Research On R&D Project Risk Assessment With Belief Rule-base Inference Method And Ensemble Learning

The highly competitive market environment makes research and development(R&D)activities a key element in determining corporate status.In order to maintain the competitive advantage and achieve considerable development,many enterprises regard R&D projects as the focus of investment.However,due to the high uncertainty and high risk of R&D projects,the success rate of R&D projects is very low,and it does not reach the ideal return on investment.Therefore,it is crucial to effectively assess various risks in the life cycle of R&D projects,so as to provide scientific basis for subsequent risk control to promote the success of projects and realize the return on R&D investment.Belief rule-based inference method has been applied in the field of risk assessment in recent years due to its strong interpretability and high prediction accuracy.However,when constructing a risk assessment model for R&D project based on the belief rulebased inference method,the existence of many risk factors will lead to the problem of combination explosion in belief rule base.This research combines the belief rule-based inference method and ensemble learning,proposes an ensembled belief rule-based inference method for R&D project risk assessment in order to improve the accuracy of project risk prediction and overcome the problem of combination explosion in belief rule base.The proposed method in this paper mainly includes three parts.The first part is R&D project risk identification,risk factors are extracted from the three aspects of market risk,technical risk and organizational risk in the project life cycle,and project risk assessment results are measured according to project time,cost,quality and the achievement of desired product goals.The second part is construction and inference of belief rule-based system.By introducing the random subspace method in ensemble learning and combining with information gain,multiple low-dimensional data subsets are randomly extracted from the original data set with probability,and belief rule-based systems are independently constructed in the data subsets for risk inference.The third part is the information fusion strategy based on evidence reasoning rule.The evidence reasoning rule is used for the fusion of inference results in belief rule-based systems.The complementary information provided by the inference results in different belief rule-based systems can be effectively combined to obtain the final prediction result.Based on the proposed method,experiments were performed in the real data of R&D projects.The experimental results show that the method proposed in this paper can effectively assess R&D projects risks.And compared with a series of other commonly used methods,the proposed method has higher prediction accuracy,can more effectively quantify the impact of risk events,and has great application potential in the field of risk assessment.