| Summary: | 碩士 === 國立臺北科技大學 === 土木與防災研究所 === 103 === The analytic hierarchy process (AHP) has proved to be a powerful multiple-criteria decision analysis to deal with complex problems, and its multi-disciplinary applications have been widely published. However, traditional AHP compels decision-makers to converge vague judgments to single numeric preferences in order to evaluate the pairwise comparisons of all pair of elements and reach the further decision alternatives. Therefore, this study, aiming to solve the abovementioned problem, proposes an innovative stochastic analytic hierarchy process (SAHP), which incorporates probability distributions and uncertainty in judgment. The constrained Monte Carlo simulations (MCS) based on Latin hypercube sampling (LHS) scheme was employed in this SAHP to establish the pairwise comparison matrix and eventually to evaluate the overall weight of each element. Moreover, the modified SAHP was applied to the combined index models for inundation and slop disaster risk assessments of disaster-prone villages in both Taipei and New Taipei cities. As a result, the differences of relative weights between indexes were decreased so that the proposed SAHP approach turned out to improve the judgment uncertainty from insufficient information. In addition, few extreme opinions in comparison were found to reflect group outlier. Finally, contrary to AHP approach, the results of risk score show that the disaster risk of 61% and 46% of villages has increased for inundation and slop disaster, respectively. Especially, Fuji Vil., Xinzhuang Dist. Gained most 111 point for inundation risk assessment. This study concludes that incorporation of MCS with traditional AHP lessens the uncertainty in the pairwise comparisons and assesses the relative significance of elements in a more objective way. Furthermore, the LHS scheme improves the sampling uniformity and significantly reduces sampling size so as to save a great amount of computing time. More importantly, the proposed SAHP enhances its applicability in various practices.
|
|---|
