cases. This paper aims to address this gap by proposing two learning-based methods for resource allocation in business processes to minimize the average cycle time of cases. The first method leverages Deep Reinforcement Learning (DRL) to learn policies by allocating resources to activities. The second method is a score-based value function approximation approach, which learns the weights of a set of curated features to prioritize resource assignments. We evaluated the proposed approaches on six distinct business processes with archetypal process flows, referred to as scenarios, and three realistically sized business processes, referred to as composite business processes, which are a combination of the scenarios. We benchmarked our methods against traditional heuristics and existing resource allocation methods. The results show that our methods learn adaptive resource allocation policies that outperform or are competitive with the benchmarks in five out of six scenarios. The DRL approach outperforms all benchmarks in all three composite business processes and finds a policy that is, on average, 12.7% better than the best-performing benchmark.
Paper Structure (19 sections, 3 equations, 8 figures, 5 tables)

This paper contains 19 sections, 3 equations, 8 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related work
  3. Preliminaries
  4. Business process models
  5. Business Process Execution
  6. Resource allocation problem
  7. Method
  8. Deep Reinforcement Learning
  9. Score-based value function approximation
  10. Evaluation
  11. Experimental setup
  12. Training setup
  13. Deep Reinforcement Learning
  14. Score-based value function approximation
  15. Evaluation protocol and benchmarks
  16. ...and 4 more sections

Figures (8)

  • Figure 1: An example of a business process model of a loan application process with two activities and two resources. Cases arrive according to some distribution with rate $\lambda$, and the loans are either accepted (AA) or rejected (RA). Each activity can be performed by both resources $r_1$ and $r_2$, according to a processing time distribution with mean $\mathbb{E}(X_{i})$.
  • Figure 2: Agent’s perspective of the agent-environment interaction, adapted from sutton_reinforcement_2018.
  • Figure 3: Six business process models with different characteristics used for evaluation.
  • Figure 4: Arrival pattern based on the BPI12W event log
  • Figure 5: Total reward and mean cycle time during training of DRL models.
  • ...and 3 more figures

Theorems & Definitions (15)

  • Definition 1: Event
  • Definition 2: Trace
  • Definition 3: Activity instance
  • Definition 4: Unassigned Activity Instance
  • Definition 5: Available Resource, Unavailable Resource
  • Definition 6: Resource Eligibility
  • Definition 7: Business Process Execution State
  • Definition 8: Assignment
  • Definition 9: State Transition
  • Definition 10: State Transition Function
  • ...and 5 more