主管:中国科学院
主办:中国优选法统筹法与经济数学研究会
   中国科学院科技战略咨询研究院

Chinese Journal of Management Science ›› 2024, Vol. 32 ›› Issue (6): 151-162.doi: 10.16381/j.cnki.issn1003-207x.2021.1861

Previous Articles     Next Articles

Dynamic Selection of New Product Development Project Portfolio Based on Online Learning with Uncertain Revenue Information

Chao Fang,Yajing Hu,Weibo Zheng(),Gengzhong Feng   

  1. Schoolof Management/The Key Lab of the Ministry of Education for Process Management & Efficiency Engineering,Xi’an Jiaotong University,Xi’an 710049,China
  • Received:2021-09-13 Revised:2022-03-02 Online:2024-06-25 Published:2024-07-03
  • Contact: Weibo Zheng E-mail:zhengweb@mail.xjtu.edu.cn

Abstract:

Because of the existence of revenue uncertainty, enterprises often face the dilemma of making accurate decision when selecting new product development (NPD) project portfolio. Most of the existing project portfolio selection models are static that optimize the project portfolio in advance before the period of development, and hence cannot provide decision makers with management flexibility to handle the uncertain. A more common phenomenon is that an enterprise gradually learns the real situation of project revenue and make dynamic adjustments to the project portfolio during the process of project development. Therefore,in this paper an online learning method is employed to study the dynamic selection of NPD project portfolio. Firstly,in view of the dynamics of development process and the uncertainty of revenue information, an optimization model containing three decision types such as project selection, continuation and abandonment is established to maximize the cumulative net profit of the selected project portfolio. Secondly, referring to the idea of the multi-armedbandit in online learning, the value function is reconstructedby transforming the objective function of the original model. Bayesian update is used to realize the online learning of the revenue information by decision makers, and then the dynamic selection policy of project portfolio is proposed. The policy gives the rules of dynamic selection of NPD project portfolio to maximize the value of the projects. Thirdly, in the part of case analysis, the proposed dynamic model is compared with the traditional static model; the influence of decision-makers' risk preference on the objective function in online learning is also explored;and the dynamic selection results of project portfolio under different risk preference are analyzed.Finally, the sensitivity analysis is conducted on the uncertainty degree of project portfolio revenue information. The results show that the dynamic selection policy of project portfolio based on online learning can improve the resource allocation efficiency between projects with different revenues, and reasonable risk preference can improve the cumulative net profit of project portfolio. Compared with the static model, the improvement rate of net profit of the dynamic model is positively correlated with the uncertainty of project revenue information. This study can provide decision support for the dynamic selection ofproject portfolio and upgrade the enterprises’ NPD investment strategy.

Key words: project portfolio dynamic selection, online learning, information uncertainty, new product development

CLC Number: