公开(公告)号：JP2010044649A

公开(公告)日：2010-02-25

申请号：JP2008209066

申请日：2008-08-14

Applicant: Internatl Business Mach Corp ,  インターナショナル・ビジネス・マシーンズ・コーポレーションInternational Business Maschines Corporation

Inventor： KUBO HARUNOBU ,  YAMAMOTO HIROMI

IPC: G06F17/30 ,  G06N5/04

Abstract: PROBLEM TO BE SOLVED: To provide a technique for comparing decision trees in detail without depending on a difference of tree structures thereof.
SOLUTION: A data set storage section stores a plurality of data sets, which are sets of a plurality of instances respectively having the same kind of target attribute. A decision tree information storage section stores a plurality of decision trees respectively generated from different data sets. A target attribute determination section determines a value of a target attribute having many instances to be classified in the process of generating a decision tree for a node as a label of the node, for each node of the decision tree. A basic frequency calculation section calculates a frequency at which an instance having the same target attribute as a label of a node is classified in the process of generating a decision tree, for each node. An application frequency calculation section makes a decision tree classify an instance which has caused another decision tree to be generated, and calculates a frequency at which the instance having the same target attribute as a label of the node is classified, for each node of the decision tree. An output section outputs a result of comparing two frequencies as a comparison result of the decision trees.
COPYRIGHT: (C)2010,JPO&INPIT

Abstract translation: 要解决的问题：提供一种用于比较决策树的技术，而不依赖于其树结构的差异。 解决方案：数据集存储部分存储分别具有相同种类的目标属性的多个实例的集合的多个数据集。 决策树信息存储部分存储分别从不同数据集生成的多个决策树。 目标属性确定部确定在决策树的每个节点的生成用于节点的决策树作为节点的标签的处理中具有许多实例的目标属性的值。 基本频率计算部分计算在每个节点生成决策树的过程中分类具有与节点的标签相同的目标属性的实例的频率。 应用频率计算部分使决策树对已经产生另一个决策树的实例进行分类，并且计算与该节点的每个节点相同的具有与节点的标签相同的目标属性的实例被分类的频率 树。 输出部分输出比较两个频率的结果作为决策树的比较结果。 版权所有（C）2010，JPO＆INPIT

公开(公告)号：JP2001318938A

公开(公告)日：2001-11-16

申请号：JP2000135928

申请日：2000-05-09

Applicant: IBM

Inventor： MORIMOTO YASUHIKO ,  KUBO HARUNOBU

IPC: G09B29/00 ,  G06F17/30 ,  G06F19/00 ,  G06T7/60

Abstract: PROBLEM TO BE SOLVED: To provide a space data mining method for finding out a distance itself and an azimuth itself for optimizing a certain purpose to be requested by many analytical operations without previously determining a distance and an azimuth and deriving a space correlation rule. SOLUTION: The space data mining device for calculating an optimum distance from a data base including space information such as an address is provided with an input means for inputting an object function necessary for distance optimization, an intermediate table preparation part 30 for generating an intermediate table by calculating a distance between a start point and a question point on the basis of start point set data and question point set data stored in a data base, and an optimum distance calculation part 39 for calculating a distance for optimizing the value of the object function inputted by the input means on the basis of the intermediate table generated by the preparation part 30.

公开(公告)号：JP2001352980A

公开(公告)日：2001-12-25

申请号：JP2000171030

申请日：2000-06-07

Applicant: IBM

Inventor： KAJIMA HISATSUGU ,  KUBO HARUNOBU ,  SUGIHARA AKIRA

IPC: C12N15/09 ,  C12N5/10 ,  C12N15/10 ,  C12Q1/68 ,  G06F17/30 ,  G06Q50/22 ,  G06F17/60

Abstract: PROBLEM TO BE SOLVED: To make it possible to identify the source of the genetic information in a DNA having the specified information by embedding the information into the base sequence of a DNA. SOLUTION: This method comprises the steps of associating a base sequence having a pattern usually not appeared in DNA to the identification information for identifying the source of the specific genetic information which the DNA has, and embedding the base sequence associated to the identification information into the DNA so as not to affect the genetic information of the DNA.

公开(公告)号：JP2009205615A

公开(公告)日：2009-09-10

申请号：JP2008049729

申请日：2008-02-29

Applicant: Internatl Business Mach Corp ,  インターナショナル・ビジネス・マシーンズ・コーポレーションInternational Business Maschines Corporation

Inventor： HITO SHOHEI ,  IDE TAKESHI ,  KAJIMA HISATSUGU ,  KUBO HARUNOBU ,  MATSUZAWA YASUSHI

IPC: G06F17/30 ,  G06N5/04

CPC classification number: G06K9/623 ,  G06N99/005

Abstract: PROBLEM TO BE SOLVED: To provide a method for efficiently solving the change analysis problem.
SOLUTION: Different virtual labels, for example, like +1 and -1, are assigned to two data sets. A change analysis problem for the two data sets is reduced to a supervised learning problem by using the virtual labels. Specifically, a classifier such as logical regression, decision tree and SVM is prepared and is trained by use of a data set obtained by merging the two data sets assigned the virtual labels. A feature selection function of the resultant classifier is used to rank and output both every attribute contributing to classification and its contribution rate.
COPYRIGHT: (C)2009,JPO&INPIT

Abstract translation: 要解决的问题：提供一种有效解决变化分析问题的方法。 解决方案：将不同的虚拟标签（例如+1和-1）分配给两个数据集。 通过使用虚拟标签将两个数据集的变化分析问题简化为监督学习问题。 具体地说，准备了诸如逻辑回归，决策树和SVM的分类器，并且通过使用通过合并分配了虚拟标签的两个数据集获得的数据集进行训练。 使用得到的分类器的特征选择功能对贡献于分类的每个属性及其贡献率进行排序和输出。 版权所有（C）2009，JPO＆INPIT