Information processing apparatus, information processing method, and program

Abstract

To solve the problem that it is difficult to improve the accuracy of prediction using a machine learning model.SOLUTION: An information processing apparatus according to the present disclosure includes an acquisition unit that acquires first data including a set of an explanatory variable and an objective variable classified into a plurality of classifications based on a correspondence relationship between the explanatory variable and the objective variable, and a first model generated to predict the objective variable from the explanatory variable based on the first data for each classification, and a generation unit that generates a second model that predicts the first model corresponding to the explanatory variable based on information representing the correspondence relationship between the explanatory variable of the first data based on the classification and each first model.SELECTED DRAWING: Figure 10

Description

[Technical Field]

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

[0002] Predictions based on input data are made using machine learning models in a variety of fields. However, the prediction accuracy of machine learning models can decline due to changes in data characteristics over time (concept drift). In such cases, the machine learning model can be retrained, as described in Patent Document 1. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 2023-170458 Summary of the Invention [Problem to be solved by the invention]

[0004] However, when a machine learning model is retrained, it is not possible to utilize past data or models, which results in the problem that the machine learning model cannot be applied to all situations, making it difficult to improve prediction accuracy.

[0005] Therefore, one of the objectives of the present disclosure is to solve the above-mentioned problem that it is difficult to improve the accuracy of predictions using machine learning models. [Means for solving the problem]

[0006] An information processing device according to an embodiment of the present disclosure includes: an acquisition unit that acquires first data consisting of pairs of explanatory variables and dependent variables classified into a plurality of categories based on correspondence relationships between the explanatory variables and dependent variables, and a first model that is generated for each category based on the first data so as to predict the dependent variables from the explanatory variables; a generation unit that generates a second model that predicts the first model corresponding to the explanatory variables based on information representing a correspondence relationship between the explanatory variables of the first data based on the classification and each of the first models; Equipped with The structure is as follows. Furthermore, an information processing method according to an embodiment of the present disclosure includes: acquiring first data consisting of pairs of explanatory variables and dependent variables classified into a plurality of categories based on correspondence relationships between the explanatory variables and dependent variables, and a first model generated for each category based on the first data so as to predict the dependent variables from the explanatory variables; generating a second model that predicts the first model corresponding to the explanatory variables based on information representing a correspondence relationship between the explanatory variables of the first data based on the classification and each of the first models; The structure is as follows. Furthermore, a program according to an embodiment of the present disclosure includes: acquiring first data consisting of pairs of explanatory variables and dependent variables classified into a plurality of categories based on correspondence relationships between the explanatory variables and dependent variables, and a first model generated for each category based on the first data so as to predict the dependent variables from the explanatory variables; generating a second model that predicts the first model corresponding to the explanatory variables based on information representing a correspondence relationship between the explanatory variables of the first data based on the classification and each of the first models; Have the computer perform the process, The structure is as follows. [Effects of the Invention]

[0007] By being configured as described above, the present disclosure can improve the accuracy of predictions using machine learning models. [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] FIG. 1 is a diagram illustrating an example of a process performed by an information processing device according to the present disclosure. [Figure 6] 10 is a flowchart illustrating an example of a processing operation of an information processing device according to the present disclosure. [Figure 7] 1 is a block diagram illustrating an example of a configuration and processing state of an information processing system according to the present disclosure. [Figure 8] FIG. 1 is a diagram illustrating an example of processing by an information processing system according to the present disclosure. [Figure 9] FIG. 1 is a block diagram illustrating an example of a hardware configuration of an information processing device according to the present disclosure. [Figure 10] 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 creates a prediction model that makes predictions by performing machine learning, and uses this prediction model to make predictions on input data. In particular, in this embodiment, the input data is classified and a prediction model is created for each classification, and a gate model that predicts a prediction model appropriate for the input data from among these multiple prediction models is also created. This allows predictions to be made using a prediction model appropriate for the input data, thereby improving the accuracy of predictions.

[0011] Here, an example of a prediction target using a prediction model is the occurrence and probability of a seizure, a patient's illness. In this case, explanatory variables, which are input data, include biological information such as the patient's body temperature and heart rate, and environmental information such as time, temperature, and weather. By inputting these into the prediction model, the occurrence and probability of a seizure are predicted as predicted values. In this case, as described below, multiple prediction models are created for each characteristic of training data, which is patient data collected in advance, and a gate model is also created that predicts a prediction model appropriate for the patient's condition. By inputting the patient's condition into the gate model, the occurrence and probability of a seizure can be predicted using a prediction model appropriate for the patient's condition, thereby improving the accuracy of the prediction. However, the prediction target using a prediction model in the present disclosure is not limited to the above and may be any content.

[0012] The configuration and operation of an information processing device 10 in this embodiment will be described below. The information processing device 10 is configured with 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 a data decomposition unit 11, a prediction model creation unit 12, a gate model learning unit 13, and a prediction unit 14. The functions of the data decomposition unit 11, the prediction model creation unit 12, the gate model learning unit 13, and the prediction 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 data storage unit 15 and a model storage unit 16. The data storage unit 15 and the model storage unit 16 are configured with a storage device.

[0013] The data decomposition unit 11 (acquisition unit, classification unit) first receives input of first data D1, which will serve as learning data, and stores it in the data storage unit 15 (step S1 in FIG. 6). The first data D1 is a data group containing a plurality of pairs of data, each of which is a pair of an explanatory variable and a response variable. The first data D1 is, for example, data acquired from actual cases or generated from simulations or probabilistic models. In the example of predicting the occurrence of a patient's seizure described above, the first data D1 is a group of data containing a large number of pairs of data, each of which is an explanatory variable consisting of biometric data and environmental information measured in the past, and a response variable consisting of the occurrence or probability of a seizure within a predetermined period of time thereafter. For example, the first data D1 can be expressed as in the following equation 1, where the explanatory variable is x and the response variable is y:

number

[0014] Then, the data decomposition unit 11 decomposes the first data D1 into each sub-data B based on the correspondence between the explanatory variable x and the response variable y (step S2 in FIG. 6). That is, the data decomposition unit 11 classifies the first data D1 into a plurality of sub-data categories by including paired data having a common characteristic in the same sub-data B according to the characteristics of the paired data of the explanatory variable x and the response variable y. As an example, the data decomposition unit 11 classifies the first data D1 into sub-data B1, B2, ..., B k , that is, as shown in the following formula 2, K sub-data B k Decompose into:

number

[0015] Specifically, the data decomposition unit 11 uses a clustering method such as the K-means method or the shortest distance method to decompose the first data D1 into a plurality of sub-data B having different correspondences between the explanatory variable x and the response variable y. For example, the data decomposition unit 11 decomposes the first data D1 into three sub-data B1, B2, and B3 as shown in Fig. 2, or decomposes the first data D1 as shown in Fig. 4(4-1) into two sub-data B1 and B2 as shown in Fig. 4(4-2). Note that in Fig. 4, one circle is shown as corresponding to one set of data consisting of a set of the explanatory variable x and the response variable y.

[0016] When the data decomposition unit 11 decomposes the first data D1 into sub-data B, it may decompose one group of data consisting of a pair of an explanatory variable x and a target variable y into multiple sub-data Bs. In this case, the data decomposition unit 11 may assign a weight corresponding to each sub-data B to one group of data and include it in each sub-data B. For example, as shown in FIG. 4(4-2), when the data is divided into two sub-data B1 and B2, a weight of 0.7 may be assigned to sub-data B1 and a weight of 0.3 may be assigned to sub-data B2 for a specific group of data, and the specific group of data may be decomposed to be included in both sub-data B1 and B2. In this case, the sub-data B can be expressed as in Equation 3 using a weight w. Note that one group of data does not need to be decomposed to be included in at least one sub-data B, and may not be included in any sub-data B.

number

[0017] The data decomposition unit 11 may repeatedly perform the decomposition process of the first data D1 together with the creation of the prediction model h and the learning of the gate model g, as will be described later.

[0018] The prediction model creation unit 12 (acquisition unit, first model generation unit) creates a prediction model h (first model) for each sub-data B obtained by decomposing the first data D1 as described above, using the paired data included in each sub-data B to predict the dependent variable y from the explanatory variable x, and stores the created prediction model in the model storage unit 16 (step S3 in FIG. 6 ). Specifically, for each sub-data B, the prediction model creation unit 12 uses the paired data of the explanatory variable x and the dependent variable y included in the sub-data B as training data, and performs machine learning on the prediction model h to minimize the error between a predicted value when the explanatory variable x is input to the prediction model h and the dependent variable y paired with the explanatory variable x. For example, a decision tree, a neural network, a gradient boosting model, or the like may be used as the prediction model h. As a result, each prediction model h is configured to output a predicted value that may be the dependent variable y for the explanatory variable x when a new explanatory variable x such as the second data D2 described below is input.

[0019] For example, when the first data D1 is decomposed into K sub-data B as described above, the prediction model creation unit 12 creates K prediction models h1, . . . , h2 corresponding to the K sub-data B. K Create a predictive model h k The predicted value by h k As an example, the prediction model creation unit 12 creates three prediction models h1, h2, and h3 from three sub-data B1, B2, and B3 of the first data D1, respectively, as shown in Fig. 2, or creates two prediction models h1 and h2 from the sub-data B1 and B2, respectively, as shown in Fig. 5 (5-1).

[0020] In addition, when the group data of the first data D1 is weighted and decomposed into sub-data B as described above, the prediction model creation unit 12 performs machine learning to minimize the loss function shown in the following formula 4, thereby obtaining the k-th sub-data B k The corresponding prediction model h k can be created.

number

[0021] In the above description, the first data D1 is decomposed into sub-data B, and a prediction model h corresponding to each sub-data B is created. However, the first data D1, each sub-data B, and each prediction model h may be prepared in advance and stored in the data storage unit 15 or the model storage unit 16. In other words, the information processing device 10 described above is not necessarily limited to including the data decomposition unit 11 and the prediction model creation unit 12. It may also acquire sub-data B obtained by decomposing the first data D1 and a prediction model h that has already been created, and use the sub-data B and the prediction model to generate a gate model g and make predictions, as will be described later.

[0022] The gate model learning unit 13 (generation unit) uses information representing the correspondence between explanatory variables x of the first data D1 decomposed into a plurality of sub-data B and the prediction models h corresponding to the explanatory variables x to generate a gate model g (second model) by machine learning, which outputs a model prediction value that predicts the prediction model h corresponding to the explanatory variables x in response to an input of the explanatory variables x, and stores the generated gate model g in the model storage unit 16 (step S4 in FIG. 6). Specifically, the gate model learning unit 13 first uses each sub-data B obtained by decomposing the first data D1 to set weight information representing the degree to which each explanatory variable x included in each sub-data B corresponds to each prediction model h. At this time, the gate model learning unit 13 focuses only on the explanatory variables x included in the sub-data B, and sets weight information representing the degree to which the explanatory variable x corresponds to each prediction model h depending on which prediction model h the explanatory variable x can be applied to, that is, which sub-data B (cluster) the explanatory variable x can belong to.

[0023] A specific example of setting weight information will be described with reference to FIG. 5 (5-2). In FIG. 5 (5-2), the explanatory variable x in the range of x1 is included only in sub-data B1, and therefore can correspond to the prediction model h1 created from the sub-data B1. Therefore, the weight information for the explanatory variable x in the range of x1 is set to 1.0 for the prediction model h1 and 0.0 for the prediction model h2, and is expressed as [1.0, 0.0]. Also, in FIG. 5 (5-2), the explanatory variable x in the range of x2 is included in both sub-data B1 and B2, and therefore can correspond to both the prediction models h1 and h2 created from the sub-data B1 and B2, respectively. Therefore, the weight information for the explanatory variable x in the range of x2 is set to 0.5 for the prediction model h1 and 0.5 for the prediction model h2, and is expressed as [0.5, 0.5]. 5(5-2), the explanatory variable x in the range of x3 is included only in sub-data B2, and therefore may correspond to the prediction model h2 created from the sub-data B2. Therefore, the weight information for the explanatory variable x in the range of x3 is set to 0.0 for the prediction model h1 and 1.0 for the prediction model h2, and is expressed as [0.0, 1.0]. Note that the above-mentioned example of setting the weight information is just an example, and the weight information may be set depending on the distribution of the explanatory variable x for each sub-data B, such as [0.3, 0.7].

[0024] In this way, the weight information w is K weights w corresponding to K prediction models h corresponding to K subdata B. k , and the values ​​of [w1, ,w K In this example, w k ,≧0, w1+···+w K = 1, but any value can be set as the weight. Also, if there is only one prediction model h corresponding to each explanatory variable x, k The weight w corresponding to k is set to 1, and the rest to 0.

[0025] Then, the gate model learning unit 13 uses the explanatory variable x of the first data D1 and the weight information w set corresponding to the explanatory variable x as learning data to machine-learn a gate model g that outputs weight information w corresponding to the input explanatory variable x. That is, by machine-learning learning data in which the explanatory variable x of the first data D1 is used as an explanatory variable and the weight information w set corresponding to the explanatory variable x is used as a target variable, a model predicted value g(x)=[w1, , w k In the example of Figure 2, a gate model g is generated that predicts weight information w, which is the degree of correspondence of explanatory variable x to each of three prediction models h1, h2, and h3.

[0026] Here, when setting weight information w for the explanatory variable x of the first data D1, the gate model learning unit 13 may set the weight information w so that the smaller the prediction error of a prediction model h for the group data of the first data D1, the greater the degree of correspondence to the prediction model h, i.e., the greater the weight value. For example, the gate model learning unit 13 may set the weight of the prediction model h with the smallest prediction error in a certain group data to 1.0 and the weights of the other prediction models to 0.0. By performing machine learning as described above using the weight information w set in this manner, the gate model g is trained so that the weight of the prediction model h with the smallest prediction error is output as a large value. More specifically, the gate model g may be trained using K subdata B to minimize the following equation (5).

number

[0027] In addition, when weights are assigned to the group data of the first data D1 and the group data is decomposed into sub-data B as described above, the gate model learning unit 13 may set the above-mentioned weight information w taking into consideration the weights assigned to the group data, and may learn the gate model g using the weight information w. For example, as in the example of FIG. 5(5-2) described above, when the weight information w for two prediction models h for the explanatory variable x of the first data D1 in the range of x2 is set to [0.5, 0.5], if weights of 0.7 and 0.3 are assigned to the first data D1 as weights for the two sub-models B, the weight information w may be set to [0.6, 0.4] taking into consideration the weights.

[0028] The prediction unit 14 acquires second data consisting only of the new explanatory variable x to be predicted (step S5 in FIG. 6), and performs prediction from the second data using the gate model g and prediction model h generated as described above (step S6 in FIG. 6). Specifically, the prediction unit 14 first acquires second data D2 consisting only of the new explanatory variable x, as shown in the following formula 6.

number

[0029] For example, the prediction unit 14 first inputs the second data D2 to all prediction models h, respectively, and obtains a predicted value that is an output from each prediction model h. Then, the prediction unit 14 multiplies the predicted value from each prediction model h by the weight of the model predicted value corresponding to each prediction model h, and calculates the sum of these as the final predicted value. Specifically, the predicted value y j Calculate.

number

[0030] An example of calculation of a predicted value by the prediction unit 14 will be described with reference to FIG. 3. In this example, as shown in FIG. 2, it is assumed that first data D1 is decomposed into three sub-data B1, B2, and B3, and three prediction models h1, h2, and h3 are created. First, the prediction unit 14 inputs a new explanatory variable x, which is second data D2, into a gate model g, and obtains weight information w, which indicates the degree to which the explanatory variable x corresponds to each prediction model h1, h2, and h3, as the output of the gate model g. In this example, g(x) = [0.7, 0.1, 0.2] can be obtained, and the explanatory variable x, which is second data D2, corresponds to prediction model h1 with a weight of 0.7, prediction model h2 with a weight of 0.1, and prediction model h3 with a weight of 0.2. In other words, the explanatory variable x, which is second data D2, is most suitable for prediction models h1, h3, and h2 in that order. Next, the prediction unit 14 inputs the new explanatory variable x, which is the second data D2, into each of the prediction models h1, h2, and h3, and obtains each predicted value, which is the output from each of the prediction models h1, h2, and h3. In this example, the prediction unit 14 obtains a predicted value of 5.0 from the prediction model h1, a predicted value of 1.0 from the prediction model h2, and a predicted value of 4.0 from the prediction model h3. The prediction unit 14 then multiplies each predicted value from each of the prediction models h1, h2, and h3 by the weight of each prediction model h1, h2, and h3, and calculates their sum. In this example, the calculation is "5.0 x 0.7 + 1.0 x 0.1 + 4.0 x 0.2 = 4.4," which is the final predicted value.

[0031] As described above, the information processing device 10 of the present disclosure creates prediction models h corresponding to each sub-data B obtained by breaking down first data D1 according to its characteristics, and generates a gate model g by machine learning that predicts the prediction model h from the explanatory variable x based on the degree of correspondence between the explanatory variable x of the first data D1 and the prediction model h. This makes it possible to predict a prediction model h corresponding to a new explanatory variable x, and by predicting a predicted value for the new explanatory variable x using a prediction model h with a high degree of correspondence, highly accurate predictions can be made. Furthermore, by having the gate model g predict the degree of correspondence between the explanatory variable x and each prediction model h using a weight, the predicted values ​​from each prediction model h can be aggregated according to the weight to obtain a final predicted value, thereby further improving prediction accuracy.

[0032] (Variation) Here, a modified example of the above-mentioned information processing device 10 will be described. As shown in Fig. 1, the above-mentioned gate model learning unit 13 may receive input of third data D3 consisting of a pair of explanatory variable x and objective variable y, and learn a gate model g using the third data D3. In this case, the third data D3 is obtained from an actual example or generated from a simulation or a probability model, and is expressed by, for example, Equation 8.

number

[0033] Specifically, when the gate model learning unit 13 acquires the third data D3, it first causes the prediction unit 14 to make a prediction. The prediction unit 14 inputs the explanatory variable x of the third data D3 into the gate model g and each prediction model h as described above, and predicts a predicted value for the explanatory variable x of the third data D3. The gate model learning unit 13 then performs machine learning to update the gate model g so as to reduce the error between this predicted value and the objective variable y paired with the explanatory variable x in the third data D3. For example, the gate model g is learned so as to minimize the loss function shown in the following equation (9).

number

[0034] As another variation, the data decomposition unit 11 may decompose the first data D1 using the prediction results from the prediction unit 14. Specifically, the prediction unit 14 first inputs the explanatory variable x of the first data D1 into the gate model g and each prediction model h as described above, and predicts a predicted value for the explanatory variable x of the first data D1. The data decomposition unit 11 then decomposes the first data D1 again into sub-data B based on the error between the predicted value and the dependent variable y paired with the explanatory variable x in the first data D1, and similarly to the above, creates a prediction model h for each sub-data B and repeatedly trains the gate model g. At this time, the data decomposition unit 11 may, for example, decompose the first data D1 so as to minimize the error between the predicted value and the dependent variable, or may decompose the first data D1 by allocating it to the sub-data corresponding to the prediction model h with the smallest error in the predicted value of each prediction model h.

[0035] (Usage example) Next, a usage example of the present disclosure will be described. Here, as described above, a case where the probability of occurrence of a seizure, which is a patient's illness, is predicted will be described as an example. First, as shown in FIG. 7 , a patient U measures vital data such as body temperature and heart rate using a wearable device or a measuring device worn by the patient, and inputs the vital data as input data (explanatory variable x) to the information processing device 10 via an information processing terminal. At this time, in addition to the vital data, environmental information such as temperature and weather may also be input as input data (explanatory variable x) to the information processing device 10. Then, the information processing device 10 predicts the probability of occurrence of a seizure in the patient using the prediction model h and gate model g generated as described above. Furthermore, the information processing device 10 outputs the predicted occurrence probability to the patient U. For example, a screen showing a graph representing the predicted occurrence probability for each date, as shown in FIG. 8, is output to be displayed on the patient U's information processing terminal. Note that the above-described usage example of the information processing device 10 is just an example, and the information processing device 10 may be used for any prediction.

[0036] <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.

[0037] 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

[0038] 9 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.

[0039] The information processing device 100 can be equipped with an acquisition unit 121 and a generation unit 122 shown in Fig. 10 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 acquisition unit 121 and the generation unit 122 described above may be constructed using dedicated electronic circuits for realizing such means.

[0040] The acquiring unit 121 acquires first data consisting of pairs of explanatory variables and dependent variables classified into a plurality of categories based on the correspondence between the explanatory variables and dependent variables, and first models generated for each category based on the first data to predict the dependent variables from the explanatory variables. The generating unit 122 generates second models that predict the first models corresponding to the explanatory variables based on information representing the correspondence between the explanatory variables of the first data based on the categories and each of the first models.

[0041] With the above configuration, the present disclosure can predict a first model corresponding to a new explanatory variable using a second model, and predict a predicted value for the new explanatory variable using the predicted first model, thereby improving the accuracy of predictions using machine learning models.

[0042] In addition, at least one of the functions of the above-mentioned acquisition unit 121 and generation unit 122 may be executed by an information processing device installed and connected anywhere on the network, that is, they may be executed by so-called cloud computing.

[0043] 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.

[0044] 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.

[0045] <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 8.2, 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 10 in the same dependent relationship as Supplements 2 to 8.2. Furthermore, not limited to Supplements 1, 9, and 10, but also within the scope of the above-described embodiments, similar hardware, software, various recording means for recording software, or systems may be similarly made to be dependent on the configurations described as Supplements and some or all of the functions of the configurations. (Appendix 1) an acquisition unit that acquires first data consisting of pairs of explanatory variables and dependent variables classified into a plurality of categories based on correspondence relationships between the explanatory variables and dependent variables, and a first model that is generated for each category based on the first data so as to predict the dependent variables from the explanatory variables; a generation unit that generates a second model that predicts the first model corresponding to the explanatory variables based on information representing a correspondence relationship between the explanatory variables of the first data based on the classification and each of the first models; An information processing device comprising: (Appendix 2) 10. The information processing device according to claim 1, a prediction unit that predicts a predicted value from the second explanatory variable by inputting a second explanatory variable into the second model and using the first model corresponding to the second explanatory variable predicted by the second model, Information processing device. (Appendix 3) 10. The information processing device according to claim 1, the generation unit generates the second model that predicts the first model corresponding to the explanatory variables based on weight information that indicates a degree to which the explanatory variables of the first data based on the classification correspond to each of the first models. Information processing device. (Appendix 4) 4. The information processing device according to claim 3, the generation unit generates the second model that predicts the degree to which the explanatory variables correspond to each of the first models. Information processing device. (Appendix 5) 4. The information processing device according to claim 3, the generation unit sets the weight information so that the explanatory variables correspond to a higher degree to the first model as the first model has a smaller prediction error in the first data, and generates the second model. Information processing device. (Appendix 6) 5. The information processing device according to claim 4, a prediction unit that inputs a second explanatory variable into the second model, predicts a degree to which the second explanatory variable corresponds to each of the first models, and predicts a predicted value from the second explanatory variable using each of the first models according to the predicted degree; Information processing device. (Appendix 7) 10. The information processing device according to claim 1, a prediction unit that predicts a predicted value from a third explanatory variable by inputting a third explanatory variable into the second model and using the first model corresponding to the third explanatory variable predicted by the second model, the generation unit generates the second model so as to reduce an error between a predicted value predicted from the third explanatory variable and a third dependent variable that is paired in advance with the third explanatory variable. Information processing device. (Appendix 8) 10. The information processing device according to claim 1, a classification unit that classifies the first data into a plurality of categories based on a correspondence between the explanatory variables and the objective variables; a first model generation unit that generates the first model for predicting the dependent variable from the explanatory variables based on the first data for each of the classifications; a prediction unit that predicts a predicted value from the explanatory variables by inputting the explanatory variables of the first data into the second model and using the first model corresponding to the predicted explanatory variables, the classifying unit classifies the first data into the plurality of categories based on an error between a predicted value predicted from the explanatory variables of the first data and the dependent variable corresponding to the explanatory variables; Information processing device. (Appendix 8.1) 10. The information processing device according to claim 1, a classification unit that classifies the first data into a plurality of categories based on a correspondence relationship between the explanatory variables and the objective variable, and assigns weights to the first data based on the categories; a first model generation unit that generates the first model for predicting the dependent variable from the explanatory variables based on the first data and weights assigned to the first data for each of the classifications; Equipped with Information processing device. (Appendix 8.2) 10. The information processing device according to claim 1, a classification unit that classifies the first data into a plurality of categories based on a correspondence relationship between the explanatory variables and the objective variable, and assigns weights to the first data based on the categories; the generation unit generates the second model based on information representing a correspondence between the explanatory variables of the first data based on the classification and each of the first models, and a weight assigned to the first data. Information processing device. (Appendix 9) acquiring first data consisting of pairs of explanatory variables and dependent variables classified into a plurality of categories based on correspondence relationships between the explanatory variables and dependent variables, and a first model generated for each category based on the first data so as to predict the dependent variables from the explanatory variables; generating a second model that predicts the first model corresponding to the explanatory variables based on information representing a correspondence relationship between the explanatory variables of the first data based on the classification and each of the first models; Information processing methods. (Appendix 9.1) 10. The information processing method according to claim 9, a prediction value from the second explanatory variable is predicted using the first model corresponding to the second explanatory variable predicted by inputting a second explanatory variable into the second model; Information processing methods. (Appendix 9.2) 10. The information processing method according to claim 9, generating the second model that predicts the first model corresponding to the explanatory variables based on weight information that indicates the degree to which the explanatory variables of the first data based on the classification correspond to each of the first models; Information processing methods. (Appendix 9.3) 9.2, a method for processing information according to claim 9.2, generating the second model that predicts the degree to which the explanatory variables correspond to each of the first models; Information processing methods. (Appendix 9.4) 9.2, a method for processing information according to claim 9.2, generating the second model by setting the weight information such that the explanatory variables correspond more to the first model as the prediction error in the first data becomes smaller; Information processing methods. (Appendix 9.5) 9.3. An information processing method according to claim 9.3, comprising: inputting a second explanatory variable into the second model to predict the degree to which the second explanatory variable corresponds to each of the first models, and predicting a predicted value from the second explanatory variable using each of the first models according to the predicted degree; Information processing methods. (Appendix 9.6) 10. The information processing method according to claim 9, predicting a predicted value from the third explanatory variable using the first model corresponding to the predicted third explanatory variable as an input to the second model; generating the second model so as to reduce an error between a predicted value predicted from the third explanatory variable and a third dependent variable that is paired in advance with the third explanatory variable; Information processing methods. (Appendix 9.7) 10. The information processing method according to claim 9, classifying the first data into a plurality of categories based on a correspondence between the explanatory variables and the objective variable; generating the first model for predicting the dependent variable from the explanatory variables based on the first data for each of the classifications; predicting a predicted value from the explanatory variables by inputting the explanatory variables of the first data into the second model and using the first model corresponding to the predicted explanatory variables; further classifying the first data into the plurality of categories based on an error between a predicted value predicted from the explanatory variables of the first data and the dependent variable corresponding to the explanatory variables; Information processing methods. (Appendix 10) acquiring first data consisting of pairs of explanatory variables and dependent variables classified into a plurality of categories based on correspondence relationships between the explanatory variables and dependent variables, and a first model generated for each category based on the first data so as to predict the dependent variables from the explanatory variables; generating a second model that predicts the first model corresponding to the explanatory variables based on information representing a correspondence relationship between the explanatory variables of the first data based on the classification and each of the first models; A program that causes a computer to perform a process. [Explanation of symbols]

[0046] 10. Information processing equipment 11 Data decomposition section 12 Prediction Model Creation Department 13 Gate model learning section 14 Prediction Department 15 Data storage unit 16 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 Acquisition Department 122 Generation part

Claims

1. an acquisition unit that acquires first data consisting of pairs of explanatory variables and dependent variables classified into a plurality of categories based on correspondence relationships between the explanatory variables and dependent variables, and a first model that is generated for each category based on the first data so as to predict the dependent variables from the explanatory variables; a generation unit that generates a second model that predicts the first model corresponding to the explanatory variables based on information representing a correspondence relationship between the explanatory variables of the first data based on the classification and each of the first models; An information processing device comprising:

2. 2. The information processing device according to claim 1, a prediction unit that predicts a predicted value from the second explanatory variable by inputting a second explanatory variable into the second model and using the first model corresponding to the second explanatory variable predicted by the second model, Information processing device.

3. 2. The information processing device according to claim 1, the generation unit generates the second model that predicts the first model corresponding to the explanatory variables based on weight information that indicates a degree to which the explanatory variables of the first data based on the classification correspond to each of the first models. Information processing device.

4. 4. The information processing device according to claim 3, the generation unit generates the second model that predicts the degree to which the explanatory variables correspond to each of the first models. Information processing device.

5. 4. The information processing device according to claim 3, the generation unit sets the weight information so that the explanatory variables correspond to a higher degree to the first model as the first model has a smaller prediction error in the first data, and generates the second model. Information processing device.

6. 5. The information processing device according to claim 4, a prediction unit that inputs a second explanatory variable into the second model, predicts a degree to which the second explanatory variable corresponds to each of the first models, and predicts a predicted value from the second explanatory variable using each of the first models according to the predicted degree; Information processing device.

7. 2. The information processing device according to claim 1, a prediction unit that predicts a predicted value from a third explanatory variable by inputting a third explanatory variable into the second model and using the first model corresponding to the third explanatory variable predicted by the second model, the generation unit generates the second model so as to reduce an error between a predicted value predicted from the third explanatory variable and a third dependent variable that is paired in advance with the third explanatory variable. Information processing device.

8. 2. The information processing device according to claim 1, a classification unit that classifies the first data into a plurality of categories based on a correspondence between the explanatory variables and the objective variables; a first model generation unit that generates the first model for predicting the dependent variable from the explanatory variables based on the first data for each of the classifications; a prediction unit that predicts a predicted value from the explanatory variables by inputting the explanatory variables of the first data into the second model and using the first model corresponding to the predicted explanatory variables, the classifying unit classifies the first data into the plurality of categories based on an error between a predicted value predicted from the explanatory variables of the first data and the dependent variable corresponding to the explanatory variables; Information processing device.

9. acquiring first data consisting of pairs of explanatory variables and dependent variables classified into a plurality of categories based on correspondence relationships between the explanatory variables and dependent variables, and a first model generated for each category based on the first data so as to predict the dependent variables from the explanatory variables; generating a second model that predicts the first model corresponding to the explanatory variables based on information representing a correspondence relationship between the explanatory variables of the first data based on the classification and each of the first models; Information processing methods.

10. acquiring first data consisting of pairs of explanatory variables and dependent variables classified into a plurality of categories based on correspondence relationships between the explanatory variables and dependent variables, and a first model generated for each category based on the first data so as to predict the dependent variables from the explanatory variables; generating a second model that predicts the first model corresponding to the explanatory variables based on information representing a correspondence relationship between the explanatory variables of the first data based on the classification and each of the first models; A program that causes a computer to perform a process.

Images