分类数据

Then by applying the method of ant swarm intelligence, classify customers on the basis of the presupposed group resemblance coefficient and the drop-pickup probability.

目前针对具体的客户分类有很多数据挖掘算法，这些方法主要集中在聚类和分类两大类，其中又以聚类分析方法更为突出[1~3]。

In this research, the SOLO (Structure of the Observed Learning Outcome) Taxonomy presented by Australian scholars Biggs and Collis (1982) was adopted, and students studying the fifth and sixth grades in Taipei area were chosen as research targets.This study was primarily implemented with a quantitative analysis and supported by a qualitative analysis. In order to carry out the research, a questionnaire had been designed with criteria suggested by Jones et al.(2000), including "describing data","organizing and reducing data","representing data" and "analyzing and interpreting data." This questionnaire consisted of open-ended questions, and after analyzed with qualitative methods, the responses were classified and scored based on standards set in SOLO Taxonomy. Then the quantitative method "Partial Credit Analysis" from Rasch model was adopted in examining the reliability, validity and consistency of outcomes from the qualitative analysis.

提出的统计思维架构－「描述资料」、「组织及简化资料」、「呈现资料」、「分析资料」作为评量试题的向度，题型为开放性问答题，在区分学童的答题表现方面，采用质的分析，参照SOLO（Structure of the Observed Learning Outcome）分类法的分类标准，将学童不同的答题反应进行分类，给予不同的分数，最后再采用Rasch model中的部份给分模式进行量化分析，以量化数据佐证质性分析的信度与效度，以及不同给分的吻合度。

In this dissertation, based on the extraction of meaningful polarization information, the polarimetric speckle reduction, polarimetric contrast enhancement, polarimetric classification and ship detection in polarimetric SAR images are systemically studied. During the study, some new concepts and methods are introduced, which are supported by the application to measured Pol-SAR data. The innovations of this dissertation are as follows: 1 A simplified schem is proposed to reduce the complexity of scattering-model-based speckle filtering. 2 Extend the expression of polarimetric correlation coefficient to help distinguishing targets. In addition, introduce the quantized parameters of polarimetric signature so as to show the differentiae in quantity in co-polar power, cross-polar power and shapes of polarization signatures among targets. 3 Define a new parameter of difference degree between targets, which can be used for express the difference between both coherent and distributed scatters.

论文在较系统的研究目标特征量提取方法和特征量所包含的物理意义的基础上，对极化SAR图像降斑、极化对比增强、极化SAR图像分类以及极化SAR图像中海上舰船目标检测等方面进行了研究，并提出了一些新的概念和方法，对实测数据的处理结果表明，这些新的概念和方法对于极化SAR信息处理是有效的，本文的主要创新成果包括以下方面： 1）针对基于预分类的MMSE极化降斑方法实现过程非常复杂的问题，提出了简化方案，并验证了方案的有效性； 2）扩展了极化相关系数的表示方法，这种扩展增强了极化相关系数对目标的区分效果；另外还提出了极化特征图量化参数，可以从量上表征不同目标间相同极化和交叉极化的回波功率差异和极化特征图形状差异； 3）定义了目标间的差异度参数，它不仅可以用于描述相干目标，还可以用于描述分布目标的差别程度。

In this paper, we present a semi-supervised algorithm to learn Gaussian process classifiers, which is combined with nonparametric semi-supervised kernels in the presence of unlabeled data.

摘要提出了一种半监督算法用于学习高斯过程分类器，其通过结合非参数的半监督核向分类器提供未标记数据信息。

Data is grouped under general categories for each country. Standardized classifications make our databases more user-friendly.

每个国家的数据都是按照通用分类标准划分，标准分类让数据库的用户使用界面更易操作

A widely used strategy, which computes the lower and upper of all decision classes, was introduced in the algorithm. Thus, the task of rule induction from inconsistent data is reduced to rule induction from consistent data.

实验表明它能获取较少的规则，规则较为简单，且规则是自然按照强度由大到小排列，便于分类决策；另外，从对Monks数据集的实验可以看出所获得的规则有较好的预测分类能力。

D based on the charac teristics of lacunarity data ,we also give two effective methods to classify. we develop the method of fractal dimension image tcxture classification.

针对空隙数据自身的特点，提出了分别利用拟合回归方法和模糊聚类分析方法进行分类的思想，从而进一步完善了影像纹理分类中的分形维方法。

The experiments based on evidence-based medicine web documents show that this method can improve classification accuracy more effectively than single SVM classifier.

基于循证医学网络文献分类的实验表明，该方法使训练数据规模大大减小，且比传统的SVM算法分类性能更好，具有较好的精确率和召回率。

The improved SVM was used to classify the dataset of WWTP, and the results have indicated that the misclassification rate of the improved SVM is 16.5% lower.

结果表明：用改进的SVM对污水处理厂的故障数据进行分类时，比未经改进的SVM错分类率低16.5%。

On the base of applied tech. and system structure, the key tech. are expressed in detail. It includes:Knowledge representation technology: it is described in two stages by "Depictive frame + Rule group"."Depictive frame" is used to reflect the factors concerning question field."Rule group" reflects the son-question in question field; knowledge management system: it is made up of knowledge-base editing component, factor classifying component, knowledge inspecting component and extractive knowledge component; uncertainty reasoning adopt method of elicitation search strategy and data drive controlled by member knowledge. It take reliability of compound proof at first. That can resolve the matching contradiction. Reasoning explanation introduce pre-text and followed method of resolving path. Knowledge discovery component take SLIQ as basic arithmetic. It provides classification model in two stage forms which aim at training suit. That is decision tree and productive rule.

在确定应用技术与系统结构的基础上，重点阐述了本文的核心技术，这包括：知识表示技术，采用"描述框架+规则组"的层次结构，"描述框架"用于描述问题域所涉及的因素，"规则组"反映问题域内的子问题；知识管理系统，由知识库编辑构件、因素分类构件、知识检测构件和知识求精构件组成；不确定性推理机采用启发式搜索策略，元知识控制的数据驱动，以复合证据可信度优先作为匹配冲突消解的策略，推理解释采用预置文本与求解路径追踪的方法；知识发现，采用SLIQ快速可伸缩算法，分类模型用判定树与产生式规则两种形式描述。

The split between the two groups can hardly be papered over.

这两个团体间的分歧难以掩饰。

This approach not only encourages a greater number of responses, but minimizes the likelihood of stale groupthink.

这种做法不仅鼓励了更多的反应，而且减少跟风的可能性。