Model training method and device, operation information generation method and device and electronic equipment

Abstract

The embodiment of the invention discloses a model training method and device, an operation information generation method and device and electronic equipment. A specific embodiment of the method comprises the following steps: acquiring value borrowing resource exchange data as training data; inputting the training data into the initial hybrid expert network model to generate an output result; inputting the output result into the initial target value borrowing resource numerical value prediction model to output value borrowing resource numerical value prediction information; according to the value borrowing resource numerical value prediction information, generating a second initial value borrowing resource numerical value diversity prediction model; and in response to determining that the second initial value borrowed resource numerical value diversity prediction model meets a predetermined model optimization target, determining the second initial value borrowed resource numerical value diversity prediction model as a value borrowed resource numerical value diversity prediction model. The implementation mode is related to artificial intelligence, and more accurate value borrowing resource numerical value information can be generated by using the value borrowing resource numerical value diversification prediction model.

Description

Technical Field

[0001] Embodiments of the present disclosure relate to the field of computer technology, and in particular to a model training method, an operation information generating method, a device, and an electronic device. Background Art

[0002] At present, with the continuous development of big data technology and artificial intelligence, it has laid an important foundation for the intelligent transformation and upgrading of related industries (for example, the financial industry). For the prediction of the numerical information of value borrowing resources, the usual method is to use the Uplift-Modeling (UM model) to achieve the prediction of value borrowing resource data information.

[0003] However, the inventors have found that when the above method is used to predict the value information of the value borrowing resource, the following technical problems often occur:

[0004] First, for the One-model in the gain model, when there are many features, the ability to distinguish the treatment feature is lower than that of the other features, resulting in a lower weight of the treatment feature in the One-model, which greatly affects the prediction of the subsequent value borrowing resource numerical information, resulting in low prediction accuracy;

[0005] Second, for the Two-model in the gain model, during the training process of the Two-model, the sample utilization is restricted (that is, the two models in the Two-model cannot utilize each other's samples), resulting in cumulative errors in the Two-model, so that the prediction accuracy of the subsequent value borrowing resource numerical information is low.

[0006] The above information disclosed in this Background section is only for enhancement of understanding of the background of the inventive concept and therefore it may contain information that does not form the prior art that is already known in this country to a person of ordinary skill in the art. Summary of the invention

[0007] The content of this disclosure is used to introduce concepts in a brief form, which will be described in detail in the detailed implementation section below. The content of this disclosure is not intended to identify the key features or essential features of the technical solution claimed for protection, nor is it intended to limit the scope of the technical solution claimed for protection.

[0008] Some embodiments of the present disclosure propose a model training method, an operation information generating method, an apparatus and an electronic device to solve the technical problems mentioned in the above background technology section.

[0009] In a first aspect, some embodiments of the present disclosure provide a model training method, including: obtaining value borrowing resource exchange data as training data, wherein the above training data has a corresponding value resource fallback label; inputting the above training data into an initial hybrid expert network model included in a first initial value borrowing resource numerical diversification prediction model to generate an output result; inputting the above output result into an initial target value borrowing resource numerical prediction model among at least two initial value borrowing resource numerical prediction models to output value borrowing resource numerical prediction information, wherein the above initial target value borrowing resource numerical prediction model and the above value resource fallback label have a A corresponding relationship is established, wherein the at least two initial value borrowing resource numerical prediction models are sub-models in the first initial value borrowing resource numerical diversification prediction model; a second initial value borrowing resource numerical diversification prediction model is generated according to the value borrowing resource numerical prediction information, wherein the second initial value borrowing resource numerical diversification prediction model includes: a hybrid expert network model and a target value borrowing resource numerical prediction model; in response to determining that the second initial value borrowing resource numerical diversification prediction model meets the predetermined model optimization target, the second initial value borrowing resource numerical diversification prediction model is determined as the value borrowing resource numerical diversification prediction model.

[0010] Optionally, the above-mentioned value resource fallback label is a first label characterizing that the value resource has not been fallen back, and the above-mentioned at least two initial value borrowing resource numerical prediction models include: a first initial value borrowing resource numerical prediction model for the above-mentioned first label; and the above-mentioned input of the above-mentioned output result into the initial target value borrowing resource numerical prediction model in at least two initial value borrowing resource numerical prediction models to output value borrowing resource numerical prediction information, including: inputting the above-mentioned output result into the above-mentioned first initial value borrowing resource numerical prediction model to output the value borrowing resource numerical prediction information under the condition that the value resource has not been fallen back.

[0011] Optionally, the above-mentioned value resource fallback label is a second label characterizing that the value resource has been fallen back, and the above-mentioned at least two initial value borrowing resource numerical prediction models include: a second initial value borrowing resource numerical prediction model for the above-mentioned second label; and the above-mentioned input of the above-mentioned output result into the initial target value borrowing resource numerical prediction model in at least two initial value borrowing resource numerical prediction models to output value borrowing resource numerical prediction information, including: inputting the above-mentioned output result into the above-mentioned second initial value borrowing resource numerical prediction model to output value borrowing resource numerical prediction information under the condition that the value resource has been fallen back.

[0012] Optionally, the above-mentioned training data includes: actual numerical information of value-borrowing resources; and generating a second initial value-borrowing resource numerical diversification prediction model based on the above-mentioned value-borrowing resource numerical prediction information, including: generating loss values ​​for the above-mentioned value-borrowing resource numerical actual information and the value-borrowing resource numerical prediction information; based on the above-mentioned loss values, performing model training on the above-mentioned initial hybrid expert network model and the above-mentioned initial target value-borrowing resource numerical prediction model to obtain a hybrid expert network model and a target value-borrowing resource numerical prediction model; based on the above-mentioned hybrid expert network model and the above-mentioned target value-borrowing resource numerical prediction model, generating the above-mentioned second initial value-borrowing resource numerical diversification prediction model.

[0013] Optionally, in response to determining that the second initial value borrowing resource numerical diversification prediction model meets the predetermined model optimization target, before determining the second initial value borrowing resource numerical diversification prediction model as the value borrowing resource numerical diversification prediction model, the method further includes: obtaining a verification data set; using the second initial value borrowing resource numerical diversification prediction model to generate at least one data indicator information for the verification data set; and determining whether the second initial value borrowing resource numerical diversification prediction model meets the predetermined model optimization target based on the at least one data indicator information.

[0014] Optionally, in response to determining that the second initial value borrowing resource numerical diversification prediction model meets the predetermined model optimization target, before determining the second initial value borrowing resource numerical diversification prediction model as the value borrowing resource numerical diversification prediction model, the method also includes: determining the data round corresponding to the data set to which the training data belongs; in response to determining that the data round reaches a preset round, determining that the second initial value borrowing resource numerical diversification prediction model meets the predetermined model optimization target.

[0015] Optionally, the method further includes: in response to determining that the second initial value borrowing resource numerical diversification prediction model does not meet the predetermined model optimization target, reacquiring training data as target training data; and continuing model training for the second initial value borrowing resource numerical diversification prediction model based on the target training data.

[0016] In a second aspect, some embodiments of the present disclosure provide a model training device, comprising: a first acquisition unit, configured to acquire value borrowing resource exchange data as training data, wherein the above-mentioned training data has a corresponding value resource fallback label; a first input unit, configured to input the above-mentioned training data into an initial hybrid expert network model included in a first initial value borrowing resource numerical diversification prediction model to generate an output result; a second input unit, configured to input the above-mentioned output result into an initial target value borrowing resource numerical prediction model among at least two initial value borrowing resource numerical prediction models to output value borrowing resource numerical prediction information, wherein the above-mentioned initial target value borrowing resource numerical prediction model is consistent with the above-mentioned value resource fallback label. There is a one-to-one correspondence between the de-labels, and the at least two initial value borrowing resource numerical prediction models are sub-models in the first initial value borrowing resource numerical diversification prediction model; the first generation unit is configured to generate a second initial value borrowing resource numerical diversification prediction model according to the value borrowing resource numerical prediction information, wherein the second initial value borrowing resource numerical diversification prediction model includes: a hybrid expert network model and a target value borrowing resource numerical prediction model; the determination unit is configured to determine the second initial value borrowing resource numerical diversification prediction model as the value borrowing resource numerical diversification prediction model in response to determining that the second initial value borrowing resource numerical diversification prediction model meets the predetermined model optimization target.

[0017] Optionally, the above-mentioned value resource fallback label is a first label characterizing that the value resource has not been fallen back, and the above-mentioned at least two initial value borrowing resource numerical prediction models include: a first initial value borrowing resource numerical prediction model for the above-mentioned first label; and the first input unit can be configured to: input the above-mentioned output result into the above-mentioned first initial value borrowing resource numerical prediction model to output the value borrowing resource numerical prediction information under the condition that the value resource has not been fallen back.

[0018] Optionally, the above-mentioned value resource fallback label is a second label characterizing that the value resource has been fallen back, and the above-mentioned at least two initial value borrowing resource numerical prediction models include: a second initial value borrowing resource numerical prediction model for the above-mentioned second label; and the first input unit can be configured to: input the above-mentioned output result into the above-mentioned second initial value borrowing resource numerical prediction model to output the value borrowing resource numerical prediction information under the condition that the value resource has been fallen back.

[0019] Optionally, the above-mentioned training data includes: actual information on the numerical values ​​of value-borrowed resources; and the first generation unit can be configured to: generate loss values ​​for the above-mentioned actual information on the numerical values ​​of value-borrowed resources and the predicted numerical values ​​of value-borrowed resources; according to the above-mentioned loss values, perform model training on the above-mentioned initial hybrid expert network model and the above-mentioned initial target value-borrowed resource numerical prediction model to obtain a hybrid expert network model and a target value-borrowed resource numerical prediction model; based on the above-mentioned hybrid expert network model and the above-mentioned target value-borrowed resource numerical prediction model, generate the above-mentioned second initial value-borrowed resource numerical diversification prediction model.

[0020] Optionally, the device also includes: obtaining a verification data set; using the above-mentioned second initial value borrowing resource numerical diversification prediction model to generate at least one data indicator information for the above-mentioned verification data set; and determining whether the above-mentioned second initial value borrowing resource numerical diversification prediction model meets the above-mentioned predetermined model optimization goal based on the above-mentioned at least one data indicator information.

[0021] Optionally, the device also includes: determining a data round corresponding to the data set to which the above-mentioned training data belongs; in response to determining that the above-mentioned data round reaches a preset round, determining that the above-mentioned second initial value borrowing resource numerical diversification prediction model meets the above-mentioned predetermined model optimization goal.

[0022] Optionally, the device also includes: in response to determining that the above-mentioned second initial value borrowing resource numerical diversification prediction model does not meet the predetermined model optimization target, re-acquiring training data as target training data; and continuing model training for the above-mentioned second initial value borrowing resource numerical diversification prediction model based on the above-mentioned target training data.

[0023] In a third aspect, some embodiments of the present disclosure provide a method for generating operation information, including: obtaining user attribute information of a target user; inputting the above user attribute information into a pre-trained value borrowing resource numerical diversification prediction model to output at least two value borrowing resource numerical prediction information, wherein each value borrowing resource numerical prediction information has a corresponding value resource fallback label, and the above value borrowing resource numerical diversification prediction model is generated based on the above method of the first aspect of the present disclosure; based on the above at least two value borrowing resource numerical prediction information, generating operation information representing whether a value resource fallback operation is performed on the above target user.

[0024] In a fourth aspect, some embodiments of the present disclosure provide an operation information generating device, comprising: a second acquisition unit, configured to acquire user attribute information of a target user; a third input unit, configured to input the above user attribute information into a pre-trained value borrowing resource numerical diversification prediction model to output at least two value borrowing resource numerical prediction information, wherein each value borrowing resource numerical prediction information has a corresponding value resource fallback label, and the above value borrowing resource numerical diversification prediction model is generated based on the above method of the first aspect of the present disclosure; a second generation unit, configured to generate operation information representing whether a value resource fallback operation is performed on the above target user based on the above at least two value borrowing resource numerical prediction information.

[0025] In a fifth aspect, some embodiments of the present disclosure provide an electronic device comprising: one or more processors; a storage device on which one or more programs are stored, and when the one or more programs are executed by the one or more processors, the one or more processors implement the method described in any one of the implementation methods in the first and third aspects.

[0026] In a sixth aspect, some embodiments of the present disclosure provide a computer-readable medium having a computer program stored thereon, wherein when the program is executed by a processor, the method described in any one of the implementation modes of the first aspect and the third aspect is implemented.

[0027] In a seventh aspect, some embodiments of the present disclosure provide a computer program product, including a computer program, which implements the method described in any implementation manner in the above-mentioned first aspect when executed by a processor.

[0028] The above-mentioned embodiments of the present disclosure have the following beneficial effects: through the model training method of some embodiments of the present disclosure, using the value borrowing resource numerical diversification prediction model, more accurate value borrowing resource numerical information can be generated. Specifically, the reason why the relevant value borrowing resource numerical information is not accurate is that: first, for the One-model (single model) in the gain model, when there are many features, the ability to distinguish the intervention action (treatment) feature is lower than the other features, resulting in a lower weight of the intervention action feature in the One-model, which greatly affects the prediction of the subsequent value borrowing resource numerical information, resulting in low prediction accuracy; second, for the Two-model (dual model) in the gain model, during the training process of the Two-model, the sample utilization is restricted (that is, the two models in the dual model cannot use each other's samples), resulting in the existence of cumulative errors in the Two-model, so that the prediction accuracy of the subsequent value borrowing resource numerical information is low. Based on this, the model training method of some embodiments of the present disclosure, first, obtains the value borrowing resource exchange data as training data. Among them, the above training data has a corresponding value resource fallback label. Here, the training data is obtained for the subsequent initial value borrowing resource numerical diversification prediction model training. By determining the value resource fallback label corresponding to the training data, the corresponding initial value borrowing resource numerical prediction model is subsequently used to select the corresponding initial value borrowing resource numerical prediction model to output the value borrowing resource numerical prediction information under the corresponding model. Then, the above training data is input into the initial hybrid expert network model included in the first initial value borrowing resource numerical diversification prediction model to generate an output result corresponding to the accurate and targeted features. Next, the above output result is input into the initial target value borrowing resource numerical prediction model in at least two initial value borrowing resource numerical prediction models to output the value borrowing resource numerical prediction information. Among them, there is a one-to-one correspondence between the above initial target value borrowing resource numerical prediction model and the above value resource fallback label. Among them, the above at least two initial value borrowing resource numerical prediction models are sub-models in the above first initial value borrowing resource numerical diversification prediction model. Here, the corresponding initial value borrowing resource numerical prediction model is selected through the value resource fallback label to output the corresponding value borrowing resource numerical prediction information, and the model parameters of the initial hybrid expert network model and the selected corresponding initial value borrowing resource numerical prediction model are updated in a targeted manner. This makes not only the selected corresponding initial value borrowing resource numerical prediction model pay more attention to the feature information corresponding to the value resource fallback label, but also makes full use of the feature information of the training data (that is, for the training data corresponding to different value resource fallback labels, the initial hybrid expert network model also needs to be trained), avoiding the problems existing in One-model and Two-model.Furthermore, according to the above-mentioned value borrowing resource numerical prediction information, a second initial value borrowing resource numerical diversification prediction model is generated. Among them, the above-mentioned second initial value borrowing resource numerical diversification prediction model includes: a hybrid expert network model and a target value borrowing resource numerical prediction model. Here, compared with the first initial value borrowing resource numerical diversification prediction model, the obtained second initial value borrowing resource numerical diversification prediction model is more accurate in predicting the value borrowing resource numerical prediction information under the value resource fallback label. Finally, in response to determining that the above-mentioned second initial value borrowing resource numerical diversification prediction model meets the predetermined model optimization goal, the above-mentioned second initial value borrowing resource numerical diversification prediction model is determined as the value borrowing resource numerical diversification prediction model. In summary, by using the value resource fallback label, the corresponding initial value borrowing resource numerical prediction model in the first initial value borrowing resource numerical diversification prediction model is selected to train the subsequent model, which can not only make full use of the data features of the training data, but also avoid the problem that the weight of the intervention action (i.e., the feature corresponding to the value resource fallback label) feature in the One-model is low, resulting in low prediction accuracy. Therefore, by using the value-borrowing resource numerical diversification prediction model, more accurate value-borrowing resource numerical information can be generated. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the accompanying drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and that components and elements are not necessarily drawn to scale.

[0030] Figure 1 is a schematic diagram of an application scenario of a model training method according to some embodiments of the present disclosure;

[0031] Figure 2 is a flowchart of some embodiments of the model training method according to the present disclosure;

[0032] Figure 3 is a schematic diagram of a hybrid expert network model in some embodiments of the model training method according to the present disclosure;

[0033] Figure 4 is a flowchart of other embodiments of the model training method according to the present disclosure;

[0034] Figure 5 is a flow chart of some embodiments of the method for generating operation information according to the present disclosure;

[0035] Figure 6 is a schematic diagram of the structure of some embodiments of the model training device according to the present disclosure;

[0036] Figure 7 is a schematic diagram of the structure of some embodiments of the operation information generating device according to the present disclosure;

[0037] Figure 8 It is a schematic diagram of the structure of an electronic device suitable for implementing some embodiments of the present disclosure. DETAILED DESCRIPTION

[0038] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments set forth herein. On the contrary, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0039] It should also be noted that, for ease of description, only the parts related to the invention are shown in the drawings. In the absence of conflict, the embodiments and features in the embodiments of the present disclosure can be combined with each other.

[0040] It should be noted that the concepts such as "first" and "second" mentioned in the present disclosure are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0041] It should be noted that the modifications of "one" and "plurality" mentioned in the present disclosure are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0042] The names of the messages or information exchanged between multiple devices in the embodiments of the present disclosure are only used for illustrative purposes and are not used to limit the scope of these messages or information.

[0043] With regard to the collection, storage, and use of data (e.g., training data) involved in this disclosure, before performing corresponding operations, relevant organizations or individuals shall fulfill obligations including conducting data security impact assessments, fulfilling the obligation to inform the data subject, and obtaining the authorization and consent of the data subject in advance.

[0044] The present disclosure will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0045] Figure 1 It is a schematic diagram of an application scenario of the model training method according to some embodiments of the present disclosure.

[0046] exist Figure 1In the application scenario, first, the electronic device 101 can obtain the value borrowing resource exchange data as the training data 102. Among them, the above training data 102 has a corresponding value resource fallback label 103. In this application scenario, the training data 102 can be "user height: 180cm, user name: Zhang**, user contact information: 13****23, loan once on January 2, loan once on January 3, and repay once on January 4, the user loaned a total of 70,000 yuan". The value resource fallback label 103 can be "interest rate cut". Then, the electronic device 101 can input the above training data 102 into the initial hybrid expert network model 105 included in the first initial value borrowing resource numerical diversification prediction model 104 to generate an output result 106. Secondly, the electronic device 101 can input the above output result 106 into the initial target value borrowing resource numerical prediction model 1072 in at least two initial value borrowing resource numerical prediction models 107 to output value borrowing resource numerical prediction information 108. Among them, there is a one-to-one correspondence between the above-mentioned initial target value borrowing resource numerical prediction model 1072 and the above-mentioned value resource fallback label 103. The above-mentioned at least two initial value borrowing resource numerical prediction models 107 are sub-models in the above-mentioned first initial value borrowing resource numerical diversification prediction model 105. In this application scenario, the value borrowing resource numerical prediction information 108 can be "100,000 yuan". Furthermore, the electronic device 101 can generate a second initial value borrowing resource numerical diversification prediction model 109 according to the above-mentioned value borrowing resource numerical prediction information 108. Among them, the above-mentioned second initial value borrowing resource numerical diversification prediction model 109 includes: a hybrid expert network model 110 and a target value borrowing resource numerical prediction model 1111. Finally, in response to determining that the above-mentioned second initial value borrowing resource numerical diversification prediction model 109 meets the predetermined model optimization target, the electronic device 101 can determine the above-mentioned second initial value borrowing resource numerical diversification prediction model 109 as the value borrowing resource numerical diversification prediction model.

[0047] It should be noted that the electronic device 101 can be hardware or software. When the electronic device is hardware, it can be implemented as a distributed cluster consisting of multiple servers or terminal devices, or it can be implemented as a single server or a single terminal device. When the electronic device is embodied as software, it can be installed in the hardware devices listed above. It can be implemented as multiple software or software modules for providing distributed services, for example, or it can be implemented as a single software or software module. No specific limitation is made here.

[0048] It should be understood that Figure 1 The number of electronic devices in the embodiment is only for illustration. Any number of electronic devices may be provided according to implementation requirements.

[0049] Continue to refer Figure 2 , shows a process 200 of some embodiments of the model training method according to the present disclosure. The model training method comprises the following steps:

[0050] Step 201, obtaining value borrowing resource exchange data as training data.

[0051] In some embodiments, the execution entity of the above model training method (for example Figure 1 The electronic device 101 shown) can obtain value borrowing resource exchange data as training data through a wired connection method or a wireless connection method. Among them, the above training data has a corresponding value resource fallback label. Among them, the value borrowing resource exchange data can be the user performance data of the user within the target period for the user's loan change. In practice, the user performance data may include: user basic information, user usage record, and user's new loan amount data within the target period. User basic information may be the user's basic information. For example, user basic information includes: user name, user height, and user residential address. User usage records may be user loan or repayment records. The value resource fallback label may be an interest rate reduction label.

[0052] For example, if the target period is "January 1-January 7", the corresponding user performance data can be "user height: 180cm, user name: Zhang**, user contact information: 13****23, one loan on January 2, one loan on January 3, one repayment on January 4, the user has a total loan of 70,000 yuan".

[0053] Step 202: input the above training data into the initial hybrid expert network model included in the first initial value borrowing resource numerical diversification prediction model to generate an output result.

[0054] In some embodiments, the execution entity may input the training data into the initial mixed expert network model included in the first initial value borrowing resource numerical diversification prediction model to generate an output result. The first initial value borrowing resource numerical diversification prediction model may be a value borrowing resource numerical diversification prediction model whose model has not yet been trained. The value borrowing resource numerical diversification prediction model may be a model for predicting the value of value borrowing resources. In practice, the value borrowing resource numerical diversification prediction model may be a model for predicting the amount of new loans for users. The initial mixed expert network model may be a mixed expert network (MoE, Mixture of Experts) model whose model has not yet been trained.

[0055] See also Figure 3 , showing a schematic diagram of the model structure of the hybrid expert model.

[0056] The hybrid expert model includes: expert network 1, expert network 2, expert network 3, expert network 4, gate network 1, gate network 2, gate network 3, gate network 4. Input is input to expert network 1 and gate network 1 to obtain a first output and a second output. Multiply the first output and the second output to obtain a first multiplication result. Input is input to expert network 2 and gate network 2 to obtain a third output and a fourth output. Multiply the third output and the fourth output to obtain a second multiplication result. Input is input to expert network 3 and gate network 3 to obtain a fifth output and a sixth output. Multiply the fifth output and the sixth output to obtain a third multiplication result. Input is input to expert network 4 and gate network 4 to obtain a seventh output and an eighth output. Multiply the seventh output and the eighth output to obtain a fourth multiplication result. Finally, the first multiplication result, the second multiplication result, the third multiplication result and the fourth multiplication result are added to obtain an output.

[0057] Step 203: input the above output result into the initial target value borrowing resource numerical prediction model in at least two initial value borrowing resource numerical prediction models to output value borrowing resource numerical prediction information.

[0058] In some embodiments, the execution subject may input the output result into the initial target value borrowing resource numerical prediction model in at least two initial value borrowing resource numerical prediction models to output value borrowing resource numerical prediction information. There is a one-to-one correspondence between the initial target value borrowing resource numerical prediction model and the value resource fallback label. The at least two initial value borrowing resource numerical prediction models are sub-models in the first initial value borrowing resource numerical diversification prediction model.

[0059] In practice, the value borrowing resource numerical prediction information can be the user's new loan amount prediction information. The corresponding initial value borrowing resource numerical prediction model is different for different value resource fallback label contents. For example, the value resource fallback label includes: a label representing a 5% interest rate cut, a label representing a 6% interest rate cut, and a label representing a 7% interest rate cut. Then at least two initial value borrowing resource numerical prediction models include: an initial value borrowing resource numerical prediction model corresponding to a label representing a 5% interest rate cut, an initial value borrowing resource numerical prediction model corresponding to a label representing a 6% interest rate cut, and an initial value borrowing resource numerical prediction model corresponding to a label representing a 7% interest rate cut.

[0060] In some optional implementations of some embodiments, the above-mentioned value resource rollback label is a first label indicating that the value resource has not been rolled back. The above-mentioned at least two initial value borrowing resource numerical prediction models include: a first initial value borrowing resource numerical prediction model for the above-mentioned first label.

[0061] In practice, the first label may be a non-rate reduction label. That is, the first label indicates that no rate reduction operation has been performed on the user corresponding to the training data during the target period. The first initial value borrowing resource numerical prediction model may be a model that generates the user's corresponding predicted value borrowing resource numerical value when no rate reduction operation has been performed on the user corresponding to the training data during the target period. In practice, the first initial value borrowing resource numerical prediction model may be a model that generates the user's corresponding predicted new loan amount when no rate reduction operation has been performed on the user corresponding to the training data during the target period. For example, the first initial value borrowing resource numerical prediction model may be a deep neural network (DNN, Deep-Learning Neural Network) network model.

[0062] Optionally, the execution entity may input the output result into the numerical prediction model for the first initial value borrowing resource to output the numerical prediction information of the value borrowing resource under the condition that the value resource is not withdrawn. In practice, the numerical prediction information of the value borrowing resource under the condition that the value resource is not withdrawn may be the predicted amount of the user's new loan under the condition that the interest rate is not reduced.

[0063] In some optional implementations of some embodiments, the above-mentioned value resource rollback tag is a second tag indicating that the value resource has been rolled back. The above-mentioned at least two initial value borrowing resource numerical prediction models include: a second initial value borrowing resource numerical prediction model for the above-mentioned second tag.

[0064] In practice, the second label may be a label of interest rate cut. That is, the second label represents that an interest rate cut operation is performed on the user corresponding to the training data during the target period. The second initial value borrowing resource numerical prediction model may be a model that generates a user's corresponding predicted value borrowing resource numerical value when an interest rate cut operation is performed on the user corresponding to the training data during the target period. In practice, the second initial value borrowing resource numerical prediction model may be a model that generates a user's corresponding predicted new loan amount when an interest rate cut operation is performed on the user corresponding to the training data during the target period. For example, the second initial value borrowing resource numerical prediction model may also be a deep neural network (DNN, Deep-Learning Neural Network) network model.

[0065] Optionally, the output result is input into the second initial value borrowing resource numerical prediction model to output the value borrowing resource numerical prediction information under the condition that the value resource has been withdrawn. In practice, the value borrowing resource numerical prediction information under the condition that the value resource has not been withdrawn can be the predicted amount of the user's new loan under the condition of interest rate reduction.

[0066] Step 204: Generate a second initial value borrowing resource value diversification prediction model based on the above value borrowing resource value prediction information.

[0067] In some embodiments, the execution entity may generate a second initial value borrowing resource numerical diversification prediction model based on the value borrowing resource numerical prediction information. The second initial value borrowing resource numerical diversification prediction model includes: a hybrid expert network model and a target value borrowing resource numerical prediction model. The hybrid expert network model is a network model obtained by training the initial hybrid expert network model based on the training data. The target value borrowing resource numerical prediction model may be a network model obtained by training the initial target value borrowing resource numerical prediction model based on the training data.

[0068] As an example, the above-mentioned execution entity can determine whether the value-borrowing resource numerical prediction information is in the value-borrowing resource numerical variation interval set for the user corresponding to the training data. In response to determining that it is, the first initial value-borrowing resource numerical diversification prediction model is determined as the second initial value-borrowing resource numerical diversification prediction model. In response to not being in, the difference between the value-borrowing resource numerical prediction information and the two interval boundary values ​​corresponding to the value-borrowing resource numerical transformation interval is determined to obtain a difference set. Then, the minimum difference in the difference set is determined as the loss value. Finally, according to the above-mentioned loss value, the model parameters of the first initial value-borrowing resource numerical diversification prediction model are updated to generate a second initial value-borrowing resource diversification prediction model.

[0069] In some optional implementations of some embodiments, the training data includes: actual value information of the value borrowing resources. The actual value information of the value borrowing resources may be the actual value information of the additional loans of the user during the target period.

[0070] Optionally, the generating of the second initial value borrowing resource numerical value diversified prediction model according to the value borrowing resource numerical value prediction information may include the following steps:

[0071] The first step is to generate loss values ​​for the above-mentioned actual information on the value borrowing resource values ​​and the predicted information on the value borrowing resource values.

[0072] As an example, the execution subject may input the actual value borrowing resource value information and the predicted value borrowing resource value information into a preset loss function to output a loss value. For example, the preset loss function may be a KL divergence and cross entropy loss function.

[0073] In the second step, based on the above loss value, the above initial hybrid expert network model and the above initial target value borrowed resource numerical prediction model are trained to obtain the hybrid expert network model and the target value borrowed resource numerical prediction model.

[0074] As an example, in response to determining that the loss value is less than a preset value, the initial hybrid expert network model and the initial target value borrowed resource numerical prediction model are broadcasted and determined as a hybrid expert network model and a target value borrowed resource numerical prediction model. In response to determining that the loss value is greater than or equal to the preset value, the model parameters of the initial hybrid expert network model and the initial target value borrowed resource numerical prediction model are updated by back propagation to obtain a hybrid expert network model and a target value borrowed resource numerical prediction model.

[0075] The third step is to generate the second initial value borrowing resource numerical diversification prediction model based on the hybrid expert network model and the target value borrowing resource numerical prediction model.

[0076] As an example, the above-mentioned execution entity can replace the initial hybrid expert network model and the initial target value borrowing resource numerical prediction model in the first initial value borrowing resource numerical diversification prediction model with the above-mentioned hybrid expert network model and the above-mentioned target value borrowing resource numerical prediction model to obtain the above-mentioned second initial value borrowing resource numerical diversification prediction model.

[0077] Step 205, in response to determining that the second initial value borrowing resource numerical diversification prediction model meets the predetermined model optimization target, determine the second initial value borrowing resource numerical diversification prediction model as the value borrowing resource numerical diversification prediction model.

[0078] In some embodiments, in response to determining that the second initial value borrowing resource numerical diversification prediction model meets a predetermined model optimization goal, the execution entity may determine the second initial value borrowing resource numerical diversification prediction model as the value borrowing resource numerical diversification prediction model. The model optimization goal may represent the training end goal of the second initial value borrowing resource numerical diversification prediction model.

[0079] In some optional implementations of some embodiments, before step 205, the steps further include:

[0080] The first step is to determine the data round corresponding to the data set to which the above training data belongs.

[0081] The data round may be the batch information of the data set to which the training data belongs.

[0082] In the second step, in response to determining that the data rounds have reached a preset round, determining that the second initial value borrowing resource numerical diversification prediction model meets the predetermined model optimization target. For example, the preset batch can be 50 times.

[0083] Optionally, in response to determining that the data round does not reach a preset round, it is determined that the second initial value borrowing resource numerical diversification prediction model does not meet the predetermined model optimization target.

[0084] In some optional implementations of some embodiments, after step 205, the steps further include:

[0085] In the first step, in response to determining that the second initial value borrowing resource numerical diversification prediction model does not meet the predetermined model optimization goal, the training data is re-acquired as the target training data.

[0086] The second step is to continue model training for the second initial value borrowing resource numerical diversification prediction model based on the target training data. Here, the implementation method of training the second initial value borrowing resource numerical diversification prediction model based on the target training data can refer to the implementation method of training the first initial value borrowing resource numerical diversification prediction model based on the training data. No further details will be given here.

[0087] The above-mentioned embodiments of the present disclosure have the following beneficial effects: through the model training method of some embodiments of the present disclosure, using the value borrowing resource numerical diversification prediction model, more accurate value borrowing resource numerical information can be generated. Specifically, the reason why the relevant value borrowing resource numerical information is not accurate is that: first, for the One-model (single model) in the gain model, when there are many features, the ability to distinguish the intervention action (treatment) feature is lower than the other features, resulting in a lower weight of the intervention action feature in the One-model, which greatly affects the prediction of the subsequent value borrowing resource numerical information, resulting in low prediction accuracy; second, for the Two-model (dual model) in the gain model, during the training process of the Two-model, the sample utilization is restricted (that is, the two models in the dual model cannot use each other's samples), resulting in the existence of cumulative errors in the Two-model, so that the prediction accuracy of the subsequent value borrowing resource numerical information is low. Based on this, the model training method of some embodiments of the present disclosure, first, obtains the value borrowing resource exchange data as training data. Among them, the above training data has a corresponding value resource fallback label. Here, the training data is obtained for the subsequent initial value borrowing resource numerical diversification prediction model training. By determining the value resource fallback label corresponding to the training data, the corresponding initial value borrowing resource numerical prediction model is subsequently used to select the corresponding initial value borrowing resource numerical prediction model to output the value borrowing resource numerical prediction information under the corresponding model. Then, the above training data is input into the initial hybrid expert network model included in the first initial value borrowing resource numerical diversification prediction model to generate an output result corresponding to the accurate and targeted features. Next, the above output result is input into the initial target value borrowing resource numerical prediction model in at least two initial value borrowing resource numerical prediction models to output the value borrowing resource numerical prediction information. Among them, there is a one-to-one correspondence between the above initial target value borrowing resource numerical prediction model and the above value resource fallback label. Among them, the above at least two initial value borrowing resource numerical prediction models are sub-models in the above first initial value borrowing resource numerical diversification prediction model. Here, the corresponding initial value borrowing resource numerical prediction model is selected through the value resource fallback label to output the corresponding value borrowing resource numerical prediction information, and the model parameters of the initial hybrid expert network model and the selected corresponding initial value borrowing resource numerical prediction model are updated in a targeted manner. This makes not only the selected corresponding initial value borrowing resource numerical prediction model pay more attention to the feature information corresponding to the value resource fallback label, but also makes full use of the feature information of the training data (that is, for the training data corresponding to different value resource fallback labels, the initial hybrid expert network model also needs to be trained), avoiding the problems existing in One-model and Two-model.Furthermore, according to the above-mentioned value borrowing resource numerical prediction information, a second initial value borrowing resource numerical diversification prediction model is generated. Among them, the above-mentioned second initial value borrowing resource numerical diversification prediction model includes: a hybrid expert network model and a target value borrowing resource numerical prediction model. Here, compared with the first initial value borrowing resource numerical diversification prediction model, the obtained second initial value borrowing resource numerical diversification prediction model is more accurate in predicting the value borrowing resource numerical prediction information under the value resource fallback label. Finally, in response to determining that the above-mentioned second initial value borrowing resource numerical diversification prediction model meets the predetermined model optimization goal, the above-mentioned second initial value borrowing resource numerical diversification prediction model is determined as the value borrowing resource numerical diversification prediction model. In summary, by using the value resource fallback label, the corresponding initial value borrowing resource numerical prediction model in the first initial value borrowing resource numerical diversification prediction model is selected to train the subsequent model, which can not only make full use of the data features of the training data, but also avoid the problem that the weight of the intervention action (i.e., the feature corresponding to the value resource fallback label) feature in the One-model is low, resulting in low prediction accuracy. Therefore, by using the value-borrowing resource numerical diversification prediction model, more accurate value-borrowing resource numerical information can be generated.

[0088] Further references Figure 4 , shows a process 400 of other embodiments of the model training method according to the present disclosure. The model training method comprises the following steps:

[0089] Step 401, obtaining value borrowing resource exchange data as training data.

[0090] Step 402: input the above training data into the initial hybrid expert network model included in the first initial value borrowing resource numerical diversification prediction model to generate an output result.

[0091] Step 403: input the above output result into the initial target value borrowing resource numerical prediction model in at least two initial value borrowing resource numerical prediction models to output value borrowing resource numerical prediction information.

[0092] Step 404: Generate a second initial value borrowing resource value diversification prediction model based on the above value borrowing resource value prediction information.

[0093] In some embodiments, the specific implementation of steps 401-404 and the technical effects thereof can be referred to in Figure 2 The steps 201-204 in the corresponding embodiment are not described in detail here.

[0094] Step 405: Obtain a verification data set.

[0095] In some embodiments, an execution entity (e.g. Figure 1 The electronic device 101 shown in the figure can obtain a verification data set. The verification data set is a data set used to verify the corresponding model performance of the second initial value borrowing resource numerical diversification prediction model.

[0096] Step 406: Generate at least one data indicator information for the verification data set by using the second initial value borrowing resource numerical diversification prediction model.

[0097] In some embodiments, the execution subject may use the second initial value borrowing resource numerical diversification prediction model to generate at least one data indicator information for the verification data set. In practice, the at least one data indicator may include: accuracy and recall. The data indicator information is the value of the data indicator.

[0098] As an example, first, the execution subject may input the verification data set into the second initial value borrowing resource numerical diversification prediction model to output the value borrowing numerical prediction information set. Then, according to the value borrowing numerical prediction information set, the corresponding accuracy and recall rate are generated.

[0099] Step 407, based on the at least one data indicator information, determine whether the second initial value borrowing resource numerical diversification prediction model meets the predetermined model optimization goal.

[0100] In some embodiments, the execution entity may determine whether the second initial value borrowing resource numerical diversification prediction model meets the predetermined model optimization goal based on the at least one data indicator information.

[0101] As an example, for each data indicator information in at least one data indicator information, first, determine whether the data indicator information is greater than a preset indicator value to obtain a determination result. In response to determining that at least one determination result corresponding to at least one data indicator information indicates that the corresponding indicator is greater than or equal to the corresponding preset indicator value, it is determined that the second initial value borrowing resource numerical diversification prediction model meets the above-mentioned predetermined model optimization target. In response to determining that any determination result in at least one determination result corresponding to at least one data indicator information indicates that the corresponding indicator is less than the corresponding preset indicator value, it is determined that the second initial value borrowing resource numerical diversification prediction model does not meet the above-mentioned predetermined model optimization target.

[0102] As another example, first, at least one historical data indicator information sequence corresponding to at least one data indicator is obtained. Then, for each data indicator, the corresponding data indicator information in the at least one data indicator information is greater than each indicator information in the corresponding historical data indicator information sequence, and the data indicator optimal information is generated, wherein the data indicator optimal information represents that the data indicator information is the optimal value. Finally, in response to determining that each data indicator information corresponds to the data indicator optimal information, it is determined that the second initial value borrowing resource numerical diversification prediction model meets the predetermined model optimization target.

[0103] Step 408, in response to determining that the second initial value borrowing resource numerical diversification prediction model meets the predetermined model optimization target, determine the second initial value borrowing resource numerical diversification prediction model as the value borrowing resource numerical diversification prediction model.

[0104] In some embodiments, the specific implementation of step 408 and the technical effects thereof can be referred to in Figure 2 The corresponding step 205 in the embodiment will not be described in detail here.

[0105] from Figure 4 It can be seen that Figure 2 Compared with the description of some corresponding embodiments, Figure 4 In the corresponding process 300 of the model training method in some embodiments, the model performance of the second initial value borrowing resource numerical diversification prediction model is evaluated by verifying at least one data indicator information corresponding to the data set, so as to obtain a value borrowing resource numerical diversification prediction model that generates more accurate value borrowing resource values.

[0106] Further references Figure 5 , shows a process 500 of another embodiment of the method for generating operation information according to the present disclosure. The method for generating operation information includes the following steps:

[0107] Step 501: Obtain user attribute information of a target user.

[0108] In some embodiments, the execution subject (e.g., electronic device) may obtain user attribute information of the target user. The user attribute information may be attribute information of user attributes. In practice, the user attributes may include but are not limited to at least one of the following: user name, user address, user deposit, user loan record.

[0109] Step 502: input the above user attribute information into a pre-trained value borrowing resource numerical diversification prediction model to output at least two value borrowing resource numerical prediction information.

[0110] In some embodiments, the execution subject may input the user attribute information into a pre-trained value borrowing resource numerical diversification prediction model to output at least two value borrowing resource numerical prediction information. Each value borrowing resource numerical prediction information has a corresponding value resource fallback label. The value borrowing resource numerical diversification prediction model is generated by the above-mentioned method of the model training method disclosed in the present invention.

[0111] Step 503: Generate operation information indicating whether to perform a value resource rollback operation on the target user based on the at least two value borrowing resource value prediction information.

[0112] In some embodiments, the execution subject may generate operation information indicating whether to perform a value resource rollback operation on the target user based on the at least two value borrowing resource numerical prediction information.

[0113] As an example, first, the above-mentioned execution entity can determine the value borrowing resource numerical prediction information with the largest value and the value borrowing resource numerical prediction information with the smallest value among at least two value borrowing resource numerical prediction information as the maximum value borrowing resource numerical prediction information and the minimum value borrowing resource numerical prediction information. Then, the maximum value borrowing resource numerical prediction information is subtracted from the minimum value borrowing resource numerical prediction information to obtain the subtraction value. Finally, in response to determining that the subtraction value is in a predetermined interval, operation information representing the value resource rollback operation for the above-mentioned target user is generated. In response to determining that the subtraction value is not in the predetermined interval, operation information representing that the value resource rollback operation is not performed on the above-mentioned target user is generated.

[0114] The above-mentioned embodiments of the present disclosure have the following beneficial effects: through the model training method of some embodiments of the present disclosure, using the value borrowing resource numerical diversification prediction model, more accurate operation information can be generated. Based on the operation information, the value resource rollback operation can be performed to obtain more value information.

[0115] Further references Figure 6 As an implementation of the methods shown in the above figures, the present disclosure provides some embodiments of a model training device, and these device embodiments are Figure 2 Corresponding to the method embodiments shown, the model training device can be specifically applied to various electronic devices.

[0116] like Figure 6As shown, a model training device 600 includes: a first acquisition unit 601, a first input unit 602, a second input unit 603, a first generation unit 604 and a determination unit 605. The first acquisition unit 601 is configured to acquire value borrowing resource exchange data as training data, wherein the above training data has a corresponding value resource fallback label; the first input unit 602 is configured to input the above training data into an initial hybrid expert network model included in a first initial value borrowing resource numerical diversification prediction model to generate an output result; the second input unit 603 is configured to input the above output result into an initial target value borrowing resource numerical prediction model in at least two initial value borrowing resource numerical prediction models to output value borrowing resource numerical prediction information, wherein the above initial target value borrowing resource numerical prediction model has a one-to-one correspondence with the above value resource fallback label. , the above-mentioned at least two initial value borrowing resource numerical prediction models are sub-models in the above-mentioned first initial value borrowing resource numerical diversification prediction model; the first generation unit 604 is configured to generate a second initial value borrowing resource numerical diversification prediction model according to the above-mentioned value borrowing resource numerical prediction information, wherein the above-mentioned second initial value borrowing resource numerical diversification prediction model includes: a hybrid expert network model and a target value borrowing resource numerical prediction model; the determination unit 605 is configured to determine the above-mentioned second initial value borrowing resource numerical diversification prediction model as the value borrowing resource numerical diversification prediction model in response to determining that the above-mentioned second initial value borrowing resource numerical diversification prediction model meets the predetermined model optimization target.

[0117] In some optional implementations of some embodiments, the above-mentioned value resource fallback label is a first label that characterizes that the value resource has not been fallen back, and the above-mentioned at least two initial value borrowing resource numerical prediction models include: a first initial value borrowing resource numerical prediction model for the above-mentioned first label; and the first input unit 602 in the above-mentioned model training device 600 can be configured to: input the above-mentioned output result into the above-mentioned first initial value borrowing resource numerical prediction model to output the value borrowing resource numerical prediction information under the condition that the value resource has not been fallen back.

[0118] In some optional implementations of some embodiments, the above-mentioned value resource fallback label is a second label that characterizes that the value resource has been fallen back, and the above-mentioned at least two initial value borrowing resource numerical prediction models include: a second initial value borrowing resource numerical prediction model for the above-mentioned second label; and the first input unit 602 in the above-mentioned model training device 600 can be configured to: input the above-mentioned output result into the above-mentioned second initial value borrowing resource numerical prediction model to output the value borrowing resource numerical prediction information under the condition that the value resource has been fallen back.

[0119] In some optional implementations of some embodiments, the above-mentioned training data includes: actual information on the numerical values ​​of value-borrowed resources; and the first generation unit 604 in the above-mentioned model training device 600 can be configured to: generate loss values ​​for the above-mentioned actual information on the numerical values ​​of value-borrowed resources and the numerical prediction information on value-borrowed resources; according to the above-mentioned loss values, perform model training on the above-mentioned initial hybrid expert network model and the above-mentioned initial target value-borrowed resource numerical prediction model to obtain a hybrid expert network model and a target value-borrowed resource numerical prediction model; based on the above-mentioned hybrid expert network model and the above-mentioned target value-borrowed resource numerical prediction model, generate the above-mentioned second initial value-borrowed resource numerical diversification prediction model.

[0120] In some optional implementations of some embodiments, the model training device 600 further includes: a third acquisition unit, a third generation unit and a first model determination unit (not shown in the figure). The third acquisition unit may be configured to: acquire a verification data set. The third generation unit may be configured to: generate at least one data indicator information for the verification data set using the second initial value borrowing resource numerical diversification prediction model. The first model determination unit may be configured to: determine whether the second initial value borrowing resource numerical diversification prediction model meets the predetermined model optimization goal based on the at least one data indicator information.

[0121] In some optional implementations of some embodiments, the model training device 600 further includes: a data round determination unit and a second model determination unit (not shown in the figure). The data round determination unit may be configured to: determine the data round corresponding to the data set to which the training data belongs. The second model determination unit may be configured to: in response to determining that the data round reaches a preset round, determine that the second initial value borrowing resource numerical diversification prediction model meets the predetermined model optimization target.

[0122] In some optional implementations of some embodiments, the model training device 600 further includes: a re-acquisition unit and a model training unit (not shown in the figure). The re-acquisition unit may be configured to: in response to determining that the second initial value borrowing resource numerical diversification prediction model does not meet the predetermined model optimization target, re-acquire training data as target training data. The model training unit may be configured to: continue model training for the second initial value borrowing resource numerical diversification prediction model according to the target training data.

[0123] It is understandable that the units recorded in the model training device 600 are similar to the reference Figure 2Therefore, the operations, features and beneficial effects described above for the method are also applicable to the model training device 600 and the units contained therein, and will not be repeated here.

[0124] Further references Figure 7 As an implementation of the methods shown in the above figures, the present disclosure provides some embodiments of an operation information generating device, and these device embodiments are similar to Figure 5 Corresponding to the method embodiments shown, the operation information generating device can be specifically applied to various electronic devices.

[0125] like Figure 7 As shown, an operation information generating device 700 includes: a second acquisition unit 701, a third input unit 702 and a second generation unit 703. The second acquisition unit 701 is configured to acquire user attribute information of a target user; the third input unit 702 is configured to input the above user attribute information into a pre-trained value borrowing resource numerical diversification prediction model to output at least two value borrowing resource numerical prediction information, wherein each value borrowing resource numerical prediction information has a corresponding value resource fallback label, and the above value borrowing resource numerical diversification prediction model is generated based on the model training method disclosed in the present invention by the above method; the second generation unit 703 is configured to generate operation information representing whether to perform a value resource fallback operation on the above target user according to the above at least two value borrowing resource numerical prediction information.

[0126] It is understandable that the units recorded in the operation information generating device 700 are similar to those in the reference Figure 5 Therefore, the operations, features and beneficial effects described above for the method are also applicable to the operation information generating device 700 and the units included therein, and will not be described in detail here.

[0127] Reference below Figure 8 , which shows an electronic device (eg, Figure 1 Schematic diagram of the structure of the electronic device 101)800. Figure 8 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0128] like Figure 8As shown, the electronic device 800 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 801, which can perform various appropriate actions and processes according to a program stored in a read-only memory 802 or a program loaded from a storage device 808 into a random access memory 803. Various programs and data required for the operation of the electronic device 800 are also stored in the random access memory 803. The processing device 801, the read-only memory 802, and the random access memory 803 are connected to each other via a bus 804. An input / output interface 805 is also connected to the bus 804.

[0129] Typically, the following devices may be connected to the input / output interface 805: an input device 806 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 807 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 808 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 809. The communication device 809 may allow the electronic device 800 to communicate with other devices wirelessly or by wire to exchange data. Although Figure 8 The electronic device 800 is shown with various devices, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead. Figure 8 Each block shown in the figure may represent one device, or may represent multiple devices as required.

[0130] In particular, according to some embodiments of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, some embodiments of the present disclosure include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In some such embodiments, the computer program can be downloaded and installed from a network through a communication device 809, or installed from a storage device 808, or installed from a read-only memory 802. When the computer program is executed by the processing device 801, the above-mentioned functions defined in the method of some embodiments of the present disclosure are executed.

[0131] It should be noted that the computer-readable medium in some embodiments of the present disclosure may be a computer-readable signal medium or a computer-readable storage medium or any combination of the two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In some embodiments of the present disclosure, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In some embodiments of the present disclosure, the computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, in which a computer-readable program code is carried. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer readable signal medium may also be any computer readable medium other than a computer readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0132] In some embodiments, the client and the server may communicate using any currently known or future developed network protocol such as HTTP (HyperText Transfer Protocol), and may be interconnected with any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), an internet (e.g., the Internet), and a peer-to-peer network (e.g., an ad hoc peer-to-peer network), as well as any currently known or future developed network.

[0133] The above-mentioned computer-readable medium may be included in the above-mentioned electronic device; or it may exist independently without being installed in the electronic device. The above-mentioned computer-readable medium carries one or more programs. When the above-mentioned one or more programs are executed by the electronic device, the electronic device: obtains value borrowing resource exchange data as training data, wherein the above-mentioned training data has corresponding value resource fallback labels; inputs the above-mentioned training data into the initial hybrid expert network model included in the first initial value borrowing resource numerical diversification prediction model to generate an output result; inputs the above-mentioned output result into the initial target value borrowing resource numerical prediction model among at least two initial value borrowing resource numerical prediction models to output value borrowing resource numerical prediction information, wherein the above-mentioned initial target value borrowing resource numerical prediction model is consistent with the above-mentioned value borrowing resource numerical prediction model; There is a one-to-one correspondence between the value resource fallback labels, and the at least two initial value borrowing resource numerical prediction models are sub-models in the first initial value borrowing resource numerical diversification prediction model; according to the value borrowing resource numerical prediction information, a second initial value borrowing resource numerical diversification prediction model is generated, wherein the second initial value borrowing resource numerical diversification prediction model includes: a hybrid expert network model and a target value borrowing resource numerical prediction model; in response to determining that the second initial value borrowing resource numerical diversification prediction model meets the predetermined model optimization target, the second initial value borrowing resource numerical diversification prediction model is determined as the value borrowing resource numerical diversification prediction model.

[0134] Computer program code for performing the operations of some embodiments of the present disclosure may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0135] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0136] The units described in some embodiments of the present disclosure may be implemented by software or by hardware. The described units may also be provided in a processor, for example, may be described as: a processor comprising a first acquisition unit, a first input unit, a second input unit, a first generation unit, and a determination unit. The names of these units do not, in some cases, constitute limitations on the units themselves, for example, the first acquisition unit may also be described as a "unit for acquiring value borrowing resource exchange data as training data".

[0137] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), and the like.

[0138] Some embodiments of the present disclosure also provide a computer program product, including a computer program, which implements any of the above-mentioned model training methods and operation information generation methods when executed by a processor.

[0139] The above descriptions are only some preferred embodiments of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present disclosure is not limited to the technical solutions formed by a specific combination of the above-mentioned technical features, but should also cover other technical solutions formed by any combination of the above-mentioned technical features or their equivalent features without departing from the above-mentioned inventive concept. For example, the above-mentioned features are replaced with the technical features with similar functions disclosed in the embodiments of the present disclosure (but not limited to) and the technical solutions formed.

Claims

1. A model training method, comprising: Acquire value borrowing resource exchange data as training data, wherein the training data has a corresponding value resource fallback label; Inputting the training data into an initial hybrid expert network model included in the first initial value borrowing resource numerical diversification prediction model to generate an output result; Inputting the output result into an initial target value borrowing resource numerical prediction model among at least two initial value borrowing resource numerical prediction models to output value borrowing resource numerical prediction information, wherein there is a one-to-one correspondence between the initial target value borrowing resource numerical prediction model and the value resource fallback label, and the at least two initial value borrowing resource numerical prediction models are sub-models in the first initial value borrowing resource numerical diversification prediction model; According to the value borrowing resource numerical prediction information, a second initial value borrowing resource numerical diversification prediction model is generated, wherein the second initial value borrowing resource numerical diversification prediction model includes: a hybrid expert network model and a target value borrowing resource numerical prediction model; In response to determining that the second initial value-borrowing resource numerical diversification prediction model satisfies a predetermined model optimization target, the second initial value-borrowing resource numerical diversification prediction model is determined as the value-borrowing resource numerical diversification prediction model.

2. The method according to claim 1, wherein: The value resource rollback tag is a first tag indicating that the value resource has not been rolled back, and the at least two initial value borrowing resource numerical prediction models include: a first initial value borrowing resource numerical prediction model for the first tag; and The step of inputting the output result into an initial target value borrowing resource numerical prediction model of at least two initial value borrowing resource numerical prediction models to output value borrowing resource numerical prediction information includes: The output result is input into the first initial value borrowing resource numerical prediction model to output the value borrowing resource numerical prediction information under the condition that the value resource is not rolled back.

3. The method according to claim 1, wherein: The value resource rollback tag is a second tag indicating that the value resource has been rolled back, and the at least two initial value borrowing resource numerical prediction models include: a second initial value borrowing resource numerical prediction model for the second tag; and The step of inputting the output result into an initial target value borrowing resource numerical prediction model of at least two initial value borrowing resource numerical prediction models to output value borrowing resource numerical prediction information includes: The output result is input into the second initial value borrowing resource numerical prediction model to output the value borrowing resource numerical prediction information under the condition that the value resource has been withdrawn.

4. The method according to claim 1, wherein: The training data includes: actual information on the value of the borrowed resources; and The step of generating a second initial value borrowing resource numerical diversification prediction model according to the value borrowing resource numerical prediction information includes: Generate a loss value for the actual value information of the value borrowing resource and the predicted value information of the value borrowing resource; According to the loss value, the initial hybrid expert network model and the initial target value borrowed resource numerical prediction model are trained to obtain a hybrid expert network model and a target value borrowed resource numerical prediction model; The second initial value borrowing resource numerical diversification prediction model is generated according to the hybrid expert network model and the target value borrowing resource numerical prediction model.

5. The method according to claim 1, wherein: Before determining the second initial value borrowing resource numerical diversification prediction model as the value borrowing resource numerical diversification prediction model in response to determining that the second initial value borrowing resource numerical diversification prediction model meets a predetermined model optimization target, the method further includes: Get the validation dataset; Generate at least one data indicator information for the verification data set by using the second initial value borrowing resource numerical diversification prediction model; According to the at least one data indicator information, determine whether the second initial value borrowing resource numerical diversification prediction model meets the predetermined model optimization goal.

6. The method according to claim 1, wherein: Before determining the second initial value borrowing resource numerical diversification prediction model as the value borrowing resource numerical diversification prediction model in response to determining that the second initial value borrowing resource numerical diversification prediction model meets a predetermined model optimization target, the method further includes: Determine the data round corresponding to the data set to which the training data belongs; In response to determining that the data round reaches a preset round, it is determined that the second initial value borrowing resource numerical diversification prediction model meets the predetermined model optimization goal.

7. The method according to claim 1, wherein: The method further comprises: In response to determining that the second initial value borrowing resource numerical diversification prediction model does not meet a predetermined model optimization target, reacquiring training data as target training data; According to the target training data, the model training of the second initial value borrowing resource numerical diversification prediction model continues.

8. A method for generating operation information, comprising: Obtain user attribute information of the target user; Inputting the user attribute information into a pre-trained value borrowing resource numerical diversification prediction model to output at least two value borrowing resource numerical prediction information, wherein each value borrowing resource numerical prediction information has a corresponding value resource fallback label, and the value borrowing resource numerical diversification prediction model is generated based on the method of one of claims 1 to 7; Based on the at least two value borrowing resource numerical prediction information, operation information indicating whether to perform a value resource rollback operation on the target user is generated.

9. A model training device, comprising: A first acquisition unit is configured to acquire value borrowing resource exchange data as training data, wherein the training data has a corresponding value resource fallback tag; A first input unit is configured to input the training data into an initial hybrid expert network model included in the first initial value borrowing resource numerical diversification prediction model to generate an output result; a second input unit configured to input the output result into an initial target value borrowing resource numerical prediction model in at least two initial value borrowing resource numerical prediction models to output value borrowing resource numerical prediction information, wherein there is a one-to-one correspondence between the initial target value borrowing resource numerical prediction model and the value resource fallback label, and the at least two initial value borrowing resource numerical prediction models are sub-models in the first initial value borrowing resource numerical diversification prediction model; A first generating unit is configured to generate a second initial value borrowing resource numerical diversification prediction model according to the value borrowing resource numerical prediction information, wherein the second initial value borrowing resource numerical diversification prediction model includes: a hybrid expert network model and a target value borrowing resource numerical prediction model; The determination unit is configured to determine the second initial value-borrowing resource numerical diversification prediction model as the value-borrowing resource numerical diversification prediction model in response to determining that the second initial value-borrowing resource numerical diversification prediction model meets a predetermined model optimization target.

10. An information generating device, comprising: A second acquisition unit is configured to acquire user attribute information of a target user; a third input unit configured to input the user attribute information into a pre-trained value borrowing resource numerical diversification prediction model to output at least two value borrowing resource numerical prediction information, wherein each value borrowing resource numerical prediction information has a corresponding value resource fallback label, and the value borrowing resource numerical diversification prediction model is generated based on the method of one of claims 1 to 7; The second generating unit is configured to generate operation information indicating whether to perform a value resource rollback operation on the target user according to the at least two value borrowing resource numerical prediction information.

11. An electronic device, comprising: one or more processors; a storage device having one or more programs stored thereon, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 8.

12. A computer readable medium having a computer program stored thereon, wherein: When the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

13. A computer program product, comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 8.

Images