Application of Linguistic Clustering to Define Sources of Risks in Technical Projects

Abstract

Risk identification is adversely affected by the still existing definitional and applicational discrepancy regarding risks and other related notions, such as hazards and impacts. A paradigm shift is beginning to be in effect, proposing the preliminary identification of risk sources to ameliorate the aforementioned adversities. However, apart from identifying risk sources from the outset, the bulk of the already conducted project risk-related research, from which risk sources could be derived, is still not free of discrepancies and is falling short of use. In this paper, a new linguistic clustering algorithm, using the k-means++ procedure in addition to the semantics tools of stop world removal and word stemming is developed and codified. Then, the algorithm is applied on a vast risk notions set, emanated from an exhaustive review of the relative literature. The clustered and semantically processed results of the application are then used for the deduction of risk sources. Thus, this paper provides a compact, general, and encompassing master set of risk sources, discretized among distinct overhead categories.