Information processing device, information processing method, and information processing program

The information processing device generates Bayesian networks using nodes, edges, and weights to automatically calculate CPTs, addressing the challenge of incorporating human experience and reducing the need for mathematical knowledge in network design.

JP7587179B2Active Publication Date: 2024-11-20NIPPON TELEGRAPH & TELEPHONE CORP
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
JP2023501969
Authority / Receiving Office
JP · JP
Patent Type
Patents
Current Assignee / Owner
Filing Date
2021-02-26
Publication Date
2024-11-20
Estimated Expiration
2041-02-26

AI Technical Summary

Technical Problem

Designing Bayesian networks requires mathematical knowledge for setting conditional probability tables (CPTs), which is difficult and labor-intensive, and does not effectively incorporate human experience.

Method used

An information processing device and method that uses nodes, edges, and weights to generate Bayesian networks, automatically calculating CPTs based on input weights, allowing human experience to be reflected without requiring mathematical knowledge.

Benefits of technology

Enables the design of Bayesian networks that accurately reflect human experience, simplifying the process and eliminating the need for complex mathematical calculations.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 0007587179000004
    Figure 0007587179000004
  • Figure 0007587179000005
    Figure 0007587179000005
  • Figure 0007587179000006
    Figure 0007587179000006
Patent Text Reader

Abstract

An information processing device 1 for designing a Bayesian network is provided with: an input unit 11 which inputs, as Bayesian network elements, information relating to a plurality of nodes, edges indicating the cause-and-effect relationships between parent nodes and child nodes included in the plurality of nodes, and weights indicating the degrees of impact of the definition content of the parent nodes on the probability distributions of the child nodes; and a generation unit 12 which generates a Bayesian network using the input information.
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to an information processing device, an information processing method, and an information processing program. [Background technology]

[0002] AIOps (Artificial Intelligence for IT Operations) is known to automate and streamline IT operations by having AI learn various types of data, such as big data used in IT operations. For example, there is an orchestrator that sequentially uses AI to identify events, AI to identify the cause of the event, AI to search for ways to deal with the event, AI to identify the impact of the event, and AI to decide how to deal with the event.

[0003] In AIOps, if AI makes a wrong decision, there is a concern that the quality of service will decline due to the long recovery time, making it unsuitable for practical operation. In addition, correcting an erroneous decision by AI requires complex responses from highly skilled operators with operational skills, machine learning skills, and mathematical knowledge. Therefore, there is a need for technology to establish AI teaching technology that can infer the validity of the output results of various AIs, and to realize safe and secure operation by speeding up failure recovery and eliminating the need for skilled operators.

[0004] Therefore, there is a technology that can safely incorporate AI into network operations by verifying AI decisions and explaining the reasons for those decisions using a Bayesian network that mimics the flexible decisions of a veteran operator. The Bayesian network software allows the design and operation of Bayesian networks to be performed on a GUI basis.

[0005] When designing a Bayesian network, nodes indicating decision elements and edges (arrow lines) indicating causal relationships between the nodes are set, and the probabilities of child nodes are set in a CPT (Conditional Probability Table). For example, in the case of a Bayesian network for determining Loses, Sick, Dry, and Loses are set, with Sick and Dry set as the connection source and Loses set as the connection destination. In addition, the probability of occurrence and non-occurrence of Loses are set for the combination of the index value or probability of Sick and the index value or probability of Dry (see Non-Patent Document 1). [Prior art documents] [Non-patent literature]

[0006] [Non-Patent Document 1] HUGIN EXPERT, “Building a Bayesian Network”, [Retrieved February 10, 2021],<URL:https: / / hugin.com / wp-content / uploads / 2016 / 05 / Building-a-BN-Tutorial.pdf> Summary of the Invention [Problem to be solved by the invention]

[0007] In a Bayesian network, a graphical model consisting of nodes, edges, and CPTs makes it possible to design a judgment model that reflects human experience. However, while nodes and edges can be intuitively set based on knowledge, the probabilities of child nodes set in the CPT are calculated manually based on statistical information from operational data, which means that there is a large amount of work involved in generating a CPT, and it requires mathematical knowledge to format data in accordance with statistical methods, as well as advanced mathematical knowledge regarding conditional probabilities such as Bayesian inference.

[0008] The present invention has been made in consideration of the above circumstances, and an object of the present invention is to provide a technique that makes it possible to design a Bayesian network without requiring mathematical knowledge. [Means for solving the problem]

[0009] An information processing device according to one aspect of the present invention is an information processing device for designing a Bayesian network, the information processing device including an input unit for inputting information regarding a plurality of nodes as elements of the Bayesian network, edges indicating causal relationships between parent nodes and child nodes in the plurality of nodes, and weights indicating the degree of influence of the definition content of the parent node on the probability distribution of the child node, and a generation unit for generating the Bayesian network using the input information. An information processing method of one embodiment of the present invention is an information processing method performed by an information processing device that designs a Bayesian network, which includes the steps of inputting information regarding a plurality of nodes as elements of the Bayesian network, edges indicating causal relationships between parent nodes and child nodes at the plurality of nodes, and weights indicating the degree of influence that the definition content of the parent node has on the probability distribution of the child node, and generating a Bayesian network using the input information.

[0010] An information processing program according to one aspect of the present invention is an information processing program that causes a computer to function as the information processing device. Effect of the Invention

[0011] According to the present invention, a technique capable of designing a Bayesian network without requiring mathematical knowledge can be provided. [Brief description of the drawings]

[0012] [Figure 1] FIG. 1 is a diagram illustrating a functional block configuration of an information processing device. [Diagram 2] FIG. 2 is a diagram illustrating a processing flow of the information processing device. [Diagram 3] FIG. 3 is a diagram illustrating an example of a Bayesian network. [Figure 4] FIG. 4 is a diagram illustrating an example of a probability distribution of a child node. [Diagram 5]FIG. 5 is a diagram showing an example of a CPT of a child node. [Figure 6] FIG. 6 is a diagram illustrating a hardware configuration of an information processing device. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0013] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. In the description of the drawings, the same parts are given the same reference numerals and the description will be omitted.

[0014] [Summary of the Invention] In the present invention, in addition to nodes, edges, and CPTs, weights indicating the degree of influence of the definition of a parent node on the probability distribution of a child node are further input as definition elements of a Bayesian network, and a Bayesian network is generated using the input information. In other words, since a Bayesian network is generated based on weights indicating the degree of influence of the definition of a parent node on the probability distribution of a child node, it is possible to design a Bayesian network that reflects human experience without requiring mathematical knowledge.

[0015] Furthermore, the present invention automatically generates the CPT of the child node based on the weights. This allows business knowledge, business experience, etc. to be reflected in the Bayesian network using only the weight parameters, without directly setting the CPT. As a result, it becomes possible to design a Bayesian network that appropriately reflects human experience, without requiring mathematical knowledge.

[0016] Furthermore, the present invention varies the average value of the probability distribution according to the range of possible probability values ​​of the child node depending on the weight and the probability of the parent node, and adjusts the bias of the probability distribution of the child node based on the varied average value. This makes it possible to reflect business knowledge, business experience, etc. in the Bayesian network using only the weight parameters, without directly setting the CPT. As a result, it becomes possible to design a Bayesian network that more appropriately reflects human experience, without requiring mathematical knowledge.

[0017] [Configuration of information processing device] 1 is a diagram showing a functional block configuration of an information processing device 1 according to the present embodiment. The information processing device 1 is a device for designing a Bayesian network. The information processing device 1 includes an input unit 11, a generation unit 12, and a display unit 13.

[0018] The input unit 11 is a functional unit that inputs information regarding the following as definition elements of a Bayesian network: multiple nodes indicating decision elements, edges (arrow lines) indicating causal relationships between parent nodes and child nodes in the multiple nodes, a CPT for setting the probability of the child nodes (e.g., occurrence probability and non-occurrence probability), and weights indicating the degree of influence that the definition content of the parent node has on the probability distribution of the child nodes.

[0019] In this embodiment, the probability of a child node is exemplified by the occurrence probability and non-occurrence probability of the child node. In addition to the two values, the child node can also have three or more values. The parent node can also have two or more values.

[0020] The generating unit 12 is a functional unit that forms a Bayesian network using input information. The generating unit 12 is a functional unit that generates a CPT of a child node based on a weight. For example, the generating unit 12 calculates the occurrence probability and non-occurrence probability of a child node based on the weight and sets the probability in the CPT to generate a CPT of the child node. For example, the generating unit 12 calculates the average value of a probability distribution according to a range of possible probability values ​​of the child node based on the weight, calculates the occurrence probability and non-occurrence probability of the child node based on the probability distribution of the child node having the average value, and sets the probability in the CPT to generate a CPT of the child node. When calculating the occurrence probability and non-occurrence probability of the child node, the generating unit 12 varies the average value of the probability distribution according to the range of possible probability values ​​of the child node according to the weight and the index value or random variable (e.g., occurrence probability and non-occurrence probability) of the parent node, and adjusts the bias of the probability distribution of the child node based on the varied average value.

[0021] The display unit 13 is a functional unit that displays the definition elements of the Bayesian network, the design diagram of the Bayesian network, the CPT, and the like on a screen.

[0022] [Operation of information processing device] FIG. 2 is a diagram showing a processing flow of the information processing device 1 according to the present embodiment.

[0023] Step S1; First, the display unit 13 displays a GUI for setting definition elements of a Bayesian network on a screen. Next, the input unit 11 inputs a plurality of nodes indicating decision elements, edges indicating causal relationships between parent nodes and child nodes in the plurality of nodes, and weights indicating the degree of influence of the definition contents of the parent node on the probability distribution of the child node, which the user inputs to the GUI.

[0024] For example, node X1 of Pollution, node X2 of Smoker, and node X3 of Cancer are input. Also, edge E1 with node X1 as the source and node X3 as the destination, and edge E2 with node X2 as the source and node X3 as the destination are input. Furthermore, weight w1 for edge E1 and weight w2 for edge E2 are input.

[0025] The weights w1 and w2 are set to a larger weight for the parent node X1, X2 that is desired to have a greater influence on the child node X3. For example, since it is considered that smokers are more likely to cause cancer than pollution (smokers have a stronger influence on cancer), 0.7 is entered into weight w2 and 0.3 is entered into weight w1. The sum of weights w1 and w2 is 1.

[0026] This will display a Bayesian network as shown in Figure 3. The letters and numbers shown to the right of the parent nodes X1 and X2 are the index values ​​of the parent nodes. Instead of the index value of 0 or 1, the random variables (occurrence probability and non-occurrence probability) of the parent nodes may be used.

[0027] Step S2; Next, the generation unit 12 calculates the probability distribution of the child node X3 using equations (1) and (2).

[0028]

number

[0029]

number

[0030] Equation (1) defines the probability distribution f(z) of the child node X3. Equation (2) defines the mean value μ of the probability distribution f(z) using the weight w input in step S1. z is the random variable of the child node X3. σ is the variance of the probability distribution f(z). x j (j∈1,2,3) is node X j where M is the index value or random variable of the parent node X1 and the parent node X2. The probability distribution f(z) of the child node X3 defined by the formulas (1) and (2) is a normal distribution as shown in FIG.

[0031] As shown in formula (2), the average value μ of the probability distribution f(z) is calculated by multiplying the weight w by the index value or random variable of node X, so the larger the index value or random variable of node X, the more biased the probability distribution f(z) of child node X3 will be so that the probability of the child node is more likely to be a high value. Also, the larger the weight w of the parent node of the edge, the more biased the probability distribution f(z) of child node X3 will be so that the probability of the child node is more likely to be a high value.

[0032] In addition, as shown in formula (2), the sum of the products of the weight w and the index value or random variable of node X is divided by M. Therefore, when the cardinality value of the parent node tends to be larger, the probability of occurrence of a highly accurate initial value is obtained when the cardinality value of the child node tends to be larger as the cardinality value of the parent node is larger.

[0033] By using equations (1) and (2), the generation unit 12 varies the average value μ of the probability distribution according to the range of possible probability values ​​of the child node X3 depending on the weight w and the index value or random variable of the parent node X, and adjusts the bias f(z) of the probability distribution of the child node X3 based on the varied average value μ.

[0034] Next, the generation unit 12 calculates the probability P of the child node X3 using equation (3).

[0035]

number

[0036] For example, when the index value of child node X3 is max(X3), that is, 1, the generation unit 12 determines the integral value in the range from max(X3) to +∞ as the occurrence probability of child node X3. The generation unit 12 calculates the occurrence probability and non-occurrence probability of child node X3 for each of all combinations of the index values ​​that can be taken by the three nodes X1 to X3.

[0037] Finally, the generation unit 12 sets the occurrence probability and non-occurrence probability of the child node X3 for all combinations in the CPT (see FIG. 5). As a result, the generation unit 12 generates a CPT for the child node X3, and designs one Bayesian network together with the definition elements input in step S1.

[0038] Equations (1) to (3) are related to each other, and can be said to be a set of equations for generating a CPT of a child node based on the weight, and a set of equations for calculating the average value of the probability distribution of the child node based on the weight, and generating a CPT of the child node based on the probability distribution of the child node having the average value.

[0039] [effect] According to this embodiment, the information processing device 1 is equipped with an input unit 11 that inputs information regarding elements of a Bayesian network, such as multiple nodes, edges indicating causal relationships between parent nodes and child nodes at the multiple nodes, and weights indicating the degree of influence that the definition content of the parent node has on the probability distribution of the child node, and a generation unit 12 that generates a Bayesian network using the input information, so that it is possible to design a Bayesian network that reflects human experience without requiring mathematical knowledge.

[0040] [others] The present invention is not limited to the above-described embodiment, and various modifications are possible within the scope of the present invention.

[0041] The information processing device 1 of the present embodiment described above can be realized, for example, by using a general-purpose computer system including a CPU 901, a memory 902, a storage 903, a communication device 904, an input device 905, and an output device 906, as shown in Fig. 6. The memory 902 and the storage 903 are storage devices. In the computer system, the CPU 901 executes a predetermined program loaded on the memory 902, thereby realizing each function of the information processing device 1.

[0042] The information processing device 1 may be implemented by one computer. The information processing device 1 may be implemented by multiple computers. The information processing device 1 may be a virtual machine implemented in a computer. The program for the information processing device 1 may be stored in a computer-readable recording medium such as an HDD, SSD, USB memory, CD, or DVD. The program for the information processing device 1 may also be distributed via a communication network. [Explanation of symbols]

[0043] 1: Information processing device 11: Input section 12: Generation part 13:Display section 901:CPU 902: Memory 903: Storage 904: Communication equipment 905: Input device 906: Output device

Claims

1. In an information processing device for designing a Bayesian network, an input unit for inputting information regarding a plurality of nodes as elements of a Bayesian network, edges indicating causal relationships between parent nodes and child nodes in the plurality of nodes, and weights indicating the degree of influence of the definition content of the parent node on the probability distribution of the child node; A generating unit that generates a Bayesian network using the input information, The generation unit is an information processing device that calculates an average value of a probability distribution according to a range of possible probability values ​​of the child node based on the weight, and generates a conditional probability table for the child node based on the probability distribution of the child node having the average value.

2. The generation unit is 2. The information processing device according to claim 1, wherein an average value of a probability distribution according to a range of possible probability values ​​of a child node is varied according to the weight and the probability of the parent node, and a bias in the probability distribution of the child node is adjusted based on the varied average value.

3. An information processing method performed by an information processing device for designing a Bayesian network, comprising: A first step of inputting information regarding a plurality of nodes as elements of a Bayesian network, edges indicating causal relationships between parent nodes and child nodes in the plurality of nodes, and weights indicating the degree of influence of the definition content of the parent node on the probability distribution of the child node; a second step of generating a Bayesian network using the input information; In the second step, An information processing method that calculates an average value of a probability distribution corresponding to the range of possible probability values ​​of a child node based on the weight, and generates a conditional probability table for the child node based on the probability distribution of the child node having the average value.

4. An information processing program that causes a computer to function as the information processing device according to claim 1 or 2.

Citation Information

Patent Citations

  • Information processor, information processing system and information processing method

    JP2011028339A

  • Device and method generating model, information processing system, and program

    JP2011039734A