Abstract: As an emerging type of service and value creation method, smart services driven by digital technology have gradually become an innovative development direction for service-oriented manufacturing. Companies utilize the vast amount of data generated by smart, interconnected products throughout their lifecycle, along with various big data analysis technologies, to form smart services driven by big data. These services play a significant role in enhancing customer satisfaction and increasing corporate performance.
When developing smart services, companies face the challenge of selecting a few from a multitude of potential smart services for further development. Solving this problem requires considering not only the decision-maker’s preferences but also the resource constraints of the company and various uncertainties, with the aim of maximizing the decision-maker’s preferences. If the chosen portfolio of smart services is not reasonable, it will inevitably lead to losses for the company.
Current research on smart service selection mainly utilizes multi-attribute decision-making methods to evaluate the weight of different services to support service selection. Portfolio decision analysis refers to the theory, methods, and practices of using mathematical models to help decision-makers select a subset of projects from a set of projects, taking into account preferences, related constraints, and uncertainties. The multilinear portfolio utility function can model richer decision-makers’ preferences in terms of multiple attributes, uncertain project outcomes, project interactions, and risk preference compared to the additive portfolio utility function. It also has the advantage of reducing the number of parameters that need to be identified to a linear level, thereby lowering the difficulty of applying the multilinear portfolio utility function. However, most existing research based on the multilinear portfolio utility function is based on deterministic information. In the context of smart service selection, the decision-makers’ preference information and resource information required for service selection often have uncertainties. How to conduct robustness analysis based on the multilinear portfolio utility function for these uncertainties in smart service selection requires further study.
The paper proposes a new approach for smart service portfolio selection based on the multilinear portfolio utility function. First, in the case of uncertain outcomes of smart services, when the utility evaluation information, decision-makers’ preference information, and resource requirements for smart services are deterministic, an optimization model based on the multilinear portfolio utility function will be constructed to solve for the optimal smart service portfolio. This includes the following steps: (1)obtaining utility evaluation information and resource requirements for smart services, (2)eliciting decision-makers’ preference information and determining the parameters of the multilinear portfolio utility function, (3)considering the computational complexity of the multilinear portfolio utility function, and transforming it into a general linear function, (4)taking into account the interaction relationships of substitutiveness and complementarity between smart services, as well as the interaction of required resources between smart services, and establishing various constraints for the optimization model, and then (5)based on the multilinear portfolio utility function and various constraints, establishing an optimization model. The above model is a mixed-integer linear programming model, which can be solved using a linear programming solver, and the solver used in this study is Gurobi.Furthermore, considering the uncertainties that may exist in decision-makers’ preference information, the utility of smart services, and the required resources, robustness analysis is conducted for the three types of uncertainties. In situations where preference information is uncertain, robustness analysis can be conducted by utilizing the obtained uncertain preference information and the hit-and-run algorithm. For situations where the utility of smart services is inaccurate, when the resources required for smart services are inaccurate, uncertainty sets for the utility of smart services and the resources needed for smart services should be constructed respectively, followed by robustness analysis. After portfolio optimization under the three scenarios, it is possible to obtain all possible portfolios of smart services and their frequencies of occurrence, and the frequency with which each smart service is selected, providing important references for decision-makers. Smart service portfolios and smart services with high occurrence frequencies can be considered as robust choices for decision-makers in subsequent decision-making; smart services with relatively low occurrence frequencies can be further analyzed in subsequent decision-making; smart services with extremely low occurrence frequencies can be considered as alternatives when resources are very abundant.
Finally, the proposed approach is applied to the problem of selecting a smart service portfolio for a heavy truck company. Through case analysis, it can be seen that this approach can quickly determine the optimal smart service portfolio based on known information. Moreover, it can provide multiple reliable portfolios and the selection status of each smart service under the circumstances where preference information is incomplete, and the utility and required resources of smart services are uncertain. Therefore, the proposed approach can effectively support decision-making in the selection of smart service portfolios.

Key words: smart services, multilinear portfolio utility functions, portfolio selection, robust analysis, portfolio decision analysis

摘要： 在进行智能服务开发时,企业往往面临从服务集合选择部分服务进行开发的问题。考虑智能服务结果的不确定性,以及决策者偏好信息和服务所需资源信息的不准确性,提出一种基于多线性组合效用函数的智能服务组合选择方法。首先,针对智能服务结果的不确定性,利用多线性组合效用函数构建组合选择目标函数,并在考虑服务之间资源交互的基础上,建立混合整数规划模型并求解最优化服务组合。然后,考虑决策者偏好、智能服务效用评定信息及所需资源信息不准确情形,进行鲁棒分析,将最优组合出现的频率及各个服务被选择的频率作为最终选择的重要参考。最后,以某重卡企业智能服务组合选择问题为例,验证了方法的可行性与有效性。

关键词: 智能服务, 多线性组合效用函数, 组合选择, 鲁棒分析, 组合决策分析