JP2003173259A - Rule forming supporting device, control method of rule forming supporting device, control program and recording medium - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To easily obtain a setting combination of an optimal certainty factor when applying an IF-THEN form rule using a certainty factor. <P>SOLUTION: A supposition output database 16 prestores supposition output data DREF having a supposition conclusion certainty factor being a conclusion certainty factor supposed in response to supposed input data. An individual forming device 17 forms a new rule including a new rule certainty factor on the basis of an existing rule including a rule certainty factor and genetic algorithm. An output device 13 forms output data DOUT including a conclusion certainty factor by applying a rule formed to input data DIN inputted via an input device 11. As a result of these, an evaluating device 19 evaluates whether or not to adopt the new rule on the basis of the supposition conclusion certainty factor (the supposed output data DREF) and the conclusion certainty factor (the output data DOUT). <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a rule generation support device, a control method, a control program and a recording medium,
In particular, the present invention relates to a rule generation support device, a control method for the rule generation support device, a control program, and a recording medium that can be applied to an expert system using confidence as a method for expressing fragmentary knowledge that represents ambiguity.

[0002]

2. Description of the Related Art As one of methods for expressing a piece of knowledge representing ambiguity in a conventional expert system,
A method of performing rule expression in the IF-THEN format using the certainty factor is known. IF-THE using this certainty factor
N-type rules form part of the knowledge.
Therefore, in order to express the entire knowledge by the IF-THEN format rule using the certainty factor, it is necessary to set and input many rules. However, as the number of rules increases, it becomes more difficult to manage the rules as a whole. In particular, regarding the certainty factor, even if an appropriate value is set for each rule, the entire rule, that is, the entire expert system may not be able to obtain an appropriate conclusion (output value) for the input value. It was supposed to happen.

In order to solve this, conventionally, the conclusion (output value) obtained for the input value by the rule creator
It was common to adjust the value of the confidence factor of each rule by trial and error so as to obtain the confidence factor assumed by. Further, in the "fuzzy inference rule automatic generation system and its apparatus" described in Japanese Patent Laid-Open No. 7-28649, a membership function is automatically corrected using a genetic algorithm for fuzzy inference. Further, in a system using a neural network, it is possible for each neuron to learn and correct the conclusion for each input.

[0004]

By the way, depending on a combination of certain input data, it is easily possible that the certainty factor of the output data, which is the conclusion, does not reach the expected value. In such a case, if the certainty factor of any one of the rules is changed for correction, the desired certainty factor is obtained before the change, even if the expected certainty factor is obtained for the input data. In addition, there is a possibility that a desired confidence level cannot be obtained with respect to other input data, resulting in a vicious circle. In this case, the certainty factor combination number X
Is expressed by the following equation, where A is the number of combinations of input data, B is the number of established rules, and C is the number of settable confidence values. X = A × B × C

In this case, the number of settable confidence values is 11 steps of step 0.1 in the range of 0.0 to 1.0 at the minimum, and the number X of set combinations of confidence is enormous. It's easy to imagine that. Therefore, there is a problem in that it is difficult to find the optimum combination of certainty factors. Fuzzy reasoning is also suitable for systems that deal with co-continuous input and output, such as when the input is speed and the output is brake strength, and it is suitable for systems that present items such as diagnostic systems and troubleshooting. There was a problem that it was difficult to apply. In addition, a system using a neural network has a problem that it is difficult to set a learning standard in the first place, and if incorrect learning is performed during learning, it becomes difficult to cope with it thereafter. . Therefore, an object of the present invention is to provide a rule generation support device, a control method, and a control program that can easily obtain an optimal confidence combination setting when applying a rule of IF-THEN format using confidence. And a recording medium.

[0006]

[Means for Solving the Problems] To solve the above problems, a rule generation support device stores in advance assumed output data having an expected conclusion certainty factor which is a certainty certainty factor associated with expected input data. An assumed output data storage unit, a rule generation unit that generates a new rule including a new rule certainty factor based on an existing rule including a rule certainty factor and a genetic algorithm, and the rule generated for the input data. An output data generation unit that applies and outputs output data including a conclusion certainty factor, and performs a rule evaluation process that evaluates whether or not to adopt the new rule based on the assumed conclusion certainty factor and the conclusion certainty factor. It is characterized by having an evaluation section. According to the above configuration, the assumed output data storage unit stores in advance the assumed output data having the expected conclusion certainty factor which is the expected certainty factor corresponding to the assumed input data. The rule generation unit generates a new rule including a new rule certainty factor based on an existing rule including a rule certainty factor and a genetic algorithm. The output data generation unit applies the generated rule to the input data and generates output data including the conclusion certainty factor. The evaluation unit performs a rule evaluation process for evaluating whether or not to adopt the new rule based on the expected conclusion certainty factor and the conclusion certainty factor.

In this case, in the rule evaluation processing, the evaluation section performs an evaluation calculation using the assumed conclusion certainty factor and the conclusion certainty factor, and the value of the evaluation calculation corresponds to the assumed conclusion certainty factor. It may be configured such that the corresponding new rule is a rule to be adopted when the value becomes equal to or less than a predetermined threshold value determined by the above.

Further, the assumed output data corresponds to a plurality of conclusion item data and a plurality of assumed item conclusion certainty factors A1 corresponding to the respective conclusion item data and constituting the assumed conclusion certainty factor.
To An, the output data includes a plurality of conclusion item data and a plurality of item conclusion certainty factors B1 to Bn corresponding to each of the conclusion item data items and constituting the conclusion certainty factor, and the value X of the evaluation operation is It may be defined by the following formula.

[0009]

[Equation 4]

Further, an adoption rule storage section for storing the rules to be adopted may be provided.

Furthermore, the evaluation unit interrupts the rule evaluation processing when a predetermined time has elapsed since the rule evaluation processing was started, and adopts the rule evaluation at the time of interruption. May be.

Further, the genetic algorithm may use the rule certainty factor corresponding to an existing rule as a gene and generate a new rule including the new rule certainty factor which is a next-generation gene. .

Further, an assumed output data storing process for pre-storing assumed output data having an expected conclusion certainty which is an expected conclusion certainty corresponding to the assumed input data, an existing rule including rule certainty and a genetic rule. A rule generation process for generating a new rule including a new rule certainty factor based on an algorithm, and an output data generation process for applying the rule generated for the input data to generate output data including a conclusion certainty factor And an evaluation process of performing a rule evaluation process for evaluating whether or not to adopt the new rule based on the assumed conclusion certainty factor and the conclusion certainty factor.

In this case, in the evaluation process, in the rule evaluation processing, an evaluation calculation is performed using the assumed conclusion certainty factor and the conclusion certainty factor, and the value of the evaluation calculation corresponds to the assumed conclusion certainty factor. It may be configured such that the corresponding new rule is a rule to be adopted when the value becomes equal to or less than a predetermined threshold value determined by the above.

Further, the assumed output data is a plurality of assumed item conclusion certainty factors A1 corresponding to a plurality of conclusion item data and each of the concluded item data and constituting the assumed conclusion certainty factor.
To An, the output data includes a plurality of conclusion item data and a plurality of item conclusion certainty factors B1 to Bn corresponding to each of the conclusion item data items and constituting the conclusion certainty factor, and the evaluation process is performed by the following formula. The evaluation may be performed based on the value X of the evaluation calculation defined by.

[0016]

[Equation 5]

Further, an adoption rule storing step for storing the rule to be adopted may be provided.

Furthermore, in the evaluation process, the rule evaluation process is interrupted when a predetermined time has elapsed since the rule evaluation process was started, and the rule evaluation at the time of interruption is adopted. May be.

In the rule generation process, the genetic algorithm having the rule certainty factor corresponding to an existing rule as a gene generates a new rule including the new rule certainty factor which is a next-generation gene. You may do it.

Further, the control program for causing the computer to function as the rule generation assisting device prestores the expected output data having the expected conclusion certainty, which is the expected conclusion certainty corresponding to the assumed input data, and stores the rule. Generate a new rule including a new rule certainty based on an existing rule including certainty and a genetic algorithm, apply the generated rule to the input data, and generate output data including a conclusion certainty Let the evaluation whether or not to adopt the new rule based on the assumed conclusion certainty and the conclusion certainty,
It is characterized by that.

In this case, in the evaluation, an evaluation calculation is performed using the assumed conclusion certainty factor and the conclusion certainty factor, and a value of the evaluation calculation is determined according to the assumed conclusion certainty factor. If the threshold value is less than or equal to the threshold value, the corresponding new rule may be a rule to be adopted.

Further, the assumed output data is a plurality of expected item conclusion certainty factors A1 corresponding to a plurality of conclusion item data and each of the conclusion item data and constituting the assumed conclusion certainty factor A1.
To An, the output data includes a plurality of conclusion item data and a plurality of item conclusion certainty factors B1 to Bn corresponding to each of the conclusion item data items and constituting the conclusion certainty factor, and is defined by the following equation. The evaluation may be performed based on the value X of the evaluation calculation.

[0023]

[Equation 6]

Further, the rules to be adopted may be stored.

Furthermore, the evaluation process may be interrupted when a predetermined time has elapsed since the evaluation process was started, and the rule evaluation at the time of interruption may be adopted.

Further, by the genetic algorithm having the rule certainty factor corresponding to the existing rule as a gene,
A new rule including the new rule certainty, which is a next-generation gene, may be generated. Further, each of the above control programs may be recorded in a recording medium.

[0027]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Next, preferred embodiments of the present invention will be described with reference to the drawings. FIG. 1 shows a schematic block diagram of the expert system of the embodiment. The expert system 10 is roughly classified into an input device 11, an execution device 12, an output device 13, and a rule-based database 1.
4, a data storage database 15, an assumed output database 16, an individual generation device 17, an individual database 18, an evaluation device 19 and an optimum individual database 20.

In this case, the assumed output database 16 functions as an assumed output data storage unit, the individual generation device 17 functions as a rule generation unit that generates a new rule (individual data), and the execution device 12 The evaluation device 19 functions as an output data generation unit, and functions as an evaluation unit. Further, the expert system 10 in the present embodiment is actually configured by a computer including a microprocessor unit (MPU), a ROM, a RAM, an external storage device, an input / output device, a display device, etc., which are not shown. Further, the functions of the execution device 12, the individual generation device 17, the evaluation device 19 and the like are realized by a program executable by the microprocessor unit.
In addition, such a program can be stored in a semiconductor memory, a CD-
It may be directly executed from a recording medium such as a ROM. Also,
It is also possible to install the program in an external storage device in advance and execute it.

The input device 11 is the expert system 1
It is a device for a user of 0 to input various input data DIN. As the input data DIN, for example, when the expert system 10 is a medical diagnosis system, data corresponding to the condition of the patient is input. The execution device 12 outputs the output data DOUT, which is the processing execution result corresponding to the input data DIN, to the output device 13 based on the rule base stored in the rule base database 14, and also the input data DIN and the output data D.
The OUT is registered in the data storage database 15.

The output device 13 displays the processing execution result from the execution device 12. Rule base database 14
Stores a pre-built rule base. The data storage database 15 stores the input data DIN and the output data DOUT under the control of the execution device 12. The assumed output database 16 is assumed output data DRE which the rule base creator assumed in advance to be output by the execution device 12.
Stores F.

The individual generation device 17 generates a new individual from an existing individual based on rule-based data and a genetic algorithm. Here, the individual data DIDV is
It is represented by a plurality of genes (combination of certainty factors for a plurality of events in this embodiment). The individual database 18 stores the individual data DIDV generated by the individual generation device 17. The evaluation device 19 uses the input data DIN
On the other hand, the output data obtained by applying the individual data DIDV stored in the individual database 18 is compared with the assumed output data DREF to determine whether or not more suitable individual data DIDV is obtained. Optimal individual database 20
Stores more suitable (ideally optimal) individual data DIDV under the control of the evaluation device 19.

Next, a specific operation will be described. By the way, in the usual expert system, the rules stored in the rule base database 14 are 1000 to 2000.
The degree is set. However, in the following description, four rules (the first rule R1
~ The case of the fourth rule R4) will be described.

First rule R1 If the upper right corner of your stomach hurts, Cholelithiasis (0.6) Cholecystitis (0.5) Duodenal ulcer (0.4)

Second rule R2 If your middle abdomen hurts, Gastric ulcer (0.5) Duodenal ulcer (0.4) Acute gastritis (0.6) Cholelithiasis (0.2)

・ Third rule R3 If the lower right corner of your stomach hurts, Colonic diverticulitis (0.1) Ureteral stones (0.2) Acute appendicitis (0.3)

4th rule R4 If your upper right stomach hurts continuously, Cholecystitis (0.4) Liver abscess (0.3) Perforated duodenal ulcer (0.2) Pneumonia, pleuritis (0.1)

The meaning of each of the rules R1 to R4 is, for example, for the first rule R1, the degree to which one can be convinced that it is cholelithiasis (confidence factor) when the data that the upper right abdomen is painful is input. Is 0.6, the degree of belief that cholecystitis is confident (confidence) is 0.5, and the degree that the belief of duodenal ulcer (confidence) is 0.
That is four. In this case, a certainty factor of 1 means an affirmative assertion, and a certainty factor of 0 means a negative assertion. That is, if the certainty factor of cholelithiasis is 1, it is determined to be gallstone disease, and conversely, if the certainty factor of cholelithiasis is 0, it is to determine that it is not gallstone disease.

Next, the operation of the embodiment when generating the optimum rule will be described. FIG. 2 shows an outline processing flowchart of the expert system. First, in the expert system 10, input data DIN is input, output data DOUT is generated by application of a rule base by the execution device, and input data DIN and output data DOUT are stored (step S1). More specifically, the expert system 10
The user inputs the input data DIN via the input device 11. For example, assume that the following data is input as the input data DIN. Input data DIN = "I have a pain in the upper right of my stomach" and "I have a pain in my lower right of my stomach"

As a result, the execution device 12 outputs the process execution result corresponding to the input data DIN to the output device 13 based on the rule base stored in the rule base database 14, that is, the above-mentioned rules R1 to R4. . At the same time, the execution device 12 registers the input data DIN in the input database 15. Next, the individual generation device 17 of the expert system 10 refers to the rule-based database 14 to generate initial individual data IDV (step S1). In the following description, the individual data in the initial state will be referred to as first generation individual data IDV1 for convenience.

By the way, the rules R1 to R4 described above correspond to one individual data IDV as a whole, and in the initial state, correspond to the first generation individual data IDV1. That is, the first generation individual data DIDV1
As shown in FIG. 3, the confidence value is expressed as a gene for each of the rules R1 to R4, and only the confidence value is stored as the four genes of the first generation individual data IDV1. . More specifically, the first generation individual data I DV1
Are the genes corresponding to the first rule R1 "0.
6, 0.5, 0.4 ", gene" 0.5, 0.4, 0.6, 0.2 "corresponding to second rule R2, third rule R
Genes corresponding to 3 "0.1, 0.2, 0.3" and genes corresponding to 4th rule R4 "0.4, 0.3,"
It is expressed as four genes of "0.2,0.1".

In the initial state, the input data DIN is input, and the output data DOUT which is the processing execution result output from the execution device 12 is the input obtained by reading the first generation individual data IDV1 from the data storage database 15 by the evaluation device 19. It is the same as the first generation output data DOUT1 obtained by applying it to the data DIN. Output data DOU in this initial state
T, that is, the first generation output data DOUT1 is displayed on the display screen of the output device 13. In addition, () is a certainty factor.・ Initial state output data DOUT = 1st generation output data DO
UT1: Cholelithiasis (0.6) Cholecystitis (0.5) Duodenal ulcer (0.4) Colon diverticulitis (0.1) Ureteral stone (0.2) Acute appendicitis (0.3)

In parallel with this, the evaluation device 19 reads the assumed output data DOUTP corresponding to the input data DIN from the assumed output database 16 and evaluates the individual data (step S3). More specifically, assumed output data DOU
The case where the TP is as follows is explained.・ Assumed output data DOUTP: Cholelithiasis (0.2) Cholecystitis (0.2) Duodenal ulcer (0.3) Colon diverticulitis (0.1) Ureteral stone (0.1) Acute appendicitis (0.7)

First, the principle operation of the evaluation processing in the evaluation device 19 will be described. As described above, the assumed output data DOUTP is composed of a plurality of assumed conclusion item data (corresponding to each item of cholelithiasis, cholecystitis, duodenal ulcer, colonic diverticulitis, ureteral stone or acute appendicitis) and each assumed conclusion item data. Corresponding corresponding assumption conclusions A plurality of assumption items conclusion confidences A1 to An (0.2, 0.2, 0.3, 0.1,
(Corresponding to each value of 0.1 and 0.7). On the other hand, the output data DOUT corresponds to a plurality of conclusion item data (corresponding to each item of cholelithiasis, cholecystitis, duodenal ulcer, colonic diverticulitis, ureteral stone or acute appendicitis) and each conclusion item data, and has a conclusion certainty factor. Consistency of a plurality of items Conclusion confidence B1 ~
Bn (0.6, 0.5, 0.4, 0.1, 0.2, 0.
(Corresponding to each value of 3) is included. Therefore, the evaluation device 19
Is the evaluation data X that is the value of the evaluation operation defined by the following equation.
Based on.

[0044]

[Equation 7]

In the case of the above example, the first generation individual data IDV1
The evaluation data X1 corresponding to the evaluation data X is as follows. X1 = | 0.6-0.2 | + | 0.5-0.2 | + | 0.4-0.3 | + | 0.1-0.1 | + | 0.2-0.1 | + | 0.3-0.7 | = 0.4 + 0.3 + 0.1 + 0.0 + 0.1 + 0.4 = 1.3 By the way, what is meant by the value of the evaluation data X is that the evaluation data X = 0. , Output data DOUTX = estimated output data DOUTP. It is assumed that the rule base creator of the expert system 10 should output the execution device 12 if the individual data IDV used when the output data DOUTX is generated is used. The output data can be output.

That is, the closer the value of the evaluation data X is to 0, the more appropriate the individual data DIDV, and thus the rule having the same content as the gene of the individual data DIDV can be adopted as the more appropriate rule. . Next, the evaluation device 19 determines whether the value of the evaluation data X is within a predetermined threshold value or whether the total time of the evaluation process exceeds the corresponding threshold value (step S4).
Here, the reason for this determination processing will be described. For the reason described above, ideally, when a value of the evaluation data X closer to 0 is obtained for the same input data DIN, the evaluation device 19 stores the evaluation data X in the optimum individual database 20. The individual data DIDV corresponding to should be registered.

However, in reality, the evaluation data X
If the value of is less than or equal to a value (threshold value) determined to be allowable in advance, there is no problem even if the evaluation device 19 registers the individual data corresponding to the evaluation data X. In addition, in the actual expert system, since the number of rules is as large as 1000 to 2000, the combinations of genes are almost infinite. Therefore, when the total processing time for obtaining the evaluation data X corresponding to the plurality of input data DIN exceeds the time (threshold value) which is determined to be allowable in advance, the time (threshold value) is exceeded. The evaluation device 1 calculates the individual data DIDV corresponding to the evaluation data X having the smallest value at the timing.
This is because it is realistic for 9 to register.

When the value of the evaluation data X is within the predetermined threshold value or the total time of the evaluation processing exceeds the corresponding threshold value in the determination of step S4 (step S4; Yes), Optimal individual database 20
At this point, the optimum individual data DIDV is stored and the process ends (step S5). On the other hand, in the determination of step S4, if the value of the evaluation data X exceeds the predetermined threshold value and the total time of the evaluation processing is less than the corresponding threshold value, the evaluation processing This means that the time given for that is left, so that in order to obtain a more optimal individual, an individual of the next generation is generated (step S2), and thereafter, the same processing is performed to obtain more optimal individual data. The DIDV is obtained, stored in the optimum individual database 20 and the process is terminated (step S5).

Specifically, the second generation individual data DIDV2
Similarly to the first-generation individual data DIDV1, as shown in FIG. 4, the value of the certainty factor is expressed as a gene for each of the rules R1 to R4, and only the value of the certainty factor is the second-generation individual data IDV2. It is stored as four genes of. In the present embodiment, the genes of the second generation individual data IDV2 are the genes “0.6, 0.5, 0.
3 ", the gene corresponding to the second rule R2 is" 0.5,0.
5, 0.6, 0.2 ", gene" 0.1, 0.2, 0.2 "corresponding to the third rule R3 and gene" 0.4, 0.3, corresponding to the fourth rule R4. It is expressed as four genes of "0.2, 0.3". Then, the same input data DIN is input, and the output data DOUT that is the processing execution result output from the execution device 12, that is, the second generation output data DOUT21, is similarly displayed on the display screen of the output device 13. In addition, () is a certainty factor.

Second generation output data DOUT2: Cholelithiasis (0.6) Cholecystitis (0.5) Duodenal ulcer (0.3) Colonic diverticulitis (0.1) Ureteral stones (0.1) Acute appendicitis (0.2)

In the case of the above example, the second generation individual data IDV2
The evaluation data X2 corresponding to the evaluation data X is as follows. X2 = | 0.6-0.2 | + | 0.5-0.2 | + | 0.3-0.3 | + | 0.1-0.1 | + | 0.1-0.1 | + | 0.2-0.7 | = 0.4 + 0.3 + 0.0 + 0.0 + 0.0 + 0.5 = 1.2

Therefore, in this case, the second-generation individual data IDV2 is larger than the first-generation individual data IDV1 in the evaluation data X.
Is small. That is, since X2 <X1, it means that the second-generation individual data IDV2 is more suitable than the first-generation individual data IDV1. Therefore, the evaluation device 19 determines that more suitable individual data is obtained for the input data DIN, and stores the second-generation individual data IDV2 in the optimum individual database 20.

Similarly, the expert system 1
A value of 0 continues to obtain more suitable individual data until the value of the evaluation data X falls within a predetermined threshold value or the total time of evaluation processing exceeds the corresponding threshold value.
When the value of the evaluation data X is within a predetermined threshold value or when the total time of the evaluation processing exceeds the corresponding threshold value, the individual data that seems to be most suitable by then is set as the optimum individual data. It will be stored in the database 20. As described above, according to the present embodiment, it is not necessary to set the confidence factor for each rule,
It is possible to prevent the certainty factor of a conclusion from being inappropriate for a certain input data by correcting the certainty factor of a rule automatically by a genetic algorithm.

In the above description, only the expert system has been described in the field of application of the present invention, but a failure diagnosis system for diagnosing various failures, a consulting system for performing various kinds of consulting, and a product that meets the customer's desired conditions. The present invention can be applied to rule making support in a product selection system that presents information to prompt product selection, an advisor system that presents various kinds of advice, or a robot action planning system that plans robot actions.

[0055]

According to the present invention, it is not necessary to set a certainty factor for each rule, and it is automatically determined by a genetic algorithm that the certainty factor of a conclusion has an inappropriate value for a certain input data. It can be avoided by modifying the confidence setting of the rule.

[Brief description of drawings]

FIG. 1 is a schematic block diagram of an expert system.

FIG. 2 is an outline processing flowchart of an expert system.

FIG. 3 is an explanatory diagram of genes of first generation individuals.

FIG. 4 is an explanatory diagram of genes of a second generation individual.

[Explanation of symbols]

10 Expert system 11 Input device 12 Execution device 13 Output device (output data generation unit) 14 Rule base database 15 Data storage database 16 Assumed output database (Assumed output data storage unit) ) 17 ... Individual generation device (rule generation unit) 18 ... Individual database 19 ... Evaluation device (evaluation unit) 20 ... Optimal individual database DIN ... Input data DOUT ... Output data DREF ... Assumed output data DIDV ... … Individual data R1 to R4 …… Rules

Claims (19)

[Claims] 1. An assumed output data storage unit for pre-storing assumed output data having an expected conclusion certainty which is an expected conclusion certainty corresponding to expected input data, an existing rule including rule certainty and a genetic rule. A rule generation unit that generates a new rule including a new rule certainty factor based on an algorithm, and applies the generated rule to the input data,
An output data generation unit that generates output data including a conclusion certainty factor, and an evaluation unit that performs a rule evaluation process that evaluates whether or not to adopt the new rule based on the assumed conclusion certainty factor and the conclusion certainty factor. A rule generation support device comprising: 2. The rule generation support device according to claim 1, wherein in the rule evaluation processing, the evaluation unit performs an evaluation calculation using the assumed conclusion certainty factor and the conclusion certainty factor, and a value of the evaluation calculation. The rule generation support device is characterized in that, when is less than or equal to a predetermined threshold value determined corresponding to the assumed conclusion certainty factor, the corresponding new rule is a rule to be adopted. 3. The rule generation support device according to claim 2, wherein the assumed output data corresponds to a plurality of conclusion item data and a plurality of assumed item conclusion beliefs corresponding to each of the conclusion item data and constituting the assumed conclusion certainty factor. Including the degrees A1 to An, the output data includes a plurality of conclusion item data and a plurality of item conclusion certainty degrees B1 to Bn corresponding to each of the conclusion item data and constituting the conclusion certainty degree, and the value of the evaluation operation. A rule generation support device characterized in that X is defined by the following equation. [Equation 1] 4. The rule generation support device according to claim 2, further comprising an adoption rule storage unit that stores the rule to be adopted. 5. The rule generation support device according to claim 1, wherein the evaluation unit interrupts the rule evaluation process when a predetermined time has elapsed from the start of the rule evaluation process, and interrupts the rule evaluation process. A rule generation support device characterized by adopting evaluation of rules at a time point. 6. The rule generation support device according to claim 1, wherein the genetic algorithm uses the rule certainty factor corresponding to an existing rule as a gene, and includes the new rule certainty factor that is a next-generation gene. A rule generation support device characterized by generating a new rule. 7. An assumed output data storing process for pre-storing assumed output data having an assumed conclusion certainty, which is an expected conclusion certainty corresponding to expected input data, and an existing rule including rule certainty and genetic information. A rule generation process of generating a new rule including a new rule certainty factor based on an algorithm, and applying the rule generated to the input data,
An output data generation process for generating output data including a conclusion certainty factor, and an evaluation process for performing a rule evaluation process for evaluating whether or not to adopt the new rule based on the assumed conclusion certainty factor and the conclusion certainty factor. A method for controlling a rule generation support device, comprising: 8. The control method of the rule generation support device according to claim 7, wherein in the evaluation process, an evaluation calculation is performed using the assumed conclusion certainty factor and the conclusion certainty factor in the rule evaluation process, and the evaluation is performed. When the value of the operation is equal to or less than a predetermined threshold value that is determined corresponding to the assumed conclusion certainty factor, the corresponding new rule is determined to be a rule to be adopted. Control method of support device. 9. The control method for a rule generation support device according to claim 8, wherein the assumed output data corresponds to a plurality of conclusion item data and a plurality of assumptions corresponding to each of the conclusion item data and constituting the assumed conclusion certainty factor. The item conclusion certainty degrees A1 to An are included, and the output data includes a plurality of conclusion item data and a plurality of item conclusion certainty degrees B1 to Bn corresponding to the respective conclusion item data and constituting the conclusion certainty degree, and the evaluation is performed. The process is the value X of the evaluation operation defined by the following equation.
A method for controlling a rule generation support device, characterized in that the evaluation is performed based on the above. [Equation 2] 10. The control method of the rule generation support device according to claim 8, further comprising an adoption rule storage step of storing the rule to be adopted. . 11. The method for controlling a rule generation support device according to claim 7, wherein the evaluation process interrupts the rule evaluation process when a predetermined time has elapsed since the rule evaluation process was started. Then, a rule generation support device control method characterized in that the evaluation of the rule at the time of interruption is adopted. 12. The method of controlling a rule generation support device according to claim 7, wherein in the rule generation process, a next-generation gene is generated by the genetic algorithm having the rule certainty factor corresponding to an existing rule as a gene. A method of controlling a rule generation assisting apparatus, comprising: generating a new rule including the certain new rule certainty factor. 13. A control program for causing a computer to function as a rule generation support device, pre-stores expected output data having an expected conclusion certainty degree which is an expected conclusion certainty degree associated with assumed input data. A new rule including a new rule certainty factor is generated based on an existing rule including certainty factor and a genetic algorithm, the generated rule is applied to the input data, and output data including a conclusion certainty factor is generated. A control program, characterized in that it is evaluated whether or not to adopt the new rule based on the assumed conclusion certainty factor and the conclusion certainty factor. 14. The control program according to claim 13, wherein in the evaluation, an evaluation calculation is performed using the assumed conclusion certainty factor and the conclusion certainty factor, and the value of the evaluation calculation corresponds to the assumed conclusion certainty factor. The control program is characterized in that the corresponding new rule is a rule to be adopted when it becomes equal to or less than a predetermined threshold value determined by the above. 15. The control program according to claim 14, wherein the assumed output data corresponds to a plurality of conclusion item data and a plurality of assumed item conclusion certainty factors A1 corresponding to the respective conclusion item data and constituting the assumed conclusion certainty factor. -An, the output data includes a plurality of conclusion item data and a plurality of item conclusion certainty factors B1 to Bn corresponding to each of the conclusion item data items and constituting the conclusion certainty factor, and is defined by the following equation. A control program that causes the evaluation to be performed based on a value X of an evaluation calculation. [Equation 3] 16. The control program according to claim 14 or 15, wherein the rule to be adopted is stored. 17. The control program according to claim 16, wherein the evaluation processing is interrupted when a predetermined time has elapsed since the evaluation processing was started, and the rule evaluation at the time of interruption is adopted. A control program characterized by: 18. The control program according to claim 16, wherein the genetic algorithm that uses the rule certainty factor corresponding to an existing rule as a gene includes a new rule certainty factor that is a next-generation gene. A control program characterized by generating rules. 19. A recording medium on which the control program according to any one of claims 13 to 18 is recorded.