Data Analysis of R&D Partnerships to Design a Decision-Making Support System for Use in Alliance Formation

这项研究的主要目标是通过开发数学和统计方法来创建一个决策支持系统，供加拿大半导体公司在研发(R&D)联盟决策中使用，从而为应用统计和优化做出贡献。为了实现这一目标，利用回归分析，拟议研究项目的第一阶段将揭示创新成功的决定因素，并在有监督的学习环境中利用这些决定因素创建一个数学模型，用于评估加拿大半导体行业创新成功的不确定性。在模型的开发中，将使用贝叶斯信息标准和Akaike信息标准来选择变量，并将检验使用特殊回归形式(如套索或岭)的潜在好处。利用第一阶段的结果，第二阶段的目标是开发一个优化规划模型，其中的决策变量是第一阶段中发现的具有统计意义的决定因素；其中的目标函数返回半导体行业内研发伙伴关系创新成功的概率。建议的研究项目对预期的研究领域以及加拿大有以下三个主要贡献：(1)与以前的许多研发伙伴选择方法不同，我打算开发的方法是更完全的数据驱动方法，注重了解预测的准确性；(2)建议的研究明确考虑了加拿大半导体行业最近面临的挑战和变化的趋势；(3)建议的研究还试图创建研发合作伙伴选择的优化模型和有效的算法来解决相应的优化问题。来自加拿大政府的数据和信息表明，加拿大半导体业最近面临多重挑战，包括快速创新和外国竞争。这些信息还表明，随着越来越多的公司以“无芯片”的方式扩张，该行业的业务经历了实质性的变化，这些公司专注于只创造半导体的知识产权(而不是制造半导体本身)。最近物联网(IoT)的趋势刺激了对行业进一步创新的预期。为了帮助公司更多地专注于研发，拟议工作的预期结果将是开发一个决策支持系统，以帮助加拿大境内的行业选择研发合作伙伴。通过指导加拿大半导体公司平稳过渡到更注重研究的环境，拟议的工作可以帮助该行业有效地应对最近的挑战。

The main objective of this research is to contribute to applied statistics and optimization by developing mathematical and statistical methodology for the creation of a decision support system that could be used by Canadian semiconductor companies in research and development (R&D) alliance decisions. For this objective, using regression analysis, Phase 1 of the proposed research project will reveal determinants of innovation success and leverage these determinants in a supervised learning environment to create a mathematical model for evaluating the uncertainty of innovation success in the Canadian semiconductor industry. In the model's development, the Bayesian information criterion and the Akaike information criterion will be employed for variable selection, and the potential benefits of using specialized forms of regression, such as LASSO or ridge, will be examined. Using the results of Phase 1, the aim of Phase 2 will be to develop an optimization programming model, of which the decision variables are the statistically significant determinants found in Phase 1; and of which the objective function returns the probability of innovation success for R&D partnerships within the semiconductor industry. The proposed research project has the following three major contributions to the intended research fields as well as Canada: (1) Unlike many previous R&D partner selection methods, the methodology that I intend to develop is a more fully data-driven approach with attention to understanding the predictive accuracy; (2) The proposed study explicitly takes into account the recent challenges and changing trends faced by the Canadian semiconductor industry; and (3) The proposed study also seeks to create an optimization model for R&D partner selection along with an efficient algorithm to solve the corresponding optimization problem. Data and information from the Government of Canada indicates that the Canadian semiconductor industry has recently faced multiple challenges, including rapid innovation and foreign competition. The information also indicates that the industry has experienced substantial changes to its operations as more companies are expanding in "chipless" manners, in which the firms are focused on creating only the intellectual property for semiconductors (and not in manufacturing the semiconductors themselves). The recent trend of the Internet of Things (IoT) spurs the expectation for further innovation in the industry. Aiding firms in this transition to a heavier focus on R&D, the anticipated outcome of the proposed work will be the development of a decision support system to assist the industry within Canada in its R&D partner selection choices. By guiding Canadian semiconductor companies in making a smooth transition to a more research-oriented environment, the proposed work can help the industry respond effectively to its recent challenges.