Relational artificial intelligence system

• US 5,473,732 A
• Filed: 02/07/1995
• Issued: 12/05/1995
• Est. Priority Date: 11/02/1993
• Status: Expired due to Fees

• Current Assignee: Hou-Mei H. Chang
• Original Assignee: Hou-Mei H. Chang
• Inventors: Chang, Hou-Mei H.
• Primary Examiner(s): Moore, David K.
• Assistant Examiner(s): HAFIZ, TARIQ R
• Application Number:
  US08/385,087
• Time in Patent Office: 301 Days
• Field of Search: 395/3, 395/11-12, 395/51, 395/54, 395/60, 395/67, 395/77, 395/900, 395/925
• US Class Current: 706/59
• CPC Class Codes: G06N 5/025 Extracting rules from data
• CPC Class Codes: G06N 5/046 Forward inferencing; Production systems

Abstract

A relational artificial intelligence system is invented and developed. It comprises a relational automatic knowledge acquisition system and a relational reasoning system. The relational automatic knowledge acquisition system is a relational learning system which discovers knowledges from spreadsheet-formed databases and generates relational knowledge bases using inductive learning technique. The relational reasoning system is a relational knowledge-based system which reasons about the generated relational knowledge bases automatically and predicts what will happen under future data readings. The feature of this invention is that every component in this system is relational. Not only the database to be read and the knowledge base to be generated, but also the inductive engine and the inference engine are relational. In the whole reading, operating, and generating processes, data are organized in the spreadsheet-form, and hence the system works with high efficiency and speed.

Claims

1. A relational artificial intelligence system in a digital computer performing automatic knowledge acquisition from a set of data records, generating a set of relational knowledge bases through the aid of a set of CPUs of said computer, and performing inferences on said set of relational knowledge bases to obtain inference results based on a set of required data, comprising:

input/output means for acquiring data and generating output;

computer storing means for storing data and computer programs;

a set of relational inductive engines being a set of executable computer programs stored in said computer storing means for automatically discovering knowledge from said set of data records and generating said set of relational knowledge bases through the aid of said set of CPUs of said computer, each one of said relational knowledge bases comprising a set of knowledge relations; and

a set of relational inference engines being a set of executable computer programs stored in said computer storing means for reasoning about said set of relational knowledge bases and obtaining stud inference results based on said set of required data means for storing all permissible values in fields of each attribute of said decision relations in said computer storing means;

means for assigning a code to each one of said permissible values;

means for translating said permissible values to code;

means for creating a set of code decision relations; and

means for translating said code to said permissible values;

wherein

A. said set of relational inductive engines comprising;

a. means for acquiring data from said set of data records through said input/output means;

b. means for selecting a set of value-attributes and a set of decision-attributes from all attributes of said set of data records;

c. means for creating a set of decision relations in said computer storing means, each one of said decision relations comprising said set of value-attributes, said set of decision-attributes, and more than one record selected from said set of data records;

d. means for clustering tuples in each one of said set of decision relations into positive instances and negative instances;

e. means for counting positive counts of said positive instances and negative counts of said negative instances in each one of said set of decision relations;

f. means for conjunctive generalization comprising means for generalizing values in value-fields in each of said set of decision relations; and

g. means for generating said set of relational knowledge bases in said computer storing means, each of said relational knowledge bases comprising a set of knowledge relations;

B. said set of relational inference engines comprising;

a. means for scanning all value-fields in each one of said set of knowledge relations of said set of relational knowledge bases generated by said set of relational inductive engines in said computer storing means;

b. means for accepting said required data for said value-attributes;

c. means for comparing said required data with values in said value-fields and performing true-false tests for said value-fields;

d. means for tuple testing, for performing true-false tests of status factors of tuples of each one of said knowledge relations;

e. means for determining a set of tuples with true status factors; and

f. means for executing decision statements in said set of tuples through said input/output means.

2. The relational artificial intelligence system of claim 1 wherein said set of data records is a database relation of a database system.

3. A relational learning system in a digital computer for automatically discovering knowledge from a set of data records and generating a set of relational knowledge bases through the aid of a set of CPUs of said computer, comprising:

input/output means for obtaining data and generating output;

computer storing means for storing data and computer programs;

a set of relational inductive engines being a set of executable computer programs for automatically discovering knowledge from said set of data records and generating said set of relational knowledge bases through the aid of said set of CPUs of said computer, each of said relational knowledge bases comprising a set of knowledge relations, comprising;

means for acquiring data from said set of data records through said input/output means;

means for selecting a set of value-attributes and a set of decision-attributes from all attributes of said set of data records;

means for creating a set of decision relations in said computer storing means, each of said decision relations comprising said set of value-attributes, said set of decision-attributes, and more than one record selected from said set of data records;

means for clustering tuples in each one of said set of decision relations into positive instances and negative instances;

means for counting positive counts of said positive instances and negative counts of said negative instances in each one of said decision relations;

means for conjunctive generalization comprising means for generalizing values in value-fields in each one of said set of decision relations; and

means for generating said set of knowledge relations in said set of relational knowledge bases in said computer storing means from said decision relations;

means for storing all permissible values of each attribute of said decision relations in said computer storing means;

means for assigning code to said permissible values;

means for translating said permissible values to said code;

means for creating a set of code decision relations; and

means for translating said code to said permissible values.

4. The relational learning system of claim 3 wherein said set of data records is a database relation of a database system.

5. A relational knowledge-based system in a digital computer, comprising:

input/output means for obtaining data and generating output;

computer storing means for storing data and computer programs, means for storing all permissible values of each attribute of said decision relations in said computer storing means;

means for assigning code to said permissible values;

means for translating said permissible values to said code;

means for creating a set of code decision relations; and

means for translating said code to said permissible values;

a set of relational knowledge bases, each of said relational knowledge bases comprising a set of knowledge relations entered by said input/output means and stored in said computer storing means, each of said knowledge relations comprising a set of attributes and a set of tuples, said set of attributes comprising a subset of value-attributes and a subset of decision-attributes, intersections of said value-attributes and said tuples being called value-fields, each of said value-fields storing a value and having a status factor, and intersections of said decision-attributes and said tuples being called decision-fields, values stored in said decision-fields being decision statements;

a set of relational inference engines being a set of executable computer programs stored in said computer storing means for reasoning about said set of knowledge relations and generating a set of inference results, comprising;

means for scanning said value-fields in each one of said set of knowledge relations of said set of relational knowledge bases in said computer storing means;

means for accepting, for accepting required data for value-attributes;

means for comparing said required data with values in said value-fields and performing true-false tests for said value-fields;

means for tuple testing, for performing true-false tests of status factors of tuples in each one of said knowledge relations;

means for determining a set of tuples with true status factors; and

means for executing said decision statements in said set of tuples through said input/output means.

6. The relational knowledge-based system of claim 5 further comprises means for storing said required data in said computer storing means.

7. The required data of claim 6 are fed from outside said relational knowledge-based system into said computer storing means directly.

8. The set of inference engines of claim 5 further comprises means for querying environments of said relational knowledge-based system to obtain said required data.