Production and preprocessing system for data mining

• US 20030018646A1
• Filed: 02/19/2002
• Published: 01/23/2003
• Est. Priority Date: 07/18/2001
• Status: Active Grant

• Current Assignee: Hitachi, Ltd.
• Original Assignee: Hitachi, Ltd.
• Inventors: Nishikawa, Tetsuo, Ohta, Yoshihiro, Ihara, Sigeo
• Granted Patent: US 6,868,423 B2
• Time in Patent Office: Days
• US Class Current: 707/100
• CPC Class Codes: G06F 16/80 of semi-structured data, e.g. markup language structured data such as SGML, XML or HTML content-based retrieval of web data G06F16/95
• CPC Class Codes: Y10S 707/99943 Generating database or data structure, e.g. via user interface
• CPC Class Codes: Y10S 707/99944 Object-oriented database structure

Abstract

Disclosed is means capable of solving trouble in managing data formats and procedures and capable of carrying out advanced preprocessing more intuitively. A data aggregate to be inputted to a mining engine is converted into hierarchical unit trees, and node conditions of the hierarchical unit trees are changed, whereby the data aggregate and a data structure are subjected to dynamic conversion/edition processing. Thus, a system is constructed, in which preprocessing for data mining is unitarily managed/semi-automated.

Claims

1. A method of preprocessing for data mining, comprising the steps of:

creating, from XML data, a hierarchical unit tree as a tree structure in which attributes of the XML data are set as a leaf node and a non-leaf node, a relationship between the attributes without including an attribute value is expressed, and a redundant parent-child relationship between the nodes is optimized by merging;

adding a change to the hierarchical unit tree; and

converting the XML data so as to reflect the change added to the hierarchical unit tree.

2. A method of preprocessing for data mining, comprising the steps of:

displaying, on a screen, a hierarchical unit tree as a tree structure in which a leaf node and a non-leaf node, and a branch expressing a parent-child relationship between the nodes are included, both of the nodes corresponding to attributes of XML data, and a redundant parent-child relationship between the nodes is optimized by merging, the hierarchical unit tree being created from the XML data;

adding a change to the hierarchical unit tree; and

converting the XML data so as to reflect the change added to the hierarchical unit tree.

3. The method of preprocessing for data mining according to claim 2, wherein an operation for adding a change to the hierarchical unit tree includes an operation for setting a plurality of nodes as child nodes of a node newly created on the same hierarchy as the plurality of nodes having the same non-leaf node as a parent.

4. The method of preprocessing for data mining according to claim 2, wherein an operation for adding a change to the hierarchical unit tree includes an operation for moving a designated node to a position of a child of the other node than a current parent of the designated node.

5. The method of preprocessing for data mining according to claim 4, wherein the operation for moving a designated node to a position of a child of the other node than a current parent of the designated node is an operation for dragging the designated node by mouse and dropping the designated node on a node newly to be a parent.

6. The method of preprocessing for data mining according to claim 2, wherein the operation for adding a change to the hierarchical unit tree includes an operation for changing attribute names of a plurality of nodes to the same attribute name, the plurality of nodes having the same non-leaf node as a parent, and merging the plurality of nodes.

7. The method of preprocessing for data mining according to claim 2, wherein an operation history for hierarchical unit trees is displayed, the hierarchical unit trees changed by operations are recorded respectively, and when a specified operation step of the operation history displayed is designated, a hierarchical unit tree corresponding to the operation step is displayed.

8. A preprocessing system for data mining, comprising:

a display unit for displaying a hierarchical unit tree as a tree structure in which a leaf node and a non-leaf node, and a branch expressing a parent-child relationship between the nodes are included, both of the nodes corresponding to attributes of XML data, and a redundant parent-child relationship between the nodes is optimized by merging, the hierarchical unit tree being created from the XML data; and

a filter selection unit for selecting a filter for adding a change to the hierarchical unit tree.

9. The preprocessing system for data mining according to claim 8, further comprising:

a history display unit for displaying a history of filters applied to the hierarchical unit tree.