Information processing system, information processing apparatus, and information processing method

Abstract

To predict an effect of an intervention from a feature representing a target when a situation of the intervention related to the target and a result related to the target cannot be acquired in association with each other.SOLUTION: The first information processing device includes a first acquisition unit configured to acquire a feature value representing a target and a result obtained with respect to the target, a first prediction unit configured to predict a status of an intervention that may affect the result based on the feature value, and a first prediction model training unit configured to train a prediction model for predicting an effect of the intervention by joint learning by the first information processing device and a second information processing device based on the feature value, the result, and the status of the intervention. The second information processing device includes a second acquisition unit that acquires a feature value representing a target and a situation of an intervention, a second prediction unit that predicts the result based on the feature value, and a second prediction model training unit that trains a prediction model by the above-described federated learning based on the feature value, the result, and the situation of the intervention.SELECTED DRAWING: Figure 1

Description

[Technical Field]

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

[0002] There are known techniques for predicting the effects of interventions implemented to affect outcomes related to a target. For example, Patent Literature 1 describes a technique for predicting an increase in the rate of product purchases (an example of the effect of an intervention) due to an intervention, namely, displaying an advertisement for a product, based on the characteristics of a user (an example of a target). According to this technique, a prediction model is generated that predicts the effect of displaying an advertisement based on the user's characteristics, using training data that links the user's characteristics, whether or not the advertisement was displayed, and whether or not the product was purchased. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] International Publication No. 2021 / 235200 Brochure Summary of the Invention [Problem to be solved by the invention]

[0004] Here, the organization that knows the status of the intervention regarding the target (for example, an advertisement distributor that knows whether an advertisement was displayed to a user) may be different from the organization that knows the results obtained regarding the target (for example, an advertiser that knows whether a product was purchased by a user). Furthermore, it may be difficult for such different organizations to exchange information in a manner that allows the status of the intervention and the results to be linked. The technology described in Patent Document 1 has a problem in that, when the status of the intervention regarding the target and the results obtained regarding the target cannot be linked and obtained, the effect of the intervention cannot be predicted from the characteristics that represent the target.

[0005] The present disclosure has been made in consideration of the above-mentioned problems, and one exemplary purpose thereof is to provide a technology that can predict the effect of an intervention from characteristics that describe a subject when it is not possible to link and obtain the status of the intervention regarding the subject and the results obtained regarding the subject. [Means for solving the problem]

[0006] An information processing system according to an exemplary aspect of the present disclosure includes a first information processing device and a second information processing device, wherein the first information processing device comprises a first acquisition means for acquiring features representing an object and results obtained for the object, a first prediction means for predicting an intervention status that may have affected the result based on the features, and a first prediction model training means for training a prediction model that predicts the effect of the intervention by associative learning between the first information processing device and the second information processing device based on the features, the result, and the intervention status, and the second information processing device comprises a second acquisition means for acquiring the features and the intervention status, a second prediction means for predicting the result based on the features, and a second prediction model training means for training the prediction model by associative learning based on the features, the result, and the intervention status.

[0007] An information processing device according to an exemplary aspect of the present disclosure includes an acquisition means for acquiring features representing a target and results obtained for the target, a prediction means for predicting an intervention status that may have affected the results based on the features, and a prediction model training means for training a prediction model that predicts the effect of the intervention by associative learning with the device itself and another information processing device that can acquire the intervention status based on the features, the results, and the intervention status.

[0008] An information processing device according to an exemplary aspect of the present disclosure includes an acquisition means for acquiring features representing a subject and the status of interventions that may have affected a result obtained for the subject, a prediction means for predicting the result based on the features, and a prediction model training means for training a prediction model that predicts the effect of the intervention based on the features, the result, and the status of the intervention through associative learning between the information processing device itself and another information processing device that can acquire the result.

[0009] An information processing method according to an exemplary aspect of the present disclosure is an information processing method executed by an information processing system including a first information processing device and a second information processing device, the information processing method including: a first acquisition process in which at least one processor included in the first information processing device acquires features representing a target and a result obtained for the target; a first prediction process in which at least one processor included in the first information processing device predicts an intervention status that may have affected the result based on the features; a first prediction model training process in which at least one processor included in the first information processing device trains a prediction model that predicts the effect of the intervention by federated learning with the first information processing device and the second information processing device based on the features, the result, and the intervention status; a second acquisition process in which at least one processor included in the second information processing device acquires the features and the intervention status; a second prediction process in which at least one processor included in the second information processing device predicts the result based on the features; and a second prediction model training process in which at least one processor included in the second information processing device trains the prediction model by federated learning based on the features, the result, and the intervention status.

[0010] An information processing method according to an exemplary aspect of the present disclosure includes an acquisition process in which at least one processor included in an information processing device acquires features representing a target and results obtained regarding the target; a prediction process in which the at least one processor predicts an intervention status that may have affected the results based on the features; and a prediction model training process in which the at least one processor trains a prediction model that predicts the effect of the intervention based on the features, the results, and the intervention status through federated learning with the information processing device and another information processing device that can acquire the intervention status.

[0011] An information processing method according to an exemplary aspect of the present disclosure includes an acquisition process in which at least one processor included in an information processing device acquires features representing a target and the status of an intervention that may have affected a result obtained for the target; a prediction process in which the at least one processor predicts the result based on the features; and a prediction model training process in which the at least one processor trains a prediction model that predicts the effect of the intervention based on the features, the result, and the status of the intervention through federated learning with the information processing device and another information processing device that can acquire the result. [Effects of the Invention]

[0012] According to one exemplary aspect of the present disclosure, an exemplary effect is provided in that a technology can be provided that can predict the effect of an intervention from characteristics that describe a subject when it is not possible to link and obtain the status of the intervention regarding the subject and the results obtained regarding the subject. [Brief explanation of the drawings]

[0013] [Figure 1] 1 is a block diagram illustrating a configuration of an information processing system according to the present disclosure. [Figure 2] FIG. 1 is a flow diagram showing the flow of an information processing method according to the present disclosure. [Figure 3] 1 is a block diagram illustrating a configuration of an information processing device according to the present disclosure. [Figure 4] FIG. 1 is a flow diagram showing the flow of an information processing method according to the present disclosure. [Figure 5] 1 is a block diagram illustrating a configuration of an information processing device according to the present disclosure. [Figure 6] FIG. 1 is a flow diagram showing the flow of an information processing method according to the present disclosure. [Figure 7] 1 is a diagram schematically illustrating an overview of an information processing system according to the present disclosure. [Figure 8] 1 is a block diagram illustrating a configuration of an information processing system according to the present disclosure. [Figure 9] FIG. 1 is a flow diagram showing the flow of an information processing method according to the present disclosure. [Figure 10] 10A to 10C are diagrams illustrating models according to modified examples of the present disclosure. [Figure 11] FIG. 10 is a flowchart showing the flow of an information processing method according to a modified example of the present disclosure. [Figure 12] FIG. 1 is a diagram schematically illustrating an application example using an information processing system according to the present disclosure. [Figure 13] FIG. 2 is a block diagram showing the hardware configuration of a computer that functions as each device according to the present disclosure. DETAILED DESCRIPTION OF THE INVENTION

[0014] The following are examples of embodiments of the present invention. However, the present invention is not limited to the exemplary embodiments shown below, and various modifications are possible within the scope of the claims. For example, embodiments obtained by appropriately combining the technologies (part or all of the products or methods) employed in the exemplary embodiments shown below may also be included in the scope of the present invention. Furthermore, embodiments obtained by appropriately omitting some of the technologies employed in the exemplary embodiments shown below may also be included in the scope of the present invention. Furthermore, the effects mentioned in the exemplary embodiments shown below are examples of effects expected in the exemplary embodiments, and do not define the scope of the present invention. In other words, embodiments that do not exhibit the effects mentioned in the exemplary embodiments shown below may also be included in the scope of the present invention.

[0015] First Exemplary Embodiment A first exemplary embodiment, which is an example of an embodiment of the present invention, will be described in detail with reference to the drawings. This exemplary embodiment is the basic form of each exemplary embodiment described later. The scope of application of each technology employed in this exemplary embodiment is not limited to this exemplary embodiment. That is, each technology employed in this exemplary embodiment can also be employed in other exemplary embodiments included in the present disclosure, to the extent that no particular technical obstacles arise. Furthermore, each technology shown in the drawings referenced to explain this exemplary embodiment can also be employed in other exemplary embodiments included in the present disclosure, to the extent that no particular technical obstacles arise.

[0016] (Overview of Information Processing System 1) The information processing system 1 is a system that generates a predictive model by collaboration between multiple information processing devices, which predicts the effect of an intervention that may affect the outcome of a subject based on the subject's features. Hereinafter, the method of generating a predictive model by collaboration between multiple information processing devices will be referred to as federated learning.

[0017] Here, the term "subject" refers to an object for which the effect of an intervention is to be predicted, and examples thereof include, but are not limited to, visitors to a website, patients at a medical institution, etc. The term "features representing the object" refers to information that describes the object, and examples thereof include, but are not limited to, age, gender, preferences, product purchase history, medical history, and pre-existing conditions. The term "intervention" refers to a process or action taken to affect the results obtained for the object, and examples thereof include, but are not limited to, "displaying an advertisement" and "treatment." The term "effect of an intervention" refers to the degree to which the results obtained for the object improve as a result of the intervention, and examples thereof include, but are not limited to, an increase in purchase rate due to displaying an advertisement, and the effect of treatment on the medical condition.

[0018] To generate a predictive model that predicts the effect of an intervention, training data is required, for example, including (i) features related to the subject, (ii) the results obtained for the subject, and (iii) the circumstances of the intervention that may have affected the results. The "circumstances of the intervention" may be, for example, but are not limited to, whether or not an intervention was performed, the type of intervention if performed, the extent of the intervention, etc. Hereinafter, "the results obtained for the subject" may be referred to as "the results for the subject" or simply as "the results." Furthermore, "the circumstances of the intervention that may have affected the results" may be referred to as "the circumstances of the intervention for the subject" or simply as "the circumstances of the intervention."

[0019] Here, there are cases where the information processing device capable of acquiring the "results" is different from the information processing device capable of acquiring the "intervention status." In such cases, the information processing system 1 can create the above-described prediction model without requiring each information processing device to disclose the results or intervention status acquired by the other information processing device.

[0020] (Configuration of Information Processing System 1) The configuration of the information processing system 1 will be described with reference to Fig. 1. Fig. 1 is a block diagram showing the configuration of the information processing system 1.

[0021] As shown in FIG. 1, the information processing system 1 includes an information processing device 10 and an information processing device 20. The information processing devices 10 and 20 are communicatively connected via a network. Each of the information processing devices 10 and 20 has at least a function of training local models as multiple clients in federated learning. Either one of the information processing devices 10 and 20 may further have a function of integrating multiple local models as a server in federated learning. Each of the information processing devices 10 and 20 may be communicatively connected to a device (not shown) that functions as a server in federated learning. The number of information processing devices 10 included in the information processing system 1 is not limited to one and may be multiple. The number of information processing devices 20 included in the information processing system 1 is not limited to one and may be multiple.

[0022] The information processing device 10 is an example of a first information processing device. The information processing device 10 is a device that can acquire features representing an object and results related to the object, but cannot acquire the status of interventions that may have affected the results.

[0023] The information processing device 10 includes a first acquisition unit 11, a first prediction unit 12, and a first prediction model training unit 13. The first acquisition unit 11 is an example of a configuration that realizes first acquisition means. The first prediction unit 12 is an example of a configuration that realizes first prediction means. The first prediction model training unit 13 is an example of a configuration that realizes first prediction model training means.

[0024] The first acquisition unit 11 acquires features representing an object and results obtained for the object. Hereinafter, the information acquired by the first acquisition unit 11 will be referred to as first input information. The first input information includes features representing an object and results obtained for the object. It is desirable that the first input information include a set of the features representing the object and the results obtained for each of a plurality of objects.

[0025] The first prediction unit 12 predicts the status of an intervention that could have affected an outcome related to a target based on features representing the target. Specifically, the first prediction unit 12 predicts the status of an intervention related to the target based on features related to each of multiple targets included in the first input information. For example, the first prediction unit 12 may make the prediction using a model that predicts the status of an intervention related to the target based on features related to the target.

[0026] The first prediction model training unit 13 trains a prediction model that predicts the effect of an intervention through associative learning by the information processing devices 10 and 20 based on the feature values, the result, and the intervention status. For example, the first prediction model training unit 13 may use the first input information and the predicted intervention status as training data, and train a first prediction model that predicts the effect of an intervention on a subject based on feature values ​​that represent the subject by machine learning with reference to a model that integrates the first prediction model and a second prediction model trained by the information processing device 20. Note that the first prediction model and the second prediction model are local models in associative learning. For example, the information processing device 10 may use the first prediction model after training as a prediction model, or may use a model that integrates the first prediction model and the second prediction model after training as a prediction model.

[0027] The information processing device 20 is an example of a second information processing device. The information processing device 20 is a device that can acquire features representing a target and the status of interventions that may have affected results related to the target, but cannot acquire the results.

[0028] The information processing device 20 includes a second acquisition unit 21, a second prediction unit 22, and a second prediction model training unit 23. The second acquisition unit 21 is an example of a configuration that realizes second acquisition means. The second prediction unit 22 is an example of a configuration that realizes second prediction means. The second prediction model training unit 23 is an example of a configuration that realizes second prediction model training means.

[0029] The second acquisition unit 21 acquires feature quantities representing the target and the status of interventions that may have affected the results obtained for the target. Hereinafter, the information acquired by the second acquisition unit 21 will be referred to as second input information. The second input information includes feature quantities representing the target and the status of interventions that may have affected the results obtained for the target. It is desirable that the second input information include a set of the feature quantities and the status of the interventions for each of a plurality of targets.

[0030] The second prediction unit 22 predicts an outcome for an object based on a feature representing the object. Specifically, the second prediction unit 22 predicts an outcome for the object based on a feature for each of a plurality of objects included in the second input information. For example, the second prediction unit 22 may make a prediction using a model that predicts an outcome for the object based on the feature for the object.

[0031] The second prediction model training unit 23 trains a prediction model that predicts the effect of the intervention based on the feature values, the results, and the status of the intervention through associative learning by the information processing devices 10 and 20. For example, the second prediction model training unit 23 may use the second input information and the predicted results as training data to train a second prediction model that predicts the effect of the intervention on the subject based on the feature values ​​representing the subject, by machine learning with reference to a model obtained by integrating the second prediction model with the first prediction model trained by the information processing device 10. As described above, the first prediction model and the second prediction model are local models in associative learning. For example, the information processing device 20 may use the second prediction model after training as a prediction model, or may use a model obtained by integrating the first prediction model and the second prediction model after training as a prediction model.

[0032] (Associative Learning) Here, as a technique for training the prediction models by the first prediction model training unit 13 and the second prediction model training unit 23, for example, associative learning is adopted. In other words, the first prediction model and the second prediction model are each trained by associative learning. In associative learning, the "integrated model" is referred to as, for example, a global model. As described above, the first prediction model and the second prediction model are each referred to as, for example, a local model in associative learning. In other words, the first prediction model, the second prediction model, and the global model are trained as prediction models by associative learning.

[0033] The targets of the associative learning may be the entire first prediction model and the entire second prediction model, or a part of the first prediction model and a part of the second prediction model. When a part of the first prediction model and a part of the second prediction model are each trained by associative learning, the remaining part of the first prediction model may be trained locally in the information processing device 10 without using associative learning. Furthermore, the remaining part of the second prediction model may be trained locally in the information processing device 20 without using associative learning.

[0034] In federated learning, the following steps are repeated: (i) the information processing device 10 trains a first prediction model by referring to a global model, and the information processing device 20 trains a second prediction model by referring to the global model; and (ii) the trained first prediction model and the trained second prediction model are integrated to generate a new global model. The training data used to train the first prediction model and the training data used to train the second prediction model do not need to be disclosed between the information processing device 10 and the information processing device 20.

[0035] (First prediction model) The first prediction model is a local model trained by the information processing device 10A to obtain, by associative learning, a model that predicts the effect of an intervention on a subject based on features that describe the subject. The first prediction model may include, for example, a third prediction model that predicts an outcome with an intervention based on the features, and a fourth prediction model that predicts an outcome without the intervention based on the features. In this case, information obtained by subtracting the output of the fourth prediction model from the output of the third prediction model may be output from the first prediction model as the effect of the intervention.

[0036] If the "intervention status" includes only whether or not an intervention was performed, the inputs of the third and fourth prediction models may be only "features." If the "intervention status" includes other factors besides whether or not an intervention was performed, such as the type and degree of intervention, the inputs of the third prediction model, which predicts outcomes when an intervention is performed, may include the intervention status in addition to the features.

[0037] For example, the first prediction model training unit 13 classifies each pair of feature values ​​and results included in the first input information into either "with intervention" or "without intervention" based on the intervention status predicted by the first prediction unit 12. As a result, the third prediction model is trained using the pair of feature values ​​and results classified as "with intervention" as training data. In other words, the third prediction model is trained so that when the feature values ​​are input, a result in the case of intervention is output. Note that, as described above, if the intervention status includes other than the presence or absence of intervention, the third prediction model may be trained so that when the feature values ​​and the intervention status are input, a result in the case of intervention is output. Furthermore, the fourth prediction model is trained using the pair of feature values ​​and results classified as "without intervention" as training data. In other words, the fourth prediction model is trained so that when the feature values ​​are input, a result in the case of no intervention is output.

[0038] Furthermore, the first prediction model is trained with reference to the global model. "Training with reference to the global model" means, for example, training the first prediction model using the parameters of the global model as the initial parameters. For example, if the first prediction model includes a third prediction model and a fourth prediction model, the global model may include a first global model and a second global model, which will be described later. In this case, the third prediction model is trained based on the first global model, and the fourth prediction model is trained based on the second global model.

[0039] (Second forecast model) The second prediction model is a local model trained by the information processing device 20A to obtain, through associative learning, a model that predicts the effect of an intervention on a subject based on features that describe the subject. The second prediction model may include, for example, a fifth prediction model that predicts an outcome with an intervention based on the features, and a sixth prediction model that predicts an outcome without the intervention based on the features. In this case, information obtained by subtracting the output of the sixth prediction model from the output of the fifth prediction model may be output from the second prediction model as the effect of the intervention.

[0040] If the "intervention status" includes only whether or not an intervention was performed, the inputs of the fifth and sixth prediction models may be only "features." If the "intervention status" includes other factors besides whether or not an intervention was performed, such as the type and degree of intervention, the inputs of the fifth prediction model, which predicts outcomes when an intervention is performed, may include the intervention status in addition to the features.

[0041] For example, the second prediction model training unit 23 classifies each pair of feature values ​​and intervention status included in the second input information into either intervention status or no intervention status based on the intervention status. As a result, the fifth prediction model is trained using, as training data, a pair of feature values ​​classified as intervention status and a result predicted based on the feature values. In other words, the fifth prediction model is trained so that, when the feature values ​​are input, a result in the case of intervention is output. Note that, as described above, if the intervention status includes other than the presence or absence of intervention, the fifth prediction model may be trained so that, when the feature values ​​and the intervention status are input, a result in the case of intervention is output. Furthermore, a sixth prediction model is trained using, as training data, a pair of feature values ​​classified as no intervention status and a result predicted based on the feature values. In other words, the sixth prediction model is trained so that, when the feature values ​​are input, a result in the case of no intervention is output.

[0042] Furthermore, the second prediction model is trained based on the global model. "Trained based on the global model" means that the second prediction model is trained using the parameters of the global model as initial parameters. For example, if the second prediction model includes a fifth prediction model and a sixth prediction model, the global model may include a first global model and a second global model, which will be described later. In this case, the fifth prediction model is trained based on the first global model, and the sixth prediction model is trained based on the second global model.

[0043] (Global model) The global model is a model in which a first prediction model and a second prediction model are integrated to generate, by associative learning, a model that predicts the effect of an intervention on a subject based on features that represent the subject. "Models are integrated" may mean, for example, that parameters defining each model are integrated. In other words, the global model is defined by the integrated parameters. Integrating parameters may mean, but is not limited to, taking the average of the parameters. Note that, when the information processing system 1 includes multiple information processing devices 10, the global model is a model in which multiple first prediction models and multiple second prediction models are integrated. Also, when the information processing system 1 includes multiple information processing devices 20, the global model is a model in which multiple first prediction models and multiple second prediction models are integrated.

[0044] For example, the global model may include a first global model in which the third prediction model and the fifth prediction model are integrated, and a second global model in which the fourth prediction model and the sixth prediction model are integrated. In this case, information obtained by subtracting the output of the second global model from the output of the first global model may be output from the global model as the effect of the intervention.

[0045] The global model is not limited to including the first global model and the second global model. For example, if the third prediction model and the fourth prediction model are linear models, the first prediction model training unit 13 may calculate the first prediction model defined by parameters obtained by subtracting the parameters of the fourth prediction model from the parameters of the third prediction model. Similarly, if the fifth prediction model and the sixth prediction model are linear models, the second prediction model training unit 23 may calculate the second prediction model defined by parameters obtained by subtracting the parameters of the sixth prediction model from the parameters of the fifth prediction model.

[0046] (Effects of Information Processing System 1) As described above, the information processing system 1 includes information processing devices 10 and 20, and the information processing device 10 is equipped with the above-mentioned first acquisition unit 11, first prediction unit 12, and first prediction model training unit 13, and the information processing device 20 is equipped with the above-mentioned second acquisition unit 21, second prediction unit 22, and second prediction model training unit 23.

[0047] Therefore, according to the information processing system 1, the information processing device 10 can obtain a prediction model (first prediction model or global model) that predicts the effect of an intervention for a subject from the feature quantities related to the subject, without needing to know the status of the intervention for the subject, which the information processing device 10 cannot obtain, and without disclosing the results related to the subject obtained by the information processing device 20. Furthermore, the information processing device 20 can obtain a prediction model (second prediction model or global model) that predicts the effect of an intervention for the subject from the feature quantities related to the subject, without needing to know the results related to the subject, which the information processing device 10 cannot obtain, and without disclosing the status of the intervention for the subject obtained by the information processing device 10. As a result, according to the information processing system 1, it is possible to obtain the effect of predicting the effect of an intervention from the feature quantities related to the subject when the status of the intervention for the subject and the results obtained for the subject cannot be linked and obtained.

[0048] (Flow of information processing method S1) The flow of information processing method S1 will be described with reference to Fig. 2. For example, when the above-described information processing devices 10 and 20 each include at least one processor, the information processing system 1 executes information processing method S1. Fig. 2 is a flow diagram showing the flow of information processing method S1. As shown in Fig. 2, information processing method S1 includes a first acquisition process S11, a first prediction process S12, a first prediction model training process S13, a second acquisition process S21, a second prediction process S22, and a second prediction model training process S23.

[0049] In the first acquisition process S11, at least one processor (for example, the first acquisition unit 11) included in the information processing device 10 acquires features representing an object and results obtained regarding the object. In other words, first input information is acquired.

[0050] In the first prediction process S12, at least one processor (e.g., the first prediction unit 12) provided in the information processing device 10 predicts the status of intervention that could have affected the outcome related to the target based on features representing the target included in the first input information.

[0051] In the first prediction model training process S13, at least one processor (e.g., first prediction model training unit 13) included in the information processing device 10 trains a prediction model that predicts the effect of the intervention based on the feature amount, the result, and the status of the intervention by associative learning using the information processing devices 10 and 20. For example, the at least one processor uses the first input information and the predicted status of the intervention as training data and trains a first prediction model that predicts the effect of the intervention on the subject based on the feature amount representing the subject by machine learning, with reference to a model obtained by integrating the first prediction model and a second prediction model trained by the information processing device 20.

[0052] In the second acquisition process S21, at least one processor (for example, the second acquisition unit 21) included in the information processing device 20 acquires a feature representing an object and an intervention status related to the object. In other words, second input information is acquired.

[0053] In the second prediction process S22, at least one processor (for example, the second prediction unit 22) included in the information processing device 20 predicts a result related to the target based on a feature representing the target included in the second input information.

[0054] In the second prediction model training process S23, at least one processor (e.g., second prediction model training unit 23) included in the information processing device 20 trains a prediction model that predicts the effect of the intervention by associative learning using the information processing devices 10 and 20, based on the feature amounts, the result, and the status of the intervention. For example, the at least one processor uses second input information and the predicted result as training data and trains a second prediction model that predicts the effect of the intervention on the subject based on the feature amounts representing the subject, by machine learning, with reference to a model in which the second prediction model and the first prediction model trained by the information processing device 10 are integrated.

[0055] The information processing device 10 repeatedly executes at least the first prediction model training process S13. For example, the information processing device 10 may repeat a series of processes including the first prediction process S12 and the first prediction model training process S13, or may repeat the first prediction model training process S13 without repeating the first prediction process S12. The information processing device 20 repeatedly executes at least the second prediction model training process S23. For example, the information processing device 20 may repeat a series of processes including the second prediction process S22 and the second prediction model training process S23, or may repeat the second prediction model training process S23 without repeating the second prediction process S22.

[0056] In the first prediction model training process S13 and the second prediction model training process S23, a model obtained by integrating the first prediction model and the second prediction model after training in the previous processes S13 and S23 is referenced as a new global model. In the processes S13 and S23, the same new global model is referenced to train the first prediction model and the second prediction model, respectively. Note that the same initial global model is referenced in the initial first prediction model training process S13 and the initial second prediction model training process S23.

[0057] Note that the process of integrating the first prediction model and the second prediction model to generate a new global model may be performed by either the information processing device 10 or the information processing device 20. For example, when the information processing device 10 performs this process, the information processing device 10 may integrate the first prediction model after training on its own device and the second prediction model after training received from the information processing device 20 to generate a new global model. Furthermore, the information processing device 10 may transmit the new global model to the information processing device 20. When the information processing device 20 performs this process, the same description can be given by replacing the information processing devices 10 and 20 and the first prediction model and the second prediction model in the above description of when the information processing device 10 performs this process.

[0058] Furthermore, the process of integrating the first prediction model and the second prediction model to generate a new global model may be performed by a server (not shown) different from either of the information processing devices 10 and 20. In this case, the server may integrate the post-training first prediction model received from the information processing device 10 and the post-training second prediction model received from the information processing device 20 to generate a new global model and transmit the new global model to each of the information processing devices 10 and 20.

[0059] In this manner, (i) the information processing device 10 trains the first prediction model by referring to the global model, and the information processing device 20 trains the second prediction model by referring to the global model, and (ii) a model obtained by integrating the first prediction model and the second prediction model becomes a new global model. The repetition may be performed a predetermined number of times, or may be performed until a first prediction model, a second prediction model, or a global model with a predetermined accuracy is obtained.

[0060] (Effect of information processing method S1) As described above, the information processing method S1 includes the first acquisition process S11, the first prediction process S12, the first prediction model training process S13, the second acquisition process S21, the second prediction process S22, and the second prediction model training process S23. Therefore, the information processing method S1 can achieve the same effects as the information processing system 1.

[0061] (Configuration of information processing device 10) FIG. 3 is a block diagram showing the configuration of the information processing device 10. The information processing device 10 is an example of a configuration that realizes an information processing device according to an exemplary aspect of the present disclosure. As shown in FIG. 3, the information processing device 10 includes a first acquisition unit 11, a first prediction unit 12, and a first prediction model training unit 13. The first acquisition unit 11 is an example of a configuration that realizes acquisition means. The first prediction unit 12 is an example of a configuration that realizes prediction means. The first prediction model training unit 13 is an example of a configuration that realizes prediction model training means. The details of the first acquisition unit 11, the first prediction unit 12, and the first prediction model training unit 13 have been described above, and therefore will not be described again.

[0062] (Effects of the information processing device 10) As described above, the information processing device 10 is configured to include the first acquisition unit 11, the first prediction unit 12, and the first prediction model training unit 13. Therefore, the information processing device 10 can obtain a first prediction model (or a global model) that predicts the effect of an intervention on a target from features related to the target, without needing to know the status of the intervention on the target, which the information processing device 10 cannot obtain, and without disclosing the results related to the target obtained by the information processing device 10 to the outside. As a result, the information processing device 10 can obtain the effect of predicting the effect of an intervention on the target from features representing the target, even when the status of the intervention on the target and the results obtained for the target cannot be obtained in association with each other.

[0063] (Flow of information processing method S10) The flow of the information processing method S10 will be described with reference to FIG. 4. For example, when the above-described information processing device 10 includes at least one processor, the information processing device 10 executes the information processing method S10. FIG. 4 is a flow diagram showing the flow of the information processing method S10. As shown in FIG. 4, the information processing method S10 includes a first acquisition process S11 (an example of an acquisition process), a first prediction process S12 (an example of a prediction process), and a first prediction model training process S13 (an example of a prediction model training process). Details of the first acquisition process S11, the first prediction process S12, and the first prediction model training process S13 have been described above, and therefore will not be described repeatedly. Note that the at least one processor repeatedly executes at least the first prediction model training process S13. For example, the at least one processor may repeat a series of processes including the first prediction process S12 and the first prediction model training process S13, or may repeat the first prediction model training process S13 without repeating the first prediction process S12.

[0064] (Effect of information processing method S10) As described above, the information processing method S10 employs a configuration including the above-described first acquisition process S11, first prediction process S12, and first prediction model training process S13. Therefore, the information processing method S10 can achieve the same effects as the information processing device 10.

[0065] (Configuration of information processing device 20) FIG. 5 is a block diagram showing the configuration of the information processing device 20. The information processing device 20 is an example of a configuration that realizes an information processing device according to an exemplary aspect of the present disclosure. As shown in FIG. 5, the information processing device 20 includes a second acquisition unit 21, a second prediction unit 22, and a second prediction model training unit 23. The second acquisition unit 21 is an example of a configuration that realizes acquisition means. The second prediction unit 22 is an example of a configuration that realizes prediction means. The second prediction model training unit 23 is an example of a configuration that realizes prediction model training means. The details of the second acquisition unit 21, the second prediction unit 22, and the second prediction model training unit 23 are as described above, and therefore will not be repeated.

[0066] (Effects of the information processing device 20) As described above, the information processing device 20 is configured to include the second acquisition unit 21, the second prediction unit 22, and the second prediction model training unit 23. Therefore, the information processing device 20 can obtain a second prediction model (or a global model) that predicts the effect of an intervention on a target from features related to the target, without needing to know the results obtained for the target, which the device itself cannot obtain, and without disclosing the status of the intervention on the target obtained by the device itself to the outside. As a result, the information processing device 20 can obtain the effect of predicting the effect of an intervention from features representing the target, even when the status of the intervention on the target and the results obtained for the target cannot be obtained in association with each other.

[0067] (Flow of information processing method S20) The flow of the information processing method S20 will be described with reference to FIG. 6. For example, when the above-described information processing device 20 includes at least one processor, the information processing device 20 executes the information processing method S20. FIG. 6 is a flow diagram showing the flow of the information processing method S20. As shown in FIG. 6, the information processing method S20 includes a second acquisition process S21 (an example of an acquisition process), a second prediction process S22 (an example of a prediction process), and a second prediction model training process S23 (an example of a prediction model training process). The details of the second acquisition process S21, the second prediction process S22, and the second prediction model training process S23 have been described above, and therefore will not be described repeatedly. Note that the at least one processor repeatedly executes at least the second prediction model training process S23. For example, the at least one processor may repeat a series of processes including the second prediction process S22 and the second prediction model training process S23, or may repeat the second prediction model training process S23 without repeating the second prediction process S22.

[0068] (Effect of information processing method S20) As described above, the information processing method S20 includes the second acquisition process S21, the second prediction process S22, and the second prediction model training process S23. Therefore, the information processing method S20 can achieve the same effects as the information processing device 20.

[0069] Second Exemplary Embodiment A second exemplary embodiment, which is one example of an embodiment of the present invention, will be described in detail with reference to the drawings. Components having the same functions as those described in the above exemplary embodiment will be assigned the same reference numerals, and their description will be omitted as appropriate. The scope of application of each technology employed in this exemplary embodiment is not limited to this exemplary embodiment. That is, each technology employed in this exemplary embodiment can also be employed in other exemplary embodiments included in the present disclosure, to the extent that no particular technical hindrance occurs. Furthermore, each technology shown in each drawing referenced to explain this exemplary embodiment can also be employed in other exemplary embodiments included in the present disclosure, to the extent that no particular technical hindrance occurs.

[0070] (Outline of Information Processing System 1A) The information processing system 1A is configured similarly to the information processing system 1, and is also configured as follows: In the information processing system 1A, a result prediction model trained by the information processing device 10A capable of acquiring results related to a target, and an intervention prediction model trained by the information processing device 20A capable of acquiring the status of intervention related to the target are exchanged between the information processing devices 10A and 20A.

[0071] 7 is a diagram showing a schematic overview of the information processing system 1A. Note that, hereinafter, a subscript (e.g., i) shown in FIG. 7 will be written with an underscore next to the character, such as "_i." Also, a tilde symbol (~) placed above a character (e.g., x) will be written with the character followed by the tilde symbol, such as "x~." Also, a hat symbol (^) placed above a character (e.g., t) will be written with the character followed by the hat symbol, such as "t^."

[0072] 7, information processing device 10A is a device capable of acquiring a result y_i related to an object. N sets (N is a natural number equal to or greater than 2, i=1, 2, ..., N) of a feature x_i representing an object and a result y_i related to the object are input to information processing device 10A as first input information. Using the first input information as training data, information processing device 10A trains a result prediction model "Y" so that y_i is output when x_i is input.

[0073] The information processing device 20A is a device capable of acquiring the intervention status t_i related to a target. M sets (M is a natural number equal to or greater than 2) of pairs of feature x~_i representing a target and intervention status t_i related to the target are input to the information processing device 20A as second input information. Note that x~_i and x_i are common feature values ​​(e.g., gender and age, etc.), but are information that is not disclosed to each other and are not necessarily feature values ​​related to the same target (e.g., the same user). The information processing device 20A trains an intervention prediction model "T" using the second input information as training data so that t_i is output when x~_i is input. The outcome prediction model "Y" and the intervention prediction model "T" are exchanged between the information processing devices 10A and 20A.

[0074] The information processing device 10A calculates "T(x_i)" using the intervention prediction model "T" obtained by the exchange, and assigns it to the first input information as a pseudo label "t^_i" indicating the intervention status. As a result, N sets of triples, each consisting of a feature x_i representing the target, a pseudo label t^_i, and a result y_i, are obtained.

[0075] The information processing device 20A calculates "Y(x~_i)" using the result prediction model "Y" obtained by the exchange, and assigns it to the second input information as a pseudo label "y^_i" indicating the result. As a result, M sets of triples are obtained: feature x~_i representing the target, intervention status t^_i, and pseudo label y_i indicating the result.

[0076] The information processing devices 10A and 20A can each use triples as training data, and as a result, can generate a prediction model "u(x)" by associative learning. The prediction model u(x) is a model that takes a feature "x" representing a target as input and outputs the effect of an intervention "u(x)." (Configuration of information processing system 1A) The configuration of the information processing system 1A will be described with reference to FIG. 8. FIG. 8 is a block diagram showing the configuration of the information processing system 1A. The information processing system 1A includes information processing devices 10A and 20A. The information processing device 10A includes a result prediction model training unit 14 and a global model generation unit 19 in addition to the first acquisition unit 11, first prediction unit 12, and first prediction model training unit 13 included in the information processing device 10. The information processing device 20A includes an intervention prediction model training unit 24 in addition to the second acquisition unit 21, second prediction unit 22, and second prediction model training unit 23 included in the information processing device 20. The result prediction model training unit 14 is an example of a configuration that realizes result prediction model training means. The intervention prediction model training unit 24 is an example of a configuration that realizes intervention prediction model training means.

[0077] The result prediction model training unit 14 trains a result prediction model that predicts an outcome related to an object based on features that represent the object. For example, the result prediction model training unit 14 uses first input information as training data to train a result prediction model that predicts an outcome related to the object based on features that represent the object through machine learning. In other words, when features included in the first input information are input, the result prediction model is trained so that the result included in the first input information is output. Furthermore, the result prediction model training unit 14 provides the result prediction model to the information processing device 20A.

[0078] The intervention prediction model training unit 24 trains an intervention prediction model that predicts the intervention status of a target based on features that represent the target. For example, the intervention prediction model training unit 24 uses second input information as training data to train an intervention prediction model by machine learning that predicts the intervention status of the target based on the features that represent the target. In other words, when the features included in the second input information are input, the intervention prediction model is trained so that the intervention status included in the second input information is output. Furthermore, the intervention prediction model training unit 24 provides the intervention prediction model to the information processing device 10A.

[0079] The first prediction unit 12 is configured similarly to the first prediction unit 12 included in the information processing device 10, and is also configured as follows: The first prediction unit 12 predicts the intervention status for a target using an intervention prediction model provided by the information processing device 20A. For example, the first prediction unit 12 inputs feature amounts representing each of the multiple targets included in the first input information into the intervention prediction model, thereby acquiring the intervention status output from the intervention prediction model.

[0080] The second prediction unit 22 is configured similarly to the second prediction unit 22 included in the information processing device 20, and is also configured as follows: The second prediction unit 22 predicts a result related to an object using a result prediction model provided by the information processing device 10A. For example, the second prediction unit 22 inputs feature amounts representing each of the multiple objects included in the second input information into the result prediction model, thereby acquiring a result output from the result prediction model.

[0081] Here, the training of the result prediction model is performed independently of the training of the prediction model. In other words, the training of the result prediction model performed in the information processing device 10 is performed independently of the training of the first prediction model, which is a local model in associative learning for generating a prediction model. This allows, for example, the training of the result prediction model to be completed before the training of the first prediction model begins. Furthermore, the trained result prediction model can be provided to the information processing device 20A before the training of the second prediction model begins. As a result, the information processing device 20A can accurately predict the "results" required for the training of the second prediction model using the trained result prediction model before starting the training of the second prediction model.

[0082] Furthermore, training of the intervention prediction model is performed independently of training of the prediction model. In other words, training of the intervention prediction model performed in the information processing device 20 is performed independently of training of the second prediction model, which is a local model in associative learning for generating a prediction model. This allows, for example, training of the intervention prediction model to be completed before training of the second prediction model begins. Furthermore, the trained intervention prediction model can be provided to the information processing device 10A before training of the first prediction model begins. As a result, the information processing device 10A can accurately predict the "intervention situation" required for training of the first prediction model using the trained intervention prediction model before starting training of the first prediction model.

[0083] The global model generation unit 19 generates an initial global model and provides it to the first prediction model training unit 13 and the second prediction model training unit 23. The initial global model is defined by initial parameters. The global model generation unit 19 also generates a new global model by integrating the first prediction model and the second prediction model and provides it to the first prediction model training unit 13 and the second prediction model training unit 23.

[0084] (Flow of information processing method S1A) The flow of information processing method S1A will be described with reference to Fig. 9. For example, when the above-described information processing devices 10A and 20A each include at least one processor, the information processing system 1A executes information processing method S1A. Fig. 9 is a flow diagram showing the flow of information processing method S1A. As shown in Fig. 9, information processing method S1A includes steps S101 to S108 executed by information processing device 10A and steps S201 to S205 executed by information processing device 20A.

[0085] In step S101, the first acquisition unit 11 acquires first input information including a feature representing an object and a result related to the object. Step S101 is an example of a first acquisition process.

[0086] In step S201, the second acquisition unit 21 acquires second input information including features representing the target and the status of intervention related to the target. Step S201 is an example of a second acquisition process. The execution order of steps S101 and S201 is not limited to this order and may be any order. Note that "the execution order of multiple steps is any order" includes executing each step in any order, executing some or all of each step in parallel, etc.

[0087] In step S102, the result prediction model training unit 14 trains the result prediction model using the first input information as training data. The result prediction model training unit 14 also transmits the trained result prediction model to the information processing device 20A. Step S102 is an example of a result prediction model training process.

[0088] In step S202, the intervention prediction model training unit 24 trains the intervention prediction model using the second input information as training data. The intervention prediction model training unit 24 also transmits the trained intervention prediction model to the information processing device 10A. Step S202 is an example of an intervention prediction model training process. The execution order of steps S102 and S202 is not limited to this order and can be any order.

[0089] In step S103, the first prediction unit 12 predicts the intervention status of a target based on features representing the target included in the first input information, using the intervention prediction model received from the information processing device 20A in step S202. As a result, a pseudo label indicating the predicted intervention status is assigned to the features representing each of the multiple targets included in the first input information. Step S103 is an example of the first prediction process.

[0090] In step S203, the second prediction unit 22 predicts a result related to an object included in the second input information based on a feature representing the object, using the result prediction model received from the information processing device 10A in step S102. As a result, a pseudo label indicating a predicted result is assigned to the feature representing each of the multiple objects included in the second input information. Step S203 is an example of a second prediction process. The order of execution of steps S103 and S203 is not limited to this order and can be any order.

[0091] In step S104, the global model generation unit 19 generates an initial global model. In other words, the global model generation unit 19 sets initial parameters that define the global model. Step S104 may be executed before step S105, and is not necessarily executed after step S103.

[0092] In step S105, the global model generation unit 19 transmits the global model to the information processing device 20A.

[0093] In step S106, the first prediction model training unit 13 trains the first prediction model by referring to the global model, using the first input information and the pseudo-label indicating the intervention status as training data. Step S106 is an example of a first prediction model training process. The details of the training of the first prediction model are as described in the first exemplary embodiment, and therefore will not be repeated.

[0094] In step S204, the second prediction model training unit 23 trains the second prediction model by referring to the global model, using the second input information and pseudo labels indicating the results as training data. Step S204 is an example of a second prediction model training process. Details of the training of the second prediction model are as described in the first exemplary embodiment, and therefore will not be repeated. The second prediction model training unit 23 transmits the trained second prediction model to the information processing device 10A. The execution order of steps S106 and S204 is not limited to this order and can be any order.

[0095] In step S107, the global model generation unit 19 generates a new global model by integrating the first prediction model trained in step S106 and the second prediction model received from the information processing device 20A in step S204.

[0096] In step S108, information processing device 10A determines whether to end the associative learning. If it is determined not to end the associative learning, information processing device 10A repeats the processing from step S105. Note that information processing device 10A may determine whether to end the associative learning based on whether the number of repetitions has reached a predetermined number, or may determine based on whether the accuracy of the first prediction model (or global model) has exceeded a threshold. However, the viewpoint for determining whether to end the associative learning is not limited to these examples.

[0097] The global model transmitted to the information processing device 20A in the next step S105 is the new global model generated in step S107 by integrating the first prediction model and the second prediction model. In addition, in the next step S106, the first prediction model is trained with reference to the "new global model generated by integrating the first prediction model and the second prediction model." Note that in each repeated step S106 (first prediction model training process), the pseudo label indicating the intervention status assigned in advance in step S103 is referenced.

[0098] In step S205, information processing device 20A determines whether or not to terminate the associative learning. If it is determined not to terminate, information processing device 20A repeats the processing from step S204. Note that information processing device 20A may determine whether or not to terminate the associative learning based on whether or not the number of repetitions has reached a predetermined number, or may determine based on whether or not the accuracy of the second prediction model (or global model) has exceeded a threshold. However, the viewpoint for determining whether or not to terminate the associative learning is not limited to these examples.

[0099] In the next step S204, the second prediction model is trained with reference to the "new global model obtained by integrating the first prediction model and the second prediction model" received from the information processing device 10A. Note that in each repeated step S204 (second prediction model training process), the pseudo label indicating the result previously assigned in step S203 is referenced.

[0100] If it is determined in step S108 or S205 that the process is to be terminated, the information processing method S1A terminates. This allows the information processing device 10A to obtain a first prediction model or a global model that predicts the effect of an intervention on a subject based on the features that represent the subject. Furthermore, the information processing device 20A can obtain a second prediction model or a global model that predicts the effect of an intervention on the subject based on the features that represent the subject.

[0101] (Advantages of this exemplary embodiment) As described above, in the information processing system 1A, the information processing device 10A further includes an outcome prediction model training unit 14 that trains an outcome prediction model that predicts an outcome for a target based on features representing the target. The information processing device 20A further includes an intervention prediction model training unit 24 that trains an intervention prediction model that predicts an intervention status for the target based on features representing the target. The first prediction unit 12 predicts the intervention status for the target using the intervention prediction model, and the second prediction unit 22 predicts an outcome for the target using the outcome prediction model. Therefore, in addition to the effects achieved by the information processing system 1, the information processing device 10A that cannot acquire the intervention status for the target can accurately predict the intervention status by using the intervention prediction model trained by the information processing device 20A that can acquire the intervention status. Furthermore, the information processing device 20A that cannot acquire the outcome for the target can accurately predict the outcome by using the outcome prediction model trained by the information processing device 10A that can acquire the outcome.

[0102] Furthermore, information processing system 1A is configured such that training of the outcome prediction model is performed independently of training of the prediction model, and training of the intervention prediction model is performed independently of training of the prediction model. Therefore, according to information processing system 1A, in addition to the effects achieved by information processing system 1, associative learning for generating a prediction model (first prediction model, second prediction model, or global model) that predicts the effect of an intervention on a subject based on features representing the subject can be initiated using as training data the intervention status and results accurately predicted by the trained outcome prediction model and intervention prediction model, resulting in the effect of rapid convergence.

[0103] [Variation 1] The above-described information processing system 1A can be modified as follows: The prediction model and the outcome prediction model share a front-end part. In other words, the first prediction model, which is a local model in the associative learning for generating the prediction model, and the outcome prediction model share a front-end part. Furthermore, the prediction model and the intervention prediction model share a front-end part. In other words, the second prediction model, which is a local model in the associative learning for generating the prediction model, and the intervention prediction model share a front-end part.

[0104] For example, the first prediction model training unit 13 may set the front-end of the trained outcome prediction model as the front-end of the first prediction model, and train the rear-end of the first prediction model by associative learning. For example, when the first prediction model and the outcome prediction model are configured using neural networks, a predetermined number of layers from the input side of the multiple layers constituting the trained outcome prediction model may be used as the front-end. Furthermore, for example, the second prediction model training unit 23 may set the front-end of the trained intervention prediction model as the front-end of the second prediction model, and train the rear-end of the second prediction model by associative learning. For example, when the second prediction model and the intervention prediction model are configured using neural networks, a predetermined number of layers from the input side of the multiple layers constituting the trained intervention prediction model may be used as the front-end. Note that it is desirable that the output from the front-end shared by the first prediction model and the outcome prediction model and the output from the front-end shared by the second prediction model and the intervention prediction model have at least the same number of dimensions.

[0105] In this modification, for example, the information processing method S1A shown in Fig. 9 is modified as follows: In step S104, the global model generation unit 19 generates an initial global model corresponding to the latter part of the first prediction model.

[0106] In step S106, instead of training the entire first prediction model, the first prediction model training unit 13 sets the front part of the result prediction model as the front part of the first prediction model and trains the rear part of the first prediction model.

[0107] In step S204, instead of training the entire second prediction model, the second prediction model training unit 23 sets the front part of the intervention prediction model as the front part of the second prediction model and trains the rear part of the second prediction model.

[0108] In step S107, the global model generation unit 19 generates a global model that integrates the latter part of the first prediction model and the latter part of the second prediction model.

[0109] (Effects of this modified example) In this modification, the first prediction model and the outcome prediction model share a front-end, and the second prediction model and the intervention prediction model share a front-end. This allows the models to be repeatedly trained by associative learning to be the rear-ends of the first prediction model and the second prediction model, thereby reducing computational costs.

[0110] [Variation 2] The above-described first variation of the information processing system 1A can be further modified as follows. The first prediction model training unit 13 and the outcome prediction model training unit 14 train the prediction model and the outcome prediction model in parallel, sharing their leading-stage parts. In other words, the first prediction model training unit 13 and the outcome prediction model training unit 14 train the first prediction model and the outcome prediction model, which are local models in the associative learning for generating the prediction model, in parallel, sharing their leading-stage parts. The second prediction model training unit 23 and the intervention prediction model training unit 24 train the prediction model and the intervention prediction model in parallel, sharing their leading-stage parts. In other words, the second prediction model training unit 23 and the intervention prediction model training unit 24 train the second prediction model and the intervention prediction model, which are local models in the associative learning for generating the prediction model, in parallel, sharing their leading-stage parts.

[0111] FIG. 10 is a diagram illustrating each model according to this modification. As shown in FIG. 10, the first model front-end is a front-end shared by the first prediction model and the outcome prediction model. A feature representing the target is input to the first model front-end. The intermediate representation output from the first model front-end is input to the first prediction model rear-end and the outcome prediction model rear-end. The first prediction model rear-end outputs the effect of the intervention. The outcome prediction model rear-end outputs the result.

[0112] Also, as shown in FIG. 10, the second model front-end is a front-end shared by the second prediction model and the intervention prediction model. Features representing the target are input to the second model front-end. The intermediate representation output from the second model front-end is input to the second prediction model rear-end and the intervention prediction model rear-end. The second prediction model rear-end outputs the effect of the intervention. The intervention prediction model rear-end outputs the status of the intervention.

[0113] Here, the first prediction model consisting of a first model front-end and a first prediction model rear-end, and the second prediction model consisting of a second model front-end and a second prediction model rear-end are trained by associative learning. The outcome prediction model rear-end and the intervention prediction model rear-end are trained independently in the information processing devices 10A and 20A, respectively, and are exchanged.

[0114] In this modification, information processing method S1B is executed instead of information processing method S1A shown in Fig. 9. Fig. 11 is a flow diagram showing the flow of information processing method S1B. As shown in Fig. 11, information processing method S1B includes steps S151 to S158 executed by information processing device 10A and steps S251 to S256 executed by information processing device 20A.

[0115] In step S151, the first acquisition unit 11 acquires first input information. Step S151 is the same as step S101 described above.

[0116] In step S251, the second acquiring unit 21 acquires second input information. Step S251 is the same as step S201 described above. The execution order of steps S151 and S251 is not limited to this order and may be any order.

[0117] In step S152, the global model generation unit 19 generates an initial global model. Step S152 is similar to the above-mentioned step S104. Step S152 may be executed before step S153 is executed, and does not necessarily have to be executed after step S151.

[0118] In step S153, the global model generation unit 19 transmits the global model to the information processing device 20 A. Step S153 is the same as step S105 described above.

[0119] In step S154, the first prediction model training unit 13 and the outcome prediction model training unit 14 of the information processing device 10A train the first prediction model and the outcome prediction model in parallel while sharing the first model front-end, using the first input information and the pseudo-label indicating the intervention status as training data. In other words, the first prediction model front-end, the first prediction model rear-end, and the outcome prediction model rear-end are trained in parallel. A known multitask learning method can be used for this parallel training process.

[0120] Among the trainings performed in parallel in step S154, the training of the first prediction model is performed with reference to the global model. That is, the parameters of the front and rear stages of the global model are set as initial parameters of the front and rear stages of the first prediction model, and training is performed.

[0121] Furthermore, among the trainings performed in parallel in step S154, training of the first prediction model requires a pseudo label indicating the intervention status, but the intervention prediction model required to assign the pseudo label has not yet been obtained when step S154 is performed for the first time. In this case, the pseudo label may be assigned using any intervention prediction model. Note that pseudo labels when step S154 is performed for the second time or later will be described later. Furthermore, in the first execution of step S154, training of the first prediction model may be omitted, and only training of the outcome prediction model may be performed.

[0122] In step S252, the second prediction model training unit 23 and the intervention prediction model training unit 24 of the information processing device 20A train the second prediction model and the intervention prediction model in parallel, sharing the second model front-end, using the second input information and pseudo labels indicating the results as training data. In other words, the second prediction model front-end, the second prediction model rear-end, and the intervention prediction model rear-end are trained in parallel. A known multi-task learning method can be used for this parallel training process.

[0123] In the training performed in parallel in step S252, the second prediction model is trained with reference to the global model, i.e., the parameters of the front and rear stages of the global model are set as the initial parameters of the front and rear stages of the second prediction model, and training is performed.

[0124] Furthermore, among the training steps performed in parallel in step S252, training of the second prediction model requires a pseudo label indicating the result, but the result prediction model required to assign the pseudo label has not yet been obtained when step S252 is performed for the first time. In this case, the pseudo label may be assigned using any result prediction model. Note that pseudo labels when step S252 is performed for the second or subsequent times will be described later. Furthermore, in the first execution of step S252, training of the second prediction model may be omitted, and only training of the intervention prediction model may be performed. The order of execution of steps S154 and S252 is not limited to this order and may be any order.

[0125] In step S155, the result prediction model training unit 14 of the information processing device 10A transmits the trained result prediction model subsequent part to the information processing device 20A.

[0126] In step S253, the intervention prediction model training unit 24 of the information processing device 20A transmits the post-training intervention prediction model subsequent part to the information processing device 10A. The execution order of steps S155 and S253 is not limited to this order and may be any order.

[0127] In step S156, the first prediction unit 12 of the information processing device 10A assigns pseudo labels indicating the intervention status to features representing each of the multiple targets included in the first input information, using the trained first model front-end and the received intervention prediction model back-end. The pseudo labels assigned in this step are referenced when the first prediction model and the outcome prediction model are trained in parallel in the next step S154.

[0128] In step S254, the second prediction unit 22 of the information processing device 20A assigns pseudo labels indicating results to features representing each of the multiple targets included in the second input information, using the trained second model front-end unit and the received result prediction model back-end unit. The pseudo labels assigned in this step are referenced when training the second prediction model and the intervention prediction model in parallel in the next step S252. The execution order of steps S156 and S254 is not limited to this order and can be any order.

[0129] In step S255, the second prediction model training unit 23 of the information processing device 20A transmits the second prediction model made up of the second model front-end part and the second prediction model back-end part to the information processing device 10A.

[0130] In step S157, the global model generation unit 19 generates a global model by integrating the first prediction model consisting of the first model front-end part and the first prediction model rear-end part with the received second prediction model.

[0131] In step S158, information processing device 10A determines whether to end the associative learning. A specific example of determining whether to end the associative learning is the same as in step S108, and therefore detailed description will not be repeated. If it is determined not to end, information processing device 10A repeats the processing from step S153.

[0132] The global model transmitted to the information processing device 20A in the next step S153 is a new global model generated in the previous step S157 by integrating the first prediction model and the second prediction model. Furthermore, in the next step S154, the front part of the new global model becomes the first model front part at the start of training. Furthermore, the rear part of the new global model becomes the first prediction model rear part at the start of training. Furthermore, the result prediction model rear part trained in the previous step S154 becomes the result prediction model rear part at the start of training. The first model front part, the first prediction model rear part, and the result prediction model rear part are trained in parallel. Furthermore, the pseudo labels assigned in the previous step S156 are referenced during the training.

[0133] In step S256, information processing device 20A determines whether to end the associative learning. A specific example of determining whether to end the associative learning is the same as in step S205, and therefore detailed description will not be repeated. If it is determined not to end, information processing device 20A repeats the processing from step S252.

[0134] In the next step S252, the front part of the "new global model obtained by integrating the first prediction model and the second prediction model" received from the information processing device 10A becomes the front part of the second model at the start of training. Also, the rear part of the new global model becomes the rear part of the second prediction model at the start of training. Also, the rear part of the intervention prediction model trained in the previous step S252 becomes the rear part of the intervention prediction model at the start of training. Then, the training of the front part of the second model, the rear part of the second prediction model, and the rear part of the intervention prediction model is performed in parallel. Also, in this training, the pseudo label assigned in the previous step S254 is referenced.

[0135] If it is determined in step S158 or S256 that the process is to end, the information processing method S1B ends. This allows the information processing device 10A to obtain a first prediction model or a global model that predicts the effect of an intervention on a subject based on the features that represent the subject. Furthermore, the information processing device 20A can obtain a second prediction model or a global model that predicts the effect of an intervention on the subject based on the features that represent the subject.

[0136] (Effects of this modified example) In this modification, in addition to the second modification, the first prediction model training unit 13 and the outcome prediction model training unit 14 train the prediction model and the outcome prediction model in parallel while sharing their front-end parts, and the second prediction model training unit 23 and the intervention prediction model training unit 24 train the prediction model and the intervention prediction model in parallel while sharing their front-end parts. This configuration allows the information processing device 10A to transfer knowledge between the task of predicting the effect of an intervention and the task of predicting an outcome, thereby improving the performance of the prediction model and the outcome prediction model. Furthermore, the information processing device 20A allows knowledge transfer between the task of predicting the effect of an intervention and the task of predicting the intervention status, thereby improving the performance of the prediction model and the intervention prediction model. Furthermore, because the front-end parts of the outcome prediction model and the intervention prediction model are generated by associative learning, the effects of associative learning can be reflected in the assignment of pseudo-labels.

[0137] [Other Modifications] The global model generation unit 19 may be included in the information processing device 20A instead of being included in the information processing device 10A, or may be included in a device different from either the information processing device 10A or 20A. In this modification, the information processing system 1A also achieves the above-described effects.

[0138] [Application example] 12 is a diagram schematically illustrating an application example in which a sales promotion effect due to advertisement display is predicted using the information processing systems 1 and 1A. For example, an advertiser requests one or more advertisement distribution platforms to distribute advertisements in order to promote product purchases.

[0139] In this case, as shown in FIG. 12 , an advertiser server managed by an advertiser can acquire the gender and age (an example of a feature) of a user (A, B, C, . . .) (an example of a target) and whether or not the user has made a purchase (an example of a result). An advertisement delivery server managed by an advertisement delivery platform can acquire the gender and age of a user and whether or not an advertisement has been displayed to the user terminal (A, B, C, . . .) used by the user (an example of an intervention status). However, due to privacy protection and other considerations, the advertiser server and the advertisement delivery server cannot mutually disclose whether or not a user has made a purchase and whether or not an advertisement has been displayed to the user. This makes it difficult to link whether or not an advertisement has been displayed to the user and whether or not the user has made a purchase. Therefore, the advertiser server is used as an example of the information processing device 10, 10A, and the advertisement delivery server is used as an example of the information processing device 20, 20A. This makes it possible to generate a prediction model that predicts the effect of advertisement display (an example of an intervention effect) based on the user's gender and age.

[0140] According to this application example, an advertiser can cooperate with an advertisement distribution platform to generate a prediction model for predicting sales promotion effects, without needing to know whether an advertisement was displayed to users who purchased a product, or without needing to disclose whether or not the user purchased a product. As a result, the advertiser can predict the sales promotion effects of advertisement display according to user characteristics, and can request the advertisement distribution platform to distribute advertisements preferentially to users with the characteristics that will be most effective.

[0141] Furthermore, according to this application example, the advertising distribution platform can cooperate with advertisers to generate a prediction model for predicting sales promotion effects, without needing to know whether users who viewed a product advertisement purchased the product, or without needing to disclose whether or not an advertisement was displayed. As a result, the advertising distribution platform can predict the sales promotion effects of advertisement display based on user characteristics, and can deliver advertisements preferentially based on the effects. This can be expected to increase incentives from advertisers, for example.

[0142] [Software implementation example] Some or all of the functions of the information processing devices 10, 10A, 20, 20A (hereinafter also referred to as "the above-mentioned devices") that make up the information processing systems 1, 1A may be realized by hardware such as an integrated circuit (IC chip), or by software.

[0143] In the latter case, each of the above devices is realized by, for example, a computer that executes instructions of a program, which is software that realizes each function. An example of such a computer (hereinafter referred to as computer C) is shown in Figure 13. Figure 13 is a block diagram showing the hardware configuration of computer C that functions as each of the above devices.

[0144] The computer C includes at least one processor C1 and at least one memory C2. The memory C2 stores a program P for causing the computer C to operate as each of the above-mentioned devices. In the computer C, the processor C1 reads and executes the program P from the memory C2, thereby realizing the functions of each of the above-mentioned devices.

[0145] The processor C1 may be, for example, a central processing unit (CPU), a graphic processing unit (GPU), a digital signal processor (DSP), a micro processing unit (MPU), a floating point number processing unit (FPU), a physics processing unit (PPU), a tensor processing unit (TPU), a quantum processor, a microcontroller, or a combination thereof. The memory C2 may be, for example, a flash memory, a hard disk drive (HDD), a solid state drive (SSD), or a combination thereof.

[0146] The computer C may further include a RAM (Random Access Memory) for expanding the program P during execution and for temporarily storing various data. The computer C may also include a communication interface for transmitting and receiving data to and from other devices. The computer C may also include an input / output interface for connecting input / output devices such as a keyboard, mouse, display, and printer.

[0147] Furthermore, the program P can be recorded on a non-transitory tangible recording medium M that can be read by the computer C. Such a recording medium M can be, for example, a tape, a disk, a card, a semiconductor memory, or a programmable logic circuit. The computer C can acquire the program P via such a recording medium M. The program P can also be transmitted via a transmission medium. Such a transmission medium can be, for example, a communication network or broadcast waves. The computer C can also acquire the program P via such a transmission medium.

[0148] Furthermore, the functions of each of the devices may be realized by a single processor provided in a single computer, by multiple processors provided in a single computer working in cooperation, or by multiple processors provided in each of multiple computers working in cooperation. Furthermore, the programs for causing each of the devices to realize the functions may be stored in a single memory provided in a single computer, or may be distributed and stored in multiple memories provided in a single computer, or may be distributed and stored in multiple memories provided in each of multiple computers.

[0149] [Appendix 1] This disclosure includes the techniques described in the following appendices. However, the present invention is not limited to the techniques described in the following appendices, and various modifications are possible within the scope of the claims.

[0150] (Appendix 1) a first information processing device and a second information processing device; The first information processing device a first acquisition means for acquiring a feature representing an object and a result obtained regarding the object; a first prediction means for predicting an intervention status that may have influenced the outcome based on the feature amount; a first prediction model training means for training a prediction model that predicts the effect of the intervention by federated learning using the first information processing device and the second information processing device, based on the feature amount, the result, and the intervention status; The second information processing device a second acquisition means for acquiring the feature amount and the intervention status; a second prediction means for predicting the result based on the feature amount; and a second prediction model training means for training the prediction model by the associative learning based on the feature amount, the result, and the intervention status. Information processing system.

[0151] (Appendix 2) the first information processing device further includes a result prediction model training means for training a result prediction model that predicts the result from the feature amount; the second information processing device further includes an intervention prediction model training means for training an intervention prediction model that predicts the intervention status from the feature amount, the first prediction means predicts the intervention status using the intervention prediction model; The second prediction means predicts the outcome using the outcome prediction model. 10. The information processing system of claim 1.

[0152] (Appendix 3) training the outcome prediction model independently of training the predictive model; training the intervention prediction model independently of training the prediction model; 10. The information processing system of claim 2.

[0153] (Appendix 4) the prediction model and the outcome prediction model share a front end; The prediction model and the intervention prediction model share a front part, 10. The information processing system of claim 2.

[0154] (Appendix 5) the first prediction model training means and the outcome prediction model training means train the prediction model and the outcome prediction model in parallel while sharing a front-end part; the second prediction model training means and the intervention prediction model training means train the prediction model and the intervention prediction model in parallel while sharing a front-end part; 5. The information processing system of claim 4.

[0155] (Appendix 6) an acquisition means for acquiring a feature representing an object and a result obtained regarding the object; a prediction means for predicting an intervention status that may have affected the outcome based on the feature amount; a prediction model training means for training a prediction model that predicts the effect of the intervention by associative learning between the device itself and another information processing device that can acquire the intervention status, based on the feature amount, the result, and the intervention status; and An information processing device comprising:

[0156] (Appendix 7) an acquisition means for acquiring features representing the object and the status of interventions that may have influenced the results obtained for the object; a prediction means for predicting the result based on the feature amount; a prediction model training means for training a prediction model that predicts the effect of the intervention by associative learning between the device itself and another information processing device that can acquire the result, based on the feature amount, the result, and the intervention status; and An information processing device comprising:

[0157] (Appendix 8) An information processing method executed by an information processing system including a first information processing device and a second information processing device, a first acquisition process in which at least one processor included in the first information processing device acquires features representing an object and results obtained regarding the object; a first prediction process in which at least one processor included in the first information processing device predicts a state of an intervention that may have affected the outcome based on the feature amount; a first prediction model training process in which at least one processor included in the first information processing device trains a prediction model that predicts the effect of the intervention through federated learning by the first information processing device and the second information processing device, based on the feature amount, the result, and the intervention status; a second acquisition process in which at least one processor included in the second information processing device acquires the feature amount and the intervention status; a second prediction process in which at least one processor included in the second information processing device predicts the result based on the feature amount; a second prediction model training process in which at least one processor included in the second information processing device trains the prediction model by the associative learning based on the feature amount, the result, and the intervention status; An information processing method, including:

[0158] (Appendix 9) an acquisition process in which at least one processor included in the information processing device acquires features representing an object and results obtained regarding the object; a prediction process in which the at least one processor predicts intervention conditions that may have affected the outcome based on the features; a prediction model training process in which the at least one processor trains a prediction model that predicts the effect of the intervention based on the feature amount, the result, and the intervention status through federated learning with the information processing device and another information processing device that can acquire the intervention status; An information processing method, including:

[0159] (Appendix 10) an acquisition process in which at least one processor included in the information processing device acquires features representing the object and a status of intervention that may have influenced the result obtained regarding the object; a prediction process in which the at least one processor predicts the outcome based on the feature amount; a prediction model training process in which the at least one processor trains a prediction model that predicts the effect of the intervention based on the feature amount, the result, and the status of the intervention through federated learning with the information processing device and another information processing device that can acquire the result; An information processing method, including:

[0160] (Appendix 11) An information processing program that causes at least one processor included in the information processing device described in Supplementary Note 6 to execute the acquisition process, the prediction process, and the prediction model training process.

[0161] (Appendix 12) An information processing program that causes at least one processor included in the information processing device described in Supplementary Note 7 to execute the acquisition process, the prediction process, and the prediction model training process.

[0162] (Appendix 13) An information processing system including a first information processing device and a second information processing device, The first information processing device includes at least one processor, and the at least one processor a first acquisition process for acquiring features representing an object and results obtained regarding the object; a first prediction process for predicting an intervention status that may have affected the outcome based on the feature amount; a first prediction model training process for training a prediction model that predicts the effect of the intervention through federated learning by the first information processing device and the second information processing device, based on the feature amount, the result, and the intervention status; The second information processing device includes at least one processor, and the at least one processor a second acquisition process for acquiring the feature amount and the intervention status; a second prediction process for predicting the result based on the feature amount; and executing a second prediction model training process to train the prediction model by the federated learning based on the feature amount, the result, and the intervention status. Information processing system. The first information processing device may further include a memory. Furthermore, the memory may store a program for causing the at least one processor included in the first information processing device to execute each of the processes. Furthermore, the second information processing device may further include a memory. Furthermore, the memory may store a program for causing the at least one processor included in the second information processing device to execute each of the processes. [Explanation of symbols]

[0163] 1. 1A Information Processing System 10, 10A, 20, 20A Information processing equipment 11 First acquisition part 12 First Prediction Section 13 First Prediction Model Training Department 14. Outcome Prediction Model Training Department 19 Global Model Generation Unit 21 Second acquisition part 22 Second Prediction Section 23 Second Prediction Model Training Department 24 Intervention Prediction Model Training Department C1 processor C2 Memory

Claims

1. a first information processing device and a second information processing device; The first information processing device a first acquisition means for acquiring a feature representing an object and a result obtained regarding the object; a first prediction means for predicting an intervention status that may have influenced the outcome based on the feature amount; a first prediction model training means for training a prediction model that predicts the effect of the intervention by federated learning using the first information processing device and the second information processing device, based on the feature amount, the result, and the intervention status; The second information processing device a second acquisition means for acquiring the feature amount and the intervention status; a second prediction means for predicting the result based on the feature amount; and a second prediction model training means for training the prediction model by the associative learning based on the feature amount, the result, and the intervention status. Information processing system.

2. the first information processing device further includes a result prediction model training means for training a result prediction model that predicts the result from the feature amount; the second information processing device further includes an intervention prediction model training means for training an intervention prediction model that predicts the intervention status from the feature amount, the first prediction means predicts the intervention status using the intervention prediction model; the second prediction means predicts the outcome using the outcome prediction model; The information processing system according to claim 1 .

3. training the outcome prediction model independently of training the predictive model; training the intervention prediction model independently of training the prediction model; The information processing system according to claim 2 .

4. the prediction model and the outcome prediction model share a front end; The prediction model and the intervention prediction model share a front part, The information processing system according to claim 2 .

5. the first prediction model training means and the outcome prediction model training means train the prediction model and the outcome prediction model in parallel while sharing a front-end part; the second prediction model training means and the intervention prediction model training means train the prediction model and the intervention prediction model in parallel while sharing a front-end part; The information processing system according to claim 4 .

6. an acquisition means for acquiring a feature representing an object and a result obtained regarding the object; a prediction means for predicting an intervention status that may have affected the outcome based on the feature amount; a prediction model training means for training a prediction model that predicts the effect of the intervention by associative learning between the device itself and another information processing device that can acquire the intervention status, based on the feature amount, the result, and the intervention status; and An information processing device comprising:

7. an acquisition means for acquiring features representing the object and the status of interventions that may have influenced the results obtained for the object; a prediction means for predicting the result based on the feature amount; a prediction model training means for training a prediction model that predicts the effect of the intervention by associative learning between the device itself and another information processing device that can acquire the result, based on the feature amount, the result, and the intervention status; and An information processing device comprising:

8. An information processing method executed by an information processing system including a first information processing device and a second information processing device, a first acquisition process in which at least one processor included in the first information processing device acquires features representing an object and results obtained regarding the object; a first prediction process in which at least one processor included in the first information processing device predicts a state of an intervention that may have affected the outcome based on the feature amount; a first prediction model training process in which at least one processor included in the first information processing device trains a prediction model that predicts the effect of the intervention through federated learning by the first information processing device and the second information processing device, based on the feature amount, the result, and the intervention status; a second acquisition process in which at least one processor included in the second information processing device acquires the feature amount and the intervention status; a second prediction process in which at least one processor included in the second information processing device predicts the result based on the feature amount; a second prediction model training process in which at least one processor included in the second information processing device trains the prediction model by the associative learning based on the feature amount, the result, and the intervention status; An information processing method, including:

9. an acquisition process in which at least one processor included in the information processing device acquires features representing an object and results obtained regarding the object; a prediction process in which the at least one processor predicts intervention conditions that may have affected the outcome based on the features; a prediction model training process in which the at least one processor trains a prediction model that predicts an effect of the intervention based on the feature amount, the result, and the intervention status through federated learning with the information processing device and another information processing device that can acquire the intervention status; An information processing method, including:

10. an acquisition process in which at least one processor included in the information processing device acquires features representing the object and a status of intervention that may have influenced the result obtained regarding the object; a prediction process in which the at least one processor predicts the outcome based on the feature; a prediction model training process in which the at least one processor trains a prediction model that predicts the effect of the intervention based on the feature amount, the result, and the status of the intervention through federated learning with the information processing device and another information processing device that can acquire the result; An information processing method, including:

Images