企业创新项目组合绩效预测的贝叶斯网络模型研究

发布时间：2018-04-10 20:23

本文选题：创新项目组合 + 绩效预测　；参考：《合肥工业大学》2017年硕士论文

【摘要】：为了维持长期的竞争优势,企业常常并行开展多个创新项目以便最大程度地利用资源,同时也使得创新项目组合管理的重要性逐渐凸显。创新项目组合的成功可以通过创新项目组合管理完成项目组合战略目标的程度即创新项目组合绩效来呈现。要保证有着较高创新项目组合绩效就需要对创新项目组合管理的整个过程进行管控,及时地发现影响成功的关键因素。创新项目组合的成功受到高管参与、项目经理胜任力、项目组合管理流程等多因素的影响,且影响因素间不仅存在线性关系,也可能存在复杂的非线性因果关系。然而,传统的绩效预测方法大部分只考虑影响因素间的线性因果关系或者本身就存在缺陷。鉴于贝叶斯网络建模方法常被用来模拟变量间的复杂因果关系并进行绩效的预测,本文将使用贝叶斯网络对创新项目组合绩效进行预测,然而,贝叶斯网络在区分因果关系和虚假关系方面却有着不足。为了解决这种不足且提高预测准确率,本文建立一个含潜变量的贝叶斯网络模型,并提出创新项目组合绩效预测的方法框架。首先,通过问卷调查识别创新项目组合绩效的影响因子,建立项目组合绩效的传统贝叶斯网络结构;然后,基于贝叶斯网络图中变量间的因果关系构建变量间的假设因果关系模型,采用偏最小二乘法进行模型参数估计,并根据路径系数的显著性识别出虚假的因果关系,形成新的贝叶斯网络模型;最后,通过参数学习预测创新项目组合绩效。此外,将本文所提出的方法应用于收集的169份企业样本数据,并使用了十折交叉法计算了人工神经网络、树扩展的贝叶斯分类器结构学习的网络结构、K2算法结构学习的网络结构以及偏最小二乘法优化后的贝叶斯网络结构对创新项目组合绩效预测的平均准确率,结果表明了本文所提出网络结构有着较高的准确性和稳定性。
[Abstract]:In order to maintain long-term competitive advantage, enterprises often carry out parallel multiple innovation projects to maximize the use of resources, but also makes the importance of the innovation of the project portfolio management gradually. The innovation project portfolio success can complete the project portfolio strategy through innovation project portfolio management level innovation project portfolio performance to ensure a presentation. High innovation project portfolio performance is needed to control the whole process of innovation of project portfolio management, the timely discovery of key factors affecting the success of innovation. The success of the project portfolio is high in tube, the project manager competency, many factors affecting the project portfolio management process, not only the linear relationship exists and influence factors, may also exist the complex nonlinear causal relationship. However, the traditional performance prediction methods only consider the most linear among the influencing factors The relationship between fruit or itself is defective. Given the predicted Bias network modeling method is often used to simulate the complex causal relationships between variables and performance, this paper will use the Bias network to predict the innovation project portfolio performance however, Bias network in distinguishing causality and false relations are insufficient to solve this problem. And improve the accuracy of prediction, this paper establishes a includinglv Bias network model, and puts forward the framework of innovation project portfolio performance prediction method. Firstly, through the influence of questionnaire identify innovative project portfolio performance factor, the traditional Bias network to establish a project portfolio performance; then, the causal relationship between variables of the Bias network diagram construction of the causal relationship between the variables hypothesis model based on model parameters were estimated using partial least squares method, and according to the path coefficient The significance of identifying false causality, Bias formed the new network model; finally, predict the innovation project portfolio performance through learning the parameters. In addition, 169 enterprises sample data of the proposed method is applied to the collection, and using ten fold cross method of artificial neural network calculation, the extended Bias tree classifier structure learning of the network structure, the average accuracy of Bias network structure K2 algorithm to learn the structure of the network structure and partial least squares optimization prediction on the innovation project portfolio performance, results show that the proposed network structure with high accuracy and stability.

【学位授予单位】：合肥工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F273.1;F224

【相似文献】