改进的POTTS模型及其数据多分类直推学习算法

借助于图上离散非局部算子,计算机视觉领域图像分割的Potts模型可直接应用于数据多分类直推学习,但为受多种约束的能量泛函极值问题。采用传统的惩罚函数方法将受约束优化问题转化为无约束优化问题的求解涉及多个难以设定的惩罚参数。通过用较少的标记函数设计每类数据的特征函数自然满足原有的Simplex约束避免了对这类约束的惩罚。通过直接投影方法保证了直推学习中预设标记点精确约束进一步减少了能量泛函中惩罚项及惩罚参数的数量。对平衡分类约束和变量分裂引起的约束通过设计ADMM(Alternating Direction Method of Multipliers)方法降低了对惩罚参数的过分依赖。通过对多个标准数据集进行数值实验验证了所提出模型和算法的有效性。

多源高维数据的多分类纵向整合分析及应用

多分类数据分析在实证研究中具有重要意义。然而,由于高维数、小样本及低信噪比等原因,现有的多分类方法仍面临信息量不足而导致的效果不佳问题。为此,学者们通过收集更多信息源数据以更全面地刻画实际问题。不同于收集相同自变量的不同源样本,目前较为流行的多源数据收集了相同样本的不同源自变量,它们的独立性和相关性为统计建模带来了新的挑战。本文提出基于典型变量回归的多分类纵向整合分析方法,其中利用惩罚技术实现变量选择,并独特地考虑不同源数据间的关联结构,提出高效的ADMM算法进行模型优化。数值模拟结果表明,该方法在变量选择和分类预测上均具有优越性。基于我国上证50的多源股票数据,利用该方法对2019年股票日收益率的影响因素进行了实证探究。研究表明,本文提出的多分类整合分析在筛选出具有解释意义变量的同时具有更好的预测效果。

基于多分类不均衡支持向量机的制造型企业财务管理研究

近年来,制造型企业在经营过程中面临的挑战日益增大。为帮助制造型企业探索科学合理的财务管理方法,更好地保障企业的健康经营和降低成本,论文首先利用哈夫曼树将多分类不均衡企业数据分解为一系列相对平衡的二分类子问题,接着以支持向量机为分类器构建模型,然后,采用算术优化算法进行模型参数寻优。结果表明,发展能力、经营能力、偿债能力对企业的财务状况影响较大。最后,论文基于结果对企业提出相关管理建议。

多分类计数数据的回归建模及其贝叶斯推断

本文使用多项-Logistic-正态模型分析多分类计数响应数据,基于贝叶斯推断方法研究模型的估计及其变量选择方法。通过引入Polya-Gamma分布的潜在变量和回归系数的“Spike-and-Slab”先验,得到了Gibbs后验抽样算法。数值模拟研究和RNA-seq基因实例数据分析验证了MLN模型及其贝叶斯推断方法的有用性。

Land Cover Classification with Multi-source Data Using Evidential Reasoning Approach

Land cover classification is the core of converting satellite imagery to available geographic data.However,spectral signatures do not always provide enough information in classification decisions.Thus,the application of multi-source data becomes necessary.This paper presents an evidential reasoning (ER) approach to incorporate Landsat TM imagery,altitude and slope data.Results show that multi-source data contribute to the classification accuracy achieved by the ER method,whereas play a negative role to that derived by maximum likelihood classifier (MLC).In comparison to the results derived based on TM imagery alone,the overall accuracy rate of the ER method increases by 7.66% and that of the MLC method decreases by 8.35% when all data sources (TM plus altitude and slope) are accessible.The ER method is regarded as a better approach for multi-source image classification.In addition,the method produces not only an accurate classification result,but also the uncertainty which presents the inherent difficulty in classification decisions.The uncertainty associated to the ER classification image is evaluated and proved to be useful for improved classification accuracy.

Housing Correlates of Under-Five Mortality in Urban Ethiopia

Little studies and analysis have been undertaken to investigate the housing determinants of under-five mortality in Ethiopia. This study, therefore, explores the impacts of urban housing variables on the levels and patterns of under-five mortality in the country based on the SPSS （Statistic Package for Social Science） file of the 2005 Ethiopian Demographic and Health Survey （EDHS）. This survey covered a sample of about 4,420 households/housing units of urban Ethiopia. The under-five deaths are computed for women in the age group 15-49 by subtracting the number of children living from children ever born （CEB） and established the proportion dead by dividing deaths by CEB corresponding to the categorical variables of housing structure, facilities, and household durables. The analytical techniques of the study included univariate, bivariate, and multivariate data analysis of the proportional variations of childhood mortality patterns being manifested by ＂bar graphs＂ with respect to housing situations as well as household durables. Amongst the categorical variables of the housing structures, facilities, and household durables with the highest no prevalence of under-five mortality levels are found to be the units of unconventional walls, thatched/leaf/reed roofing, animal dung flooring, shared pit latrine/use of bucket/bush, using kerosene, firewood/straw/charcoal for cooking, unconventional lighting, unprotected water supply, households with no durables.

A multi-resolution global land cover dataset through multisource data aggregation

Recent developments of 30 m global land characterization datasets （e.g., land cover, vegetation continues field） represent the finest spatial resolution inputs for global scale studies. Here, we present results from further improvement to land cover map- ping and impact analysis of spatial resolution on area estimation for different land cover types. We proposed a set of methods to aggregate two existing 30 m resolution circa 2010 global land cover maps, namely FROM-GLC （Finer Resolution Observa- tion and Monitoring-Global Land Cover） and FROM-GLC-seg （Segmentation）, with two coarser resolution global maps on development, i.e., Nighttime Light Impervious Surface Area （NL-ISA） and MODIS urban extent （MODIS-urban）, to produce an improved 30 m global land cover map-FROM-GLC-agg （Aggregation）. It was pos-processed using additional coarse res- olution datasets （i.e., MCD12Q1, GlobCover2009, MOD44W etc.） to reduce land cover type confusion. Around 98.9% pixels remain 30 m resolution after some post-processing to this dataset. Based on this map, majority aggregation and proportion ag- gregation approaches were employed to create a multi-resolution hierarchy （i.e., 250 m, 500 m, 1 km, 5 km, 10 km, 25 km, 50 km, 100 km） of land cover maps to meet requirements for different resolutions from different applications. Through accuracy assessment, we found that the best overall accuracies for the post-processed base map （at 30 m） and the three maps subse- quently aggregated at 250 m, 500 m, 1 km resolutions are 69.50%, 76.65%, 74.65%, and 73.47%, respectively. Our analysis of area-estimation biases for different land cover types at different resolutions suggests that maps at coarser than 5 km resolution contain at least 5% area estimation error for most land cover types. Proportion layers, which contain precise information on land cover percentage, are suggested for use when coarser resolution land cover data are required.