基于贡献度随机森林模型的公司债信用风险实证分析

运用贡献度随机森林方法（CRF）方法探讨公司债财务指标比率与其违约率的关系.运用连续属性离散化方法（OB）进行财务指标最优降维；运用WOE变换进行模型变量约简.研究表明，CRF模型的分类性能显著优于其他模型，测试集评估总体正确率达90.47%，AUC统计量、AR比率及K-S值分别提升了2.6%、7.6%、4.38%，变量贡献度量化了各财务指标对违约率影响，为诠释随机森林预测机制提供了依据.     

New Algorithms for Data Discretization Based on Rough Set Theory

The discretization of real values is always one of the key problems to be solved in the domain of machine learning for its great contribution to speeding up the followed learning algorithms, cutting down the real demand of algorithms on running space and time, and improving the clustering capability of the ultimate learning results. The basic characteristics and framework of discretization approaches based on rough set model are analyzed at first, then the different measurements of the importance of candidate cuts are discussed and researched. Two new heuristic algorithms are put forward to finally select the useful cuts from a candidate set. The selected cuts of the two algorithms will adequately maintain the discernible relation of information systems for their full considering the specialty of rough set, which perfectly embodies the advantages of this theory. Moreover, excellent discretization results may be expected through these heuristic algorithms.     

SWAT模型子流域划分引起的土地利用变化对径流和输沙模拟结果的影响

为揭示水文模型计算单元划分引起的土地利用变化对模拟结果的影响,采用SWAT模型,通过设置11种子流域阈值情景(4 000,3 200,2 000,1 500,1 100,800,400,200,100,50,25 hm²),并保持地形和土壤数据不变,定量分析了子流域划分引起的土地利用空间离散化对径流输沙模拟结果的影响。结果表明:(1)随着子流域数量的增加,流域内土地利用中面积最大的林地呈先增加后减少趋势,所有情景下林地面积均高于实际林地面积; 而随着土地利用离散程度的增加耕地面积呈减少趋势,水域、草地和建设用地面积则呈增加趋势并逐渐接近实际面积;(2)土地利用空间离散化导致年均径流量(1.85%)和输沙量(65.84%)减少,且对输沙量影响更显著;(3)土地利用空间离散化造成洪水过程和输沙峰值显著降低,并导致年最大1 d、连续最大5 d和连续最大7 d输沙量显著下降(p<0.01)。综上所述,SWAT模型不同子流域划分显著改变了流域土地利用分布,进而引起径流输沙有关参数变化,最终导致径流和输沙模拟结果的变化。     

果品产量统计数据空间离散化研究——以北京平谷区为例

果品产量是衡量区域农产品生产功能的重要指标。果品产量的空间离散化,有助于揭示区域内部要素的空间差异性和获取微观尺度的果品产量数据。该研究依据地形起伏度和土地利用分区将平谷区划分为平原区、浅山区和深山区,并分别建立离散化模型进行修正,将村果品产量离散到500 m×500 m格网下的果园图斑上。结果表明:(1)村果品模拟产量与果品实际产量的相关性系数由总体样本建模的0.585提高到分区建模的0.690,表明分区建模是一种有效的方法。(2)各区影响因素与产量的关系存在差异。平原区果园面积对果品产量呈现线性影响;浅山区和深山区果园面积对果品产量的影响较为剧烈,呈幂指数关系,并且增加海拔高度和坡度2个变量后模型的拟合效果均有改善。(3)在自然环境、社会经济等综合影响下,平谷区果品产量呈现地带性分布。其中,产量较高的区域集中分布于地形相对平缓、自然本底较好、农民劳作便利的浅山带;而平原区和深山区分别受政策限制和地形限制,果品产量相对较低。     

摆动从动杆盘形凸轮机构最小尺寸的完全解法

摆动从动杆盘形凸轮机构的最小尺寸综合问题,迄今未获得圆满解决.基于类速度图原理,通过引入“辅助平行直线族”,求解其与限制凸轮轴心位置若干直线族的交点族中的最左端、最右端者,理论上严格保证其集合为限制凸轮轴心位置的边界点、边界线,进而给出限制凸轮轴心位置的边界点、边界线的精确解析解和离散数值解法,据此直接求解凸轮的最小基圆半径、摆杆长度和直观显示凸轮轴心的容许选取区域,从而解决了按许用压力角设计最小尺寸的摆动从动杆盘形凸轮机构的问题.通过机构设计实例,验证了所提方法的正确性和有效性.     

土壤湿度估计的理想同化试验研究

通过使用集合Kalman滤波,建立基于不同离散化方法的同化土壤湿度的理想试验。通过设置不同的初始状态,以集合预报的形式建立了openloop和模拟的理想＂真值＂,并将来自不同离散化方法得到的理想＂真值＂集合生成多模型集合预报的理想＂真值＂。分别使用单个的离散化方法同化来自各自理想＂真值＂和多模型集合预报的理想＂真值＂,来检验通过同化表层土壤湿度后对整个湿度廓线的改进能力。结果表明,由于不同的离散化方法建立的土壤层之间的相关性不同,不同的离散化方法通过同化表层土壤湿度后土壤湿度廓线均有改进,但是改进的效率不同;当使用多模型的集合预报时,能够更有效地吸收观测,比任何单个模型的同化效果都要好。     

Improved Information Value Model and Its Application in the Spatial Prediction of Landslides

Information Value Model is used frequently in domestic landslide hazard zonation prediction, which plays a very important role for land-use planning especially in mountain area. Through theoretical analysis, the prerequisite of Information Value Model which should be used with influence-factors independent from each other is studied. Factor Analysis is adopted to evaluate and eliminate the influence of correlation between influence-factors in spatial prediction of landslides which are showed through case study. Relationships between landslides and influence-factors are studied based on a certain watershed in which the Information Value Model and Factor Analysis are applied. The probability of landslides changes regularly with the influence-factors, and based on which a method for the discretization of influence-factors is advised. The distribution of landslide susceptibility of above watershed is calculated and validated based on the area-ratio index with landslides occurring in nearly 5 years, which proves the credibility of Information Value Model.     

流动模拟二阶精度的简易保持研究

首先分析了各类低阶和高阶差分格式，指出二阶精度格式具有综合优势．接着以二维对流扩散方程为例，推导出延迟修正的CDS格式，并定义为DCDS格式．通过标量绕静点输运、自然对流及混合对流等实例，并与UDS，CDS及QUICK格式比较，论证了DCDS格式不仅能维持计算结果的二阶精度，而且极大地约束了CDS格式所固有的振荡和越界行为，边界处理也很简单．总之，在程序的效率、调试、边界处理、计算精度及稳定性等因素权衡下．DCDS格式是保持二阶数值精度的简易途径。     

Correlation Between Granular Entropy and Classification Accuracy in Discretization

This paper discusses the correlation between the number of cut points, granular entropy and classification accuracy in discretization. It is proven that granular entropy decreases if the number of cut points increases. A hybrid discretization algorithm is proposed to provide discretization schemes for studying these measures. The simulation experiments show that the absolute value of the correlation coefficient between number of cut points and classification accuracy is quite large, as it for granular entropy and classification accuracy. Sometimes, the correlation between the granular entropy and classification accuracy is smaller than that between the cut points and classification accuracy.     

遥感、GIS辅助下流域空间离散化方法研究

以江西兴国潋水河流域为研究区 ,设计了流域—子流域—水文响应单元的空间离散方案和实现步骤。在遥感、GIS技术支持下 ,应用了基于栅格的数字高程模型 (DEM)的流域水文建模方法 ,进行了流域河网的自动生成 ,子流域的自动划分以及流域边界生成研究 ,根据流域地形特征将整个流域分割成多个子流域。在分割出的每个子流域内部 ,通过数字土壤图和土地利用图的叠加统计分析生成单一土地利用类型和土壤类型组合的水文响应单元 ,将研究区离散成为 6 2个子流域 ,399个水文响应单元 ,实现了分布式流域建模的空间离散化 ,在这个过程中 ,研究解决了平坦区河网精确生成和流域边界的误差改正等技术难题。