基于文本挖掘词频反文档频率方法的疾病症状权重挖掘研究

词频反文档频率是文档特征权值表示的常用方法,用以评估单词对于语料库中的其中一份文件的重要程度。通过合理映射病例库与语料库的词汇关系,对词频反文档频率模型进行改进,并将改进后的词频反文档频率模型应用到针灸处方疾病症状权重挖掘中,计算出疾病所对应的各种症状权重。实验统计了106种疾病,其中与临床诊疗经验相符合的有84种,准确率达79.2%,实验表明改进的模型能得到较好的疾病症状权重鉴别效果。     

利用LS_AR模型对极移参数的中长期预报

为了满足深空探测器自主导航定位对极移参数中长期预报的需求,阐述了基于LS_AR模型的极移参数中长期预报和精度评定的原理,提出了4种改进方案对LS_AR模型的构建进行优化,并利用IERS提供的1990～1996年的极移参数的时间序列检验4种优化方案,得到了最优的预报模型,在400 d跨度上,其预报结果的平均绝对误差比未优化的模型小3 mas左右。     

Study on Variation of Fitting Goodness of Relation Between Earthquake Frequency and Magnitude before Moderately Strong Earthquakes (Ms≥5.0)

In this paper,characteristics of spatial and temporal variation of linear fitting goodness before some moderately strong earthquakes(Ms≥5.0)in the eastern part of China(east of longitude 180)are studied according to the famous Gutenberg-Richter's relation expressed as lgN=a-bM,by using the moderate and small events that occurred in and around the source area.The results show that the linear goodness of fitting varies abnormally prior to these moderately strong earthquakes.In the early stage of the earthquake preparatory process,distribution of the energy released through small events in and around the source area is isostatic and the fitting goodness approximates 1,while the distribution of the energy turns to be isostatic before moderately strong earthquakes,leading to the obvious decrease of the linear goodness of fitting.This phenomenon could be a medium term anomaly and a medium term criterion for moderately strong earthquake prediction.     

天文学英语新词自动提取系统

科技名词中文译名标准化和规范化工作是推动我国科技进步和科学知识传播的重要基础。新的科学概念和技术名词层出不穷,如何及时发现并确定新生术语的中文译名是一项普遍的社会需求。介绍了一套全新的天文学英语新词自动提取系统。该系统综合使用脚本过滤、术语识别、正则表达匹配等多种方法,能够自动追踪Ar Xiv论文数据库的更新,分析天文学论文的内容,生成推荐术语列表,从而将学科专家从繁重的科技新词收集整理工作中解放出来,把有限的精力集中到更能体现专业素养的新词审定工作当中。这个系统将为推动天文学等基础学科的新词收集,乃至学科标准化等工作发挥积极作用。     

三套NDVI长时间序列植被指数的对比——以玛曲为例

NDVI／MODIS、NDVI／GIMMS和NDVI／NSMC是时间长度不同、空间分辨率相差甚远的3套ND—VI数据集,如何集成应用这些不同时间长度、不同分辨率的数据进行相关研究,数据集间的比较是最基础的工作。本文以甘肃省甘南州玛曲县为例,用直方图、相关分析、趋势分析等方法研究了这3套NDVI产品数据集的相互关系。结果表明：1）NDVI／NSMC与NDVI/MODIS的直方图具有类似的图像分布特征,但是NDVI／MODIS数据分布范围更大;2）3套NDVI在数值上表现为NDVI／MODIS〉NDVI／GIMMS〉NDVI／NSMC;3）3套数据集空间图像特征一致,两两间均具有十分显著的空间相关性,其中1月份相对最弱,5、10月份最强,三者相比NDVIfNSMC与NDVI／MODIS的空间相关性更强;4）1—3月、5—8月及年均的NDVI／GIMMS与NDVI／NSMC值存在显著的时间相关性,但两者逐年变化趋势存在较大差别,两者气候倾向率相差最大的高达5倍之多。NDVIfNSMC数据集在处理过程中可能未进行大气订正及交叉定标,这是造成共同源的NDVIfGIMMS与NDVI／NSMC差异较大的重要原因。     

太原城市化进程对城市热岛的影响

采用城市化的多项要素指标,系统考察了太原1980-2009年期间城市化的发展对其城市热岛效应的影响关系,发现城市化的发展是导致城市热岛效应形成和加剧的主要原因.认为:①近30年,太原市郊区的增温率为0.51℃/10a,市区的增温率为0.90℃/10a,热岛强度的增温率为0.39℃/10a,热岛效应不断增强;②各类城市发展指数与城市热岛效应都有很好的相关关系;③太原城郊温差和总人口对数呈线性相关关系,其长期变化相关系数为0.81;④太原城郊温差和城市房屋建筑面积对数呈线性关系,其长期变化相关系数为0.85.     

用前期海温预报四川夏季气温的EOF迭代方案

应用ＥＯＦ（经验正交函数）迭代方案，考虑多时次历史资料，在全球海区海温与四川盆地气温非同步联系的基础上，以海温为预报因子进行了夏季气温的长期预报。结果表明：西太平洋高温区等关键海区海温的异常对未来四川盆地夏季气温变化有重要影响，由此建立的引入多时次海温的ＥＯＦ迭代长期温度预报方法，具有较强的预报能力     

A new class of bulk viscous universe with time dependent deceleration parameter and Λ-term

A new class of exact solutions of Einstein’s field equations with a bulk viscous fluid for an LRS Bianchi type-Ia obtained by using a time dependent deceleration parameter and cosmological term Λ. The coefficient of bulk viscosity is assumed to be a power function of mass density (ξ=ξ ₀ ρ ⁿ ). We have obtained a general solution of the field equations from which six models of the universe are derived: exponential, polynomial and sinusoidal form respectively. The behaviour of these models of the universe are also discussed in the frame of reference of recent supernovae Ia observations.      

Long term variations in solar flare activity

A statistical analysis of the contemporary (1954-1975) solar flare particle events has been made for the parametersF (integrated, proton fluence in cm^-2 in an event with kinetic energy above 10 MeV) andR ₀ (the characteristic rigidity). These data are compared with the long-term averaged values determined from stable- and radio-nuclide measurements of lunar samples. The analysis shows that the ancient solar flare proton spectrum was harder (higher R₀ values) compared to that observed in contemporary flares. A similar analysis can not be made for the mean long-term averaged flux (ˉJ, cm^-2 S^-1), since the contemporary averages suffer from an uncertainty due to the statistics of a single event. However, the average flux estimates for time durations 〈T〉 exceeding 10³ yr, are free from such uncertainties. The long-term averaged ˉJ values obtained over different time scales (10⁴ - 10⁶ yr) suggest a possible periodic variation in solar flare activity, with enhanced flux level during the last 10⁵ yr. The available data rule out the occurrence of giant flares, with proton fluence exceeding 10¹⁵ cm^-2 during the last million years.