滑坡灾害危险性评价模型比较

通过对庆元地区滑坡灾害的详细研究，比较讨论了单变量评价、决策树评价与逻辑回归法等3种滑坡危险性评价模型在实际应用中的效果。单变量模型最为简单，计算方便，得到的结果也具有较高的准确性，当收集到的滑坡因素越多，对滑坡发育的刻画越精确，则最终评价结果也越可信；决策树算法能迅速从大量样本中归纳出一般规则，并具有较高的精度，但其结果也受到训练样本数量大小和样本维数即参与评价的因素集数量的影响；逻辑回归模型具有计算方法简单、评价结果物理意义明确等突出优点，但需要有大量样本数据的支持才能取得较好的效果。

Comparison of models for hazard assessment of landslide

Through the agency of research on the landslide disaster particularly in Qingyuan Administrative Region, the authors comparatively discussed the effect of different susceptibility assessment model such as single-variate superimposing model, decision-making tree model and logistic regression model for landslide hazard in application. The single-variate superimposing model which is simple and easy to use, acquires the most accurate result. The more factors obtained, the more accurate and credible the result is. The decision-making tree model can sum up rapidly the general rule from the massive samples with accurate result, but the result is influenced by the size and the dimension of training sample. The strong points of logistic regression model are uncomplicated calculation and obvious physical meaning for the assessment.