Space engineering risk analysis from risk assessment matrix using text mining

Abstract

Risk Assessment Matrix(RAM) is a traditional method used for risk analysis of space engineering. The essential aspect in RAM is to identity the risk event and classify the risk level. Normally, this work is absolutely dependent on judgement of experts. But in the past 50 years, there are huge risk text in RAM structure which contain valuable knowledge about risk analysis. And with the arising era of big data, the risk analysis certainly will step into an intelligentized age. In this paper, we proposed a text mining system for risk event recognition and risk level classification intelligently. Our system has two process. The first one identifies the features of risk event based on a BIO format. The second one relies on machine learning techniques to map the relationship between features using a composite kernel-based method which consists of a shallow linguistic kernel and an extended dependency tree kernel function. A Supported Vector Machines is used in data train and test. The obtained results are encouraging after comparing the system with other methods, such as KNN, Naive Bayes.