Information processing apparatus, information processing method, and program

Abstract

To solve the problem that it is difficult to find a proper rule set model.SOLUTION: An operation recognition device comprises: a generation unit which generates a plurality of rule set models satisfying a number-of-rule constraint representing a constraint on the number of rules to be combined, on the basis of prediction performance for training data by a rule set model in which rules for prescribed prediction on the training data are combined; and a selection unit which selects and outputs a model group consisting of a combination of rule set models satisfying a number-of-model constraint representing a constraint on the number of models to be combined, on the basis of positions to which the rule set models correspond in a space having the prediction performance and the number of rules as axes.SELECTED DRAWING: Figure 7

Description

[Technical Field]

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

[0002] Predictions for input data are made using machine learning models in a variety of fields. One example of a machine learning model is a rule-based model that is easy to interpret, as described in Patent Document 1. In a rule-based model, a set of training cases is used to learn a rule set model consisting of a combination of multiple rules. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] International Publication No. 2022 / 044221 Summary of the Invention [Problem to be solved by the invention]

[0004] However, in a rule set model consisting of a combination of multiple rules, there is a trade-off between predictive performance and interpretability, which makes it difficult to find an appropriate rule set model that has excellent predictive performance and interpretability.

[0005] Therefore, one of the objectives of the present disclosure is to solve the above-mentioned problem of the difficulty in finding an appropriate rule set model. [Means for solving the problem]

[0006] An information processing device according to an embodiment of the present disclosure includes: a generation unit that generates a plurality of rule set models that satisfy a number of constraint rules representing a constraint on the number of rules that can be combined, based on prediction performance for training data by the rule set models, each of which is formed by a combination of rules that make a predetermined prediction for the training data; a selection unit that selects and outputs a model group made up of combinations of the rule set models that satisfy a constraint model number that represents a constraint on the number of models that can be combined, based on a position corresponding to the rule set model in a space whose axes are prediction performance and the number of rules; Equipped with The structure is as follows. Furthermore, an information processing method according to an embodiment of the present disclosure includes: generating a plurality of rule set models that satisfy a number of constraint rules representing constraints on the number of rules that can be combined, based on prediction performance for the training data by the rule set models each consisting of a combination of rules that make a predetermined prediction for the training data; selecting and outputting a model group consisting of a combination of the rule set models that satisfies a constraint model number that represents a constraint on the number of models that can be combined, based on a position corresponding to the rule set model in a space whose axes are prediction performance and the number of rules; The structure is as follows. Furthermore, a program according to an embodiment of the present disclosure includes: generating a plurality of rule set models that satisfy a number of constraint rules representing constraints on the number of rules that can be combined, based on prediction performance for the training data by the rule set models each consisting of a combination of rules that make a predetermined prediction for the training data; selecting and outputting a model group consisting of a combination of the rule set models that satisfies a constraint model number that represents a constraint on the number of models that can be combined, based on a position corresponding to the rule set model in a space whose axes are prediction performance and the number of rules; Have the computer perform the process, The structure is as follows. [Effects of the Invention]

[0007] With the present disclosure configured as described above, it is possible to easily find an appropriate rule set model. [Brief explanation of the drawings]

[0008] [Figure 1] 1 is a block diagram illustrating an example of a configuration of an information processing device according to the present disclosure. [Figure 2] FIG. 1 is a diagram illustrating an example of a process performed by an information processing device according to the present disclosure. [Figure 3] FIG. 1 is a diagram illustrating an example of a process performed by an information processing device according to the present disclosure. [Figure 4] FIG. 1 is a diagram illustrating an example of a process performed by an information processing device according to the present disclosure. [Figure 5] 10 is a flowchart illustrating an example of a processing operation of an information processing device according to the present disclosure. [Figure 6] FIG. 1 is a block diagram illustrating an example of a hardware configuration of an information processing device according to the present disclosure. [Figure 7] 1 is a block diagram illustrating an example of a configuration of an information processing device according to the present disclosure. DETAILED DESCRIPTION OF THE INVENTION

[0009] First Embodiment A first embodiment of the present disclosure will be described with reference to the drawings, which may be relevant to any embodiment.

[0010] The information processing device 10 in this embodiment generates a rule-based model using training case data. In particular, in this embodiment, the information processing device 10 generates a rule set model consisting of a combination of rules, and further selects and outputs a model group consisting of a combination of multiple rule set models with good predictive performance and interpretability. This allows the user to be presented with a model group consisting of multiple rule set models with good predictive performance and interpretability, and to use one of these rule set models. In this embodiment, a rule set model with good interpretability means that the rule set model contains a small number of rules.

[0011] The information processing device 10 is composed of one or more information processing devices each including a calculation device and a storage device. As shown in FIG. 1, the information processing device 10 includes an input unit 11, a model generation unit 12, a group selection unit 13, and a model search unit 14. The functions of the input unit 11, the model generation unit 12, the group selection unit 13, and the model search unit 14 can be realized by the calculation device executing a program for realizing each function stored in the storage device. The information processing device 10 also includes a training example storage unit 15, a candidate rule storage unit 16, and a model storage unit 17. The training example storage unit 15, the candidate rule storage unit 16, and the model storage unit 17 are each composed of a storage device. The functions and operations of each component will be described below.

[0012] The input unit 11 accepts input of a set of training case data (training data) that the information processing device 10 uses to train the rule set model, and stores it in the training case storage unit 15 (step S1 in FIG. 5). For example, the training case data consists of pairs of explanatory variables (x1, x2, . . . ) and objective variables (y).

[0013] The input unit 11 also receives input of a set of candidate rules (rules) that the information processing device 10 uses to make predictions on training case data, and stores the rules in the candidate rule storage unit 16 (step S1 in FIG. 5). For example, a candidate rule is composed of a condition (IF) and a predicted value (THEN), and as an example, the candidate rule is composed of a condition for an explanatory variable and a predicted value, such as "x3<5.0 AND x4>1.5 AND x6<2.0". Note that, in the following, only conditions are exemplified as candidate rules.

[0014] Furthermore, the input unit 11 receives input of the number of constraint rules, which is a parameter used by the information processing device 10 when generating a rule set model (step S1 in FIG. 5). For example, the number of constraint rules is the maximum number K of rules that can be combined and included in the rule set model. Note that the number of constraint rules may be a numerical value that lists the number of rules that can be combined and included in the rule set model, or may be any information that represents the number of rules that is allowed. In this embodiment, as an example, it is assumed that the number of constraint rules K=10 is input.

[0015] Furthermore, the input unit 11 accepts input of the number of constrained models, which is a parameter used by the information processing device 10 when selecting a model group consisting of a combination of rule set models (step S1 in FIG. 5). For example, the number of constrained models is the number S of models that can be combined and included in a model group. Note that the number of constrained models S may be a numerical value that lists the number of models that can be combined and included in a model group, or may be any information that represents the number of allowable models. In this embodiment, as an example, it is assumed that the number of constrained models S=4 is input.

[0016] The data received by the input unit 11, ie, the training case data and candidate rules, may be stored in the information processing device 10 in advance.

[0017] The model generation unit 12 (generation unit) generates a plurality of rule set models each consisting of a combination of candidate rules (step S2 in FIG. 5). At this time, the model generation unit 12 generates rule set models each combining candidate rules with a rule count r up to the maximum value K, which is the number of constraint rules, based on the prediction performance of the rule set model for training example data. In particular, in this embodiment, the model generation unit 12 generates one rule set model for each rule count r equal to or less than the maximum value K, which is the number of constraint rules. For example, when the maximum value K=10, which is the number of constraint rules, the model generation unit 12 generates rule set models including candidate rules with each rule count r ranging from 1 to 10. As an example, as shown in FIG. 3, a rule set model m2 with a rule count r=2 is generated, which is composed of two rules (rules 1 and 2), and a rule set model m3 with a rule count r=3 is generated, which is composed of three rules (rules 1, 2, and 3).

[0018] More specifically, a method for generating a rule set model by the model generation unit 12 will be described. The model generation unit 12 generates a rule set model m for each rule number r that is equal to or less than the maximum value K of the number of constraint rules. r When generating the rule set model m r That is, the model generation unit 12 generates a rule set model m r When generating the rule set model m, candidate rules are selected and added one by one using a greedy algorithm until the number of target rules reaches r. r For example, if the number of target rules r is 2, the greedy algorithm is used to first select the first candidate rule with good prediction performance, then select and add the second candidate rule with good prediction performance, and thus a rule set model m2 consisting of two candidate rules is set. In this way, for each of the target rule numbers r=1, 2, . . . , K, a rule set model m consisting of candidate rules for each rule number r is set. r Generate.

[0019] As described above, the model generation unit 12 generates each rule set model m corresponding to each rule number r by the greedy method. r By generating a rule set model m with a given approximation guarantee for the rule set model with the optimal predictive performance, r In other words, as described above, due to the submodularity in the optimization problem of the combination of candidate rules, it is possible to generate a rule set model with an approximation rate α=0.63 for the optimal predictive performance. Here, Figure 3 (3-1) shows a coordinate space with the number of rules on the horizontal axis and the predictive performance on the vertical axis, and on this coordinate space, each rule set model with the optimal predictive performance for each number of rules r is plotted as a white circle. Then, Figure 3 (3-2) further shows on the coordinate space each rule set model m generated by the greedy method as described above for each number of rules r. r In this way, the model generation unit 12 generates a rule set model m that has a predetermined approximation guarantee for each rule number r, with respect to the rule set model that has the optimal prediction performance. r However, the model generation unit 12 does not necessarily generate the rule set model m by the greedy method. r This is not limited to generating a rule set model m r At this time, the model generation unit 12 may generate a rule set model m with a predetermined approximation guarantee for each rule set model having the optimal prediction performance for each number of rules r. r It is recommended to generate

[0020] The group selection unit 13 (selection unit) selects rule set models m corresponding to the number r of rules generated as described above. r Among them, the number of rule set models m that is set as the number of constraint models S is r (Step S3 in FIG. 5). Specifically, the group selection unit 13 selects a model group that combines each rule set model m in a coordinate space with the number of rules and prediction performance as axes, as shown in FIG. 3 (3-2) above. rBased on the position of the points corresponding to the rule set model m consisting of the number of models S, r For example, when the number of models S, which is the number of constraint models, is 4, the combination of four rule set models m r A model group consisting of a combination of these is defined as each rule set model m r The number of constraint models S may represent a maximum value or a range of the number of models to be selected, and in this case, the group selection unit 13 may select a number of rule set models that is equal to or less than the maximum value or within the range.

[0021] More specifically, a method for selecting a model group by the group selection unit 13 will be described. The group selection unit 13 selects each rule set model m r The rule set model m of the set number S of models is selected so as to maximize the area of ​​the region A formed by each point of the set number S of models. r In this case, the area A to be maximized is selected from the rule set model m r and a fixed point P that is set to a value smaller than the point on the axis of prediction performance and larger on the axis of the number of rules. For example, the rule set model m r The area A is a region surrounded by sides parallel to each axis passing through the point of each rule set model and the fixed point P. For example, the fixed point P is set to the coordinate (K', 0) in the coordinate space where the prediction performance is 0 and the number of rules is a value K' that is larger than the maximum value K. Then, the group selection unit 13 adds points of the rule set models one by one by a greedy algorithm so that the area of ​​the region A formed by the sides parallel to each axis passing through the point of each rule set model and the fixed point P is maximized, and finally S rule set models m r That is, the group selection unit 13 selects a model group consisting of a rule set model m that maximizes the hypervolume exponential function corresponding to the area in the coordinate space as a hypervolume subset selection problem. r Select the point.

[0022] An example of the selection process of a group model by the group selection unit 13 will be described with reference to Fig. 4. Here, it is assumed that the number of constraint models S = 4. The group selection unit 13 first selects one rule set model m by a greedy algorithm, which maximizes the area of ​​the region A1 formed by the point of one rule set model and the fixed point P(K', 0), that is, the region A1 surrounded by sides parallel to each axis passing through the point of the rule set model and the fixed point P. r For example, as shown in gray in FIG. 4(4-1), a point of the rule set model m4 is selected based on the area of ​​a rectangular area A1 having opposite vertices of the point of the rule set model m4 and the fixed point P. Next, the group selection unit 13 selects the rule set model m4 with the largest area A by a greedy method. r Add and select points one by one to create four rule set models m r As a result, four rule set models m2, m4, m6, and m8 are selected based on the area of ​​the gray area A4 in Figure 4(4-2).

[0023] The group selection unit 13 selects a rule set model by the greedy method as described above, and selects a rule set model m that has a predetermined approximation guarantee for the area of ​​the region A that can be the largest. r In other words, as described above, due to the submodularity in the optimization problem of the combination of rule set models, it is possible to select a rule set model with an approximation rate β=0.63 for the maximum area. In addition to the approximation rate α when the rule set model is generated by the model generation unit 12, the selection of the rule set model by the group selection unit 13 has an approximation guarantee of αβ for the optimal selection. However, the group selection unit 13 does not necessarily select the rule set model m by the greedy method. r Not limited to selecting a rule set model m r may be selected.

[0024] The group selection unit 13 selects the rule set model m rThe group selection unit 13 outputs a model group consisting of a combination of the four rule set models m2, m4, m6, and m8 to the user (step S3 in FIG. 5). For example, in the above example, a model group consisting of four rule set models m2, m4, m6, and m8 is output to the user. This allows the user to use any of the rule set models included in the output model group in an actual operation case to make predictions in that case. Note that the group selection unit 13 selects the selected rule set model m r A model group consisting of the combination of the above is stored in the model storage unit 17.

[0025] The model search unit 14 (search unit) performs a solution-finding process on the training example data using each rule set model constituting the model group selected as described above as an initial solution, and searches for a new rule set model (step S4 in FIG. 5). Specifically, the model search unit 14 searches for a solution on the training example data for each rule set model, and if changing a candidate rule included in the rule set model improves prediction performance, changes the candidate rule and updates the rule set model.

[0026] Then, when the rule set model is updated, the group selection unit 13 may select again the rule set models that constitute the model group. When the rule set model is updated and a new rule set model is selected to update the model group, the group selection unit 13 outputs the model group.

[0027] As described above, the present disclosure can find an appropriate rule set model that has excellent predictive performance and interpretability and present it to the user. In other words, it is possible to find a rule set model that has good interpretability and good predictive performance because it contains a small number of rules. Furthermore, the present disclosure can also provide the advantage that, because multiple rule set models are selected using a greedy algorithm, the differences between each rule set model are easy to understand.

[0028] <Second embodiment> Next, a second embodiment of the present disclosure will be described with reference to the drawings. In this embodiment, an outline of the information processing device and the like described in the above-mentioned embodiment is shown. Note that the drawings may be relevant to any of the embodiments.

[0029] First, a description will be given of the hardware configuration of the information processing device 100 in the present disclosure. The information processing device 100 is configured as a general information processing device, and is equipped with the following hardware configuration, as an example, as shown in FIG. ·CPU(Central Processing Unit)101(Arithmetic unit) ROM (Read Only Memory) 102 (storage device) RAM (Random Access Memory) 103 (storage device) Programs 104 loaded into RAM 103 A storage device 105 for storing a group of programs 104 A drive device 106 that reads and writes from a storage medium 110 external to the information processing device A communication interface 107 that connects to a communication network 111 outside the information processing device Input / output interface 108 for inputting and outputting data Bus 109 connecting each component

[0030] 6 shows an example of the hardware configuration of the information processing device 100, and the hardware configuration of the information processing device is not limited to the above-described case. For example, the information processing device may be configured with a part of the above-described configuration, such as not including the drive device 106. Furthermore, the information processing device may use a GPU (Graphics Processing Unit), a DSP (Digital Signal Processor), an MPU (Micro Processing Unit), an FPU (Floating point number Processing Unit), a PPU (Physics Processing Unit), a TPU (Tensor Processing Unit), a quantum processor, a microcontroller, or a combination thereof, instead of the above-described CPU.

[0031] The information processing device 100 can be equipped with the generation unit 121 and the selection unit 122 shown in Fig. 7 by having the CPU 101 acquire and execute the program group 104. The program group 104 is stored in advance in the storage device 105 or the ROM 102, for example, and is loaded into the RAM 103 and executed by the CPU 101 as needed. The program group 104 may be supplied to the CPU 101 via the communication network 111, or may be stored in advance in the storage medium 110, and the drive device 106 may read out the programs and supply them to the CPU 101. However, the generation unit 121 and the selection unit 122 described above may be constructed using dedicated electronic circuits for realizing such means.

[0032] The generation unit 121 generates a plurality of rule set models that satisfy the number of constraint rules that represent constraints on the number of rules that can be combined, based on prediction performance for the training data by the rule set models each consisting of a combination of rules that makes a predetermined prediction for the training data. The selection unit 122 selects and outputs a model group made up of a combination of the rule set models that satisfies the number of constraint models that represent constraints on the number of models that can be combined, based on the corresponding position of the rule set model in a space whose axes are prediction performance and the number of rules.

[0033] With the present disclosure configured as described above, it is possible to easily find an appropriate rule set model.

[0034] At least one of the functions of the generation unit 121 and the selection unit 122 described above may be executed by an information processing device installed and connected anywhere on the network, that is, may be executed by so-called cloud computing.

[0035] The above-described program can be stored and supplied to a computer using various types of non-transitory computer-readable media. Non-transitory computer-readable media include various types of tangible storage media. Examples of non-transitory computer-readable media include magnetic recording media (e.g., flexible disks, magnetic tapes, hard disk drives), magneto-optical recording media (e.g., magneto-optical disks), CD-ROMs (Read Only Memory), CD-Rs, CD-R / Ws, and semiconductor memories (e.g., mask ROMs, PROMs (Programmable ROMs), EPROMs (Erasable PROMs), flash ROMs, and RAMs (Random Access Memory)). The program can also be supplied to a computer by various types of transitory computer-readable media. Examples of transitory computer-readable media include electrical signals, optical signals, and electromagnetic waves. The transitory computer-readable media can supply the program to a computer via a wired communication path such as an electric wire or optical fiber, or via a wireless communication path.

[0036] Although the present disclosure has been described above with reference to the embodiments, the present disclosure is not limited to the above-described embodiments. Various modifications that can be understood by those skilled in the art can be made to the configuration and details of the present disclosure within the scope of the present disclosure. Furthermore, each of the above-described embodiments can be combined with other embodiments as appropriate.

[0037] <Additional Notes> Some or all of the above embodiments may be described as follows: The following provides an overview of the configurations of an information processing device, an information processing method, and a program according to the present disclosure. However, the present disclosure is not limited to the configurations described in the following supplementary notes. Note that the configurations described in Supplements 2 to 8, which are dependent on Supplementary Note 1 below, and some or all of the functions of the configurations, may also be dependent on other Supplements 9 and 17 in the same dependent relationship as Supplements 2 to 8. Furthermore, not limited to Supplements 1, 9, and 17, but also within the scope of the above-described embodiments, similar hardware, software, various recording means for recording software, or systems may similarly be made to be dependent on the configurations described as Supplements and some or all of the functions of the configurations. (Appendix 1) a generation unit that generates a plurality of rule set models that satisfy a number of constraint rules representing a constraint on the number of rules that can be combined, based on prediction performance for training data by the rule set models, each of which is formed by a combination of rules that make a predetermined prediction for the training data; a selection unit that selects and outputs a model group made up of combinations of the rule set models that satisfy a constraint model number that represents a constraint on the number of models that can be combined, based on a position corresponding to the rule set model in a space whose axes are prediction performance and the number of rules; An information processing device comprising: (Appendix 2) 10. The information processing device according to claim 1, the generation unit generates the rule set model whose prediction performance for the training data is determined to be high according to a preset criterion for each rule number that satisfies the number of constraint rules. Information processing device. (Appendix 3) 10. The information processing device according to claim 2, the generation unit generates the rule set model for each number of rules ranging from 1 to a maximum value set as the number of constraint rules. Information processing device. (Appendix 4) 10. The information processing device according to claim 1, the selection unit selects and outputs the model group based on the size of an area formed by points representing corresponding positions of the rule set models constituting a combination of the rule set models, the number of which is equal to the number of models that satisfies the number of constraint models, in the space. Information processing device. (Appendix 5) 5. The information processing device according to claim 4, the selection unit selects and outputs the model group in which an area formed by the points in the space corresponding to the rule set models constituting a combination of the rule set models of the number of models that satisfies the number of constraint rules, and a fixed point set to a value smaller than the points on the axis of prediction performance and larger on the axis of the number of rules, is determined to be larger according to a preset criterion. Information processing device. (Appendix 6) 10. The information processing device according to claim 1, the generator is configured to generate the rule set model having a predetermined approximation guarantee for the predictive performance of the optimal rule set model. Information processing device. (Appendix 7) 10. The information processing device according to claim 1, the generation unit is configured to select each rule by a greedy method based on the predictive performance of the rule and generate the rule set model including the selected rule. Information processing device. (Appendix 8) 10. The information processing device according to claim 1, a search unit that uses the rule set model that constitutes the selected model group as an initial solution, solves the training data, and searches for a new rule set model; Information processing device. (Appendix 9) generating a plurality of rule set models that satisfy a number of constraint rules representing constraints on the number of rules that can be combined, based on prediction performance for the training data by the rule set models each consisting of a combination of rules that make a predetermined prediction for the training data; selecting and outputting a model group consisting of a combination of the rule set models that satisfies a constraint model number that represents a constraint on the number of models that can be combined, based on a position corresponding to the rule set model in a space whose axes are prediction performance and the number of rules; Information processing methods. (Appendix 10) 10. The information processing method according to claim 9, generating the rule set model whose prediction performance for the training data is determined to be high according to a preset criterion for each rule whose number of constraint rules satisfies the number of constraint rules; Information processing methods. (Appendix 11) 11. The information processing method according to claim 10, further comprising: generate the rule set model for each number of rules ranging from 1 to the maximum value set as the number of constraint rules; Information processing methods. (Appendix 12) 10. The information processing method according to claim 9, selecting and outputting the model group based on the size of an area formed by points representing corresponding positions of the rule set models constituting a combination of the rule set models in a number of models that satisfies the number of constraint models in the space; Information processing methods. (Appendix 13) 13. The information processing method according to claim 12, further comprising: select and output the model group in which an area formed by the points in the space corresponding to the rule set models constituting the combination of the rule set models of the number of models that satisfies the number of constraint rules, and a fixed point set to a value smaller than the points on the axis of prediction performance and larger on the axis of the number of rules, is determined to be larger according to a preset criterion; Information processing methods. (Appendix 14) 10. The information processing method according to claim 9, and generating the rule set model with predetermined approximation guarantees for predictive performance of the optimal rule set model. Information processing methods. (Appendix 15) 10. The information processing method according to claim 9, and selecting each rule using a greedy method based on the predictive performance of the rule to generate the rule set model including the rule. Information processing methods. (Appendix 16) 10. The information processing method according to claim 9, using the rule set model constituting the selected model group as an initial solution, solve the training data, and search for a new rule set model. Information processing methods. (Appendix 17) generating a plurality of rule set models that satisfy a number of constraint rules representing constraints on the number of rules that can be combined, based on prediction performance for the training data by the rule set models each consisting of a combination of rules that make a predetermined prediction for the training data; selecting and outputting a model group consisting of a combination of the rule set models that satisfies a constraint model number that represents a constraint on the number of models that can be combined, based on a position corresponding to the rule set model in a space whose axes are prediction performance and the number of rules; A program that causes a computer to perform a process. [Explanation of symbols]

[0038] 10. Information processing equipment 11 Input section 12 Model Generation Unit 13 Group selection section 14 Model Search Unit 15 Training case storage section 16 Candidate rule memory 17 Model memory section 100 Information processing device 101 CPU 102 ROM 103 RAM 104 Programs 105 Storage device 106 Drive device 107 Communication Interface 108 Input / Output Interface 109 Bus 110 Storage medium 111 Communication Network 121 Generation part 122 Selection section

Claims

1. a generation unit that generates a plurality of rule set models that satisfy a number of constraint rules representing a constraint on the number of rules that can be combined, based on prediction performance for training data by the rule set models, each of which is formed by a combination of rules that make a predetermined prediction for the training data; a selection unit that selects and outputs a model group made up of combinations of the rule set models that satisfy a constraint model number that represents a constraint on the number of models that can be combined, based on a position corresponding to the rule set model in a space whose axes are prediction performance and the number of rules; An information processing device comprising:

2. 2. The information processing device according to claim 1, the generation unit generates the rule set model whose prediction performance for the training data is determined to be high according to a preset criterion for each rule number that satisfies the number of constraint rules. Information processing device.

3. 3. The information processing device according to claim 2, the generation unit generates the rule set model for each number of rules ranging from 1 to a maximum value set as the number of constraint rules. Information processing device.

4. 2. The information processing device according to claim 1, the selection unit selects and outputs the model group based on the size of an area formed by points representing corresponding positions of the rule set models constituting a combination of the rule set models, the number of which is equal to the number of models that satisfies the number of constraint models, in the space. Information processing device.

5. 5. The information processing device according to claim 4, the selection unit selects and outputs the model group in which an area formed by the points in the space corresponding to the rule set models constituting a combination of the rule set models of the number of models that satisfies the number of constraint rules, and a fixed point set to a value smaller than the points on the axis of prediction performance and larger on the axis of the number of rules, is determined to be larger according to a preset criterion. Information processing device.

6. 2. The information processing device according to claim 1, the generator is configured to generate the rule set model having a predetermined approximation guarantee for the predictive performance of the optimal rule set model. Information processing device.

7. 2. The information processing device according to claim 1, the generation unit is configured to select each rule by a greedy method based on the predictive performance of the rule and generate the rule set model including the selected rule. Information processing device.

8. 2. The information processing device according to claim 1, a search unit that uses the rule set model that constitutes the selected model group as an initial solution, solves the training data, and searches for a new rule set model; Information processing device.

9. generating a plurality of rule set models that satisfy a number of constraint rules representing constraints on the number of rules that can be combined, based on prediction performance for the training data by the rule set models each consisting of a combination of rules that make a predetermined prediction for the training data; selecting and outputting a model group consisting of a combination of the rule set models that satisfies a constraint model number that represents a constraint on the number of models that can be combined, based on a position corresponding to the rule set model in a space whose axes are prediction performance and the number of rules; Information processing methods.

10. generating a plurality of rule set models that satisfy a number of constraint rules representing constraints on the number of rules that can be combined, based on prediction performance for the training data by the rule set models each consisting of a combination of rules that make a predetermined prediction for the training data; selecting and outputting a model group consisting of a combination of the rule set models that satisfies a constraint model number that represents a constraint on the number of models that can be combined, based on a position corresponding to the rule set model in a space whose axes are prediction performance and the number of rules; A program that causes a computer to perform a process.

Images