基于信息检索方法的统计翻译系统训练数据选择与优化
Corpus Selection and Optimization for Statistical Machine Translation System Based on Information Retrieval Method
黄瑾[1,2,3] 吕雅娟[1,3] 刘群[1,3]
双语平行语料库是构造高质量统计机器翻译系统的重要基础。与传统的通过扩大双语平行语料库规模来提高翻译质量的策略不同，本文旨在尽可能地挖掘现有资源的潜力来提高统计机器翻译的性能。文中提出了一种基于信息检索模型的统计机器翻译训练数据选择与优化方法，通过选择现有训练数据资源中与待翻译文本相似的句子组成训练子集，可在不增加计算资源的情况下获得与使用全部数据相当甚至更优的机器翻译结果。通过将选择出的数据子集加入原始训练数据中优化训练数据的分布可进一步提高机器翻译的质量。实验证明，该方法对于有效利用现有数据资源提高统计机器翻译性能有很好的效果。[著者文摘]
Parallel corpora are an indispensable resource for translation model training in statistical machine translation （SMT） system. Instead of collecting more and more parallel training corpora, this paper aims to improve the performance of SMT system by exploiting full potential of the existing parallel corpora. We propose an approach to select and optimize training corpus by using information retrieval method. First, sentences similar to the test text are selected to form a small and adapted training data. This allows us to get a comparable or even better performance with only a subset of the total data and the less hardware need, Second, we add the selected subset to the entire corpus to optimize the data distribution and get a better result, The experiments show that this method can effectively improve the performance of SMT system .[著者文摘]
单因子利率期限结构模型参数估计的数据选择
潘冠中
本文证明了单因子利率模型参数估计数据选择的相关性原则，并提出另一原则：选择交易最频繁、成交量最大的利率品种。依据这两个原则，R007是瞬时利率rt的最佳近似替代，在使用中国货币市场利率估计单因子利率模型的参数时，应该选择R007作为估计数据。
从曲线性数据的直线化过程中选择预测模型的方法
王文莲
看似简单的反映自变量与因变量关系的线性函数，如果选择、应用得当，可以明显地反映一种因素的变动对另一种因素的影响，在经营、管理等活动的决策分析中有着广泛的应用。因此科学、合理地选择线性函数的模型并恰当地应用，成为预测理论与实践上的重要课题。[第一段]
基于非集计选择模型的长春市居民出行数据分析
Disaggregated Model-based Changchun Residents＇ Travel Times Data Analysis
富晓艳[1] 隽志才[2] 宗芳[1]
传统的四步骤模型中的出行生成模型是预测研究范围内每个交通小区的出行发生次数和出行吸引次数，直接将居民的出行次数简单地按交通小区进行集计处理，并没有充分考虑每个人的出行次数，难以反映个人和家庭的社会经济等因素对居民个人出行次数的影响．根据2003年吉林省长春市居民出行调查数据，利用非集计模型建立居民个人的出行次数选择模型，并应用相关统计软件对模型进行标定，进而分析居民个人的出行次数，从而求得居民个人出行次数的期望值，初步尝试探索非集计模型和四步骤模型的综合应用问题．[著者文摘]
The Travel Generation Model of the Four Process Model System is one typical aggregated model, which simply forecasts the travel generation times and travel attraction times of every travel zone in the research district. It does not consider well the travel times of every resident, and it just does the sum of the travel times of the zone residents It can not reflect the effect of the individual and social factors and other factors on the residents＇ travel times. According to the 2003 Changchun data, the paper utilizes the disaggregated theory to formulate disaggregated travel times forecasting model to analyze the residents＇ travel times, then it has made the sensitivity analysis of the residents＇ travel times. At last it has calculated the individual travel time expectation. It is a beneficial research on the integrative study and application of the disaggregated model and the Four Process Model System.[著者文摘]
Web数据集成系统基于QC模型的物化视图选择
高军 唐世渭 杨冬青 王腾蛟
在Web数据集成系统中，物化视图能够有效地减少网络传输代价，提高系统的查询效率．如何选择查询进行物化，使得选中的查询满足集成层的空间限制，同时获取最大物化收益，成为集成系统中一个迫切需要解决的问题．传统方法没有考虑到海量XML查询之间的包含关系，其选择的物化视图中可能包含冗余的信息．针对上述问题，提出了①Web数据集成系统中海量查询集合的QC(query containment)模型，该模型能够捕捉查询之间最常见的包含关系；②基于QC模型的物化视图选择算法，算法考虑了物化视图选择相关的主要因素，包括查询提交的频率、空间代价、查询重写能力和查询结果的完备性，提出了查询位图的物化视图组织方式，从而获取更加合理的物化视图选择方案．实验结果证明了该方法的有效性．
论电子商务中选择数据挖掘技术的依据
On the Basis of the Selection of DataMining Techniques in Electronic Commerce
祝娟
数据挖掘中应用的技术包括经典的统计、近邻、聚类，也包括最新发展起来的决策树、神经网络和关联规则等一些较新的方法。结合数据挖掘技术在电子商务中的应用，提出选择数据挖掘技术的两个重要依据．以便开发出有效、实用的数据挖掘系统。[著者文摘]
The techniques applied in data mining include some new methods such as the latest decision tree, neural network and association rules as well as classic statistics, nearest neighbour and clustering. This paper proposes two important foundations of the selection of data mining techniques, combining the application of it in electronic commerce, so as to develop effective and practical data mining system.[著者文摘]
基于神经网络的石油价格预测
Oil Price Forecasting Using Neural Networks Method
孙大利 来君
当前石油价格研究在石油价格数据选择、数据预处理和预测方法选择上存在数据时段选择不当、直接套用原始数据代入模型以及价格预测模型和训练数据类型不相匹配等问题，需要予以解决。在采用同期通货膨胀率指数调整、滑动平均周期项以及随机项滤波等方法对石油价格时间序列数据进行预处理的基础上，利用神经网络方法，以纽约商品交易所（NYMEX）为例对轻质原油期货即期价格时间序列数据建立预测模型。最后用油价波动趋势进行拟合分析，将预处理后的石油价格时间序列数据输入到神经网络预测模型，模型的预测结果和直接套用原始数据得到的预测结果相比较，其平均偏差率显著降低。[著者文摘]
At present, there are some mistakes in choice, pretreatment and forecasting of time series data of oil price in some researches, such as improper data set, using raw data indiscriminately and improper model corresponding to data set, which need to be improved. Using inflation rate adjustment, moving- average method and stochastic filter, oil future front price data series was pretreated. Based on this a forecast model of time series oil price about NYMEX was presented with BP neural networks. In the end, the oil price fluctuation trend was analyzed. The processed oil price time series datas were inputed into neural networks forcasting model. The forecasting results between raw data and proce.ssed data using neural networks was compared and the latter effect is better, and the average deviation rate remarkably reduced[著者文摘]
数据仓库中用于视图选择的增强遗传算法
Modified Genetic Algorithm for View Selection in Data Warehouse
王自强 孙霞 张德贤
物化视图选择问题是数据仓库设计中最重要的问题之一，为了高效地解决这一问题．提出了一个如何选择物化视图集的增强遗传算法，以便在存储空间约束的条件下，取得较好的查询性能和较低的视图维护代价．这一算法的核心思想在于，首先，运用一个基于单位空间最大收益值的预处理算法来生成初始解，然后，该初始解经采用了多种优化策略的遗传算法进行提高，这些优化策略包括：基于改进的锦标赛和精英选择相结合的选择算子、基于半均匀交叉算子及自适应变异算子．并且，在进化过程中产生的无效解用损失函数加以修补．试验结果表明，该算法在寻优性能上优于启发式算法和经典遗传算法．[著者文摘]
Materialized view selection problem is one of the most important decisions in designing a data warehouse. In order to efficiently solve the problem, a modified genetic algorithm for how to select a set of views to be materialized so as to achieve both good query performance and low view maintenance cost under a storage space constraint is proposed. The core idea of the algorithm is as follows. First, a pre-process algorithm based on the maximum benefit per unit space is used to generate initial solutions. Then, the initial solutions are improved by genetic algorithm having the mixture of optimal strategies, such as selection operator based on combination of elitism and modified k-tournament method, crossover operator based on Hamming Distance and self-adaptive mutation operator. Furthermore, the generated infeasible solutions during the evolution process are repaired by loss function, The experimental results show that the proposed algorithm outperforms heuristic algorithm and canonical genetic algorithm.[著者文摘]
高光谱遥感数据最佳波段选择方法试验研究
刘建平 赵英时 等
分析了多光谱遥感数据最佳波段选择的联合熵，行列式值及最佳指数等信息量计算方法的内在联系，说明了信息量方法用于高光谱遥感数据最佳波段选择的局限性，提出了基于类间可分性的最佳波段选择原则和方法，通过试验，说明了各种处理方法的有效性，局限性和计算复杂性。
试论空间数据库数据模型的选择
杨春成 张清浦 等
本文通过对纯关系数据模型，面向对象数据模型和对象关系数据模型优缺点的分析，认为利用基于对象关系数据模型来开发空间数据库是目前较理想的选择，该模型应用面向对象思想来扩展关系模型，在原来只支持简单数据类型的基础上进行扩充，使之能支持用户定义的数据类型，支持对象的属性与方法等，从而既继承了关系数据模型的优点，又能解决地理信息系统等应用中要求支持复杂数据类型的问题，不失为空间数据库建立过程中一种理想的数据模型。