Cause inferring device

• US 5,950,183 A
• Filed: 02/04/1997
• Issued: 09/07/1999
• Est. Priority Date: 08/09/1994
• Status: Expired due to Fees

• Current Assignee: Komatsu Ltd.
• Original Assignee: Komatsu Ltd.
• Inventors: Minobe, Kaori, Yamaguchi, Hiroyoshi
• Primary Examiner(s): Downs, Robert W.
• Application Number:
  US08/793,422
• Time in Patent Office: 945 Days
• Field of Search: 395/183.01, 395/183.02, 706/52, 706/911, 706/916
• US Class Current: 706/52
• CPC Class Codes: G06N 5/025 Extracting rules from data

Abstract

A cause inferring device which infers causes with improved accuracy and efficiency. A phenomenon item (F) other than phenomenon items (A, B and C) for which frequency-of-occurrence data are already inputted is selected based on the degree of association data regarding cause items (2, 3 and 4) having the highest certainty among cause items (1-5) whose certainties are calculated. The frequency-of-occurrence data of the selected phenomenon item (F) and already inputted frequency-of-occurrence data are inputted again and the certainties of the cause items are computed. By repeating such selection and arithmetic processing at least once, causes are narrowed down as a final inference.

Claims

1. A cause inferring device wherein degree of association data indicating degree of association between various phenomenon items and various cause items are prepared as matrix-form data in which one of the phenomenon items and the cause items constitute rows while the other constitute columns, frequency of occurrence data indicating frequency of occurrence of prescribed phenomenon item of the various phenomenon items are input, certainties of the cause items are calculated using the frequency of the input occurrence data and the matrix-form degree of association data, and a cause is inferred from the calculated certainties, wherein the device comprises:

phenomenon item selection means for selecting a phenomenon item, other than the phenomenon items whose degree of association data have been input, in accordance with the degree of association data relating to cause items of highest certainty, of the cause items whose certainties have been calculated;

certainty arithmetic processing means that calculates the occurrence data of the selected phenomenon item and the already-input frequency of occurrence data; and

repetition means for effecting at least a single repetition of the arithmetic processing performed by the certainty arithmetic processing means and selection processing performed by the phenomenon item selection means,

the cause being inferred from the certainties of the cause items calculated as a result of the repetition performed by the repetition means.

2. A cause inferring device according to claim 1 characterized in that, when there is prepared a knowledge base of decision tree structure in which each node represents a phenomenon item of various kinds and each conclusion represents a cause item of various kinds, the knowledge base of decision tree structure is converted to the matrix-form degree of association data and inference of causes is performed by using the matrix-form degree of association data obtained by the conversion process.

3. A cause inferring device according to claim 2, characterized in that there is prepared a plurality of knowledge bases of the decision tree structure, and that the device comprises:

first conversion means whereby prescribed cause items are selected from the various cause items and the plurality of knowledge bases of decision tree structure is converted to matrix-form data in which one of the selected cause items and phenomenon items of various kinds constitute rows while the other constitute columns, and wherein matrix components are degree of association data indicating the degree of association of the cause items and phenomenon items; and

second conversion means whereby the matrix-form degree of association data obtained by the conversion performed by the first conversion means are converted to a knowledge base of decision tree structure wherein the selected prescribed cause items constitute the conclusions and the phenomenon items of various kinds constitute the nodes,

the cause being inferred, after converting the knowledge base of decision tree structure obtained by the conversion performed by the second conversion means into the matrix-form degree of association data, using the matrix-form degree of association data obtained by the conversion process.

4. A cause inferring device wherein there is prepared a knowledge base of decision tree structure wherein various phenomenon items are taken as nodes and various cause items are taken as conclusions, causes being inferred using the knowledge base of decision tree structure, characterized in that the device comprises:

when there is prepared a plurality of knowledge bases of decision tree structure,

first conversion means whereby prescribed cause items are selected from the various cause items and the plurality of knowledge bases of decision tree structure is converted to matrix-form data in which one of the selected cause items and phenomenon items of various kinds constitute rows while the other constitute columns, and wherein matrix components are degree of association data indicating degree of association of the cause items and phenomenon items; and

second conversion means whereby the matrix-form degree of association data obtained by the conversion performed by the first conversion means are converted to a knowledge base of decision tree structure wherein the selected prescribed cause items constitute conclusions and the phenomenon items of various kinds constitute nodes,

the cause being inferred from the selected prescribed cause items using the knowledge base of decision tree structure obtained by the conversion performed by the second conversion means.

5. An inferring device according to claim 4, characterized in that the first conversion means, if branches of the nodes of the knowledge base of decision tree structure are related to conclusions, sets corresponding degree of association data to prescribed values depending on a branch content and, if the node branches are not related to conclusions, does not set a value for the corresponding degree of association data, and that the second conversion means, if the values of the matrix-form degree of association data are prescribed values, sets the branches of the corresponding nodes to a branch content corresponding to the prescribed value and, if the values of the matrix-form degree of association data are not set, effects conversion such that a new branch is provided between corresponding conclusion and node.

6. A cause inferring device wherein degree of association data indicating degree of association between various phenomenon items and various cause items are prepared first as a knowledge base of decision tree structure in which each node represents a phenomenon item of various kinds and each conclusion represents a cause item of various kinds, and the knowledge base of decision tree structure is then converted to the matrix-form degree of association data in which one of the phenomenon items and the cause items constitute rows while the other constitute columns, frequency of occurrence data indicating frequency of occurrence of prescribed phenomenon item of the various phenomenon items are input, certainties of the cause items are calculated using the frequency of the input occurrence data and the matrix-form degree of association data, and a cause is inferred from the calculated certainties, wherein the device comprises:

phenomenon item selection means for selecting a phenomenon item, other data of the phenomenon items whose degree of association data have been input, in accordance with the degree of association data relating to cause items of highest certainty, of the cause items whose certainties have been calculated;

certainty arithmetic processing means that calculates the occurrence data of the selected phenomenon item and the already-input frequency of occurrence data; and

repetition means for effecting at least a single repetition of the arithmetic processing performed by the certainty arithmetic processing means and selection processing performed by the phenomenon item selection means,

the cause being inferred from the certainties of the cause items calculated as a result of the repetition performed by the repetition means.