Liability assessment model, liability assessment result generation method, device, equipment and medium

Abstract

This disclosure presents embodiments of a method for generating a liability assessment model, a method for generating liability assessment results, an apparatus, an electronic device, and a computer-readable medium. One specific implementation of the liability assessment model generation method includes: selecting at least one meta-model from a pre-trained meta-model set based on at least one sub-business scenario information corresponding to business scenario information, wherein each meta-model corresponds to one sub-business scenario information; performing transfer analysis on the at least one meta-model to obtain an initial liability assessment model; and training the initial liability assessment model based on a training sample set to obtain a liability assessment model for judging the liability of both parties corresponding to an order corresponding to the business scenario information, thereby obtaining a liability assessment result. The training samples in the training sample set include order features and order annotation information. This implementation achieves rapid generation of the liability assessment model.

Description

Technical Field

[0001] The embodiments of this disclosure relate to the field of computer technology, specifically to a method for generating a judgment model, a method for generating a judgment result, an apparatus, an electronic device, and a computer-readable medium. Background Technology

[0002] With the development of internet technology, various applications have emerged. Ride-hailing and other service applications have greatly facilitated people's travel. As the number of users increases, the number of orders also grows. At the same time, the number of order cancellations due to various reasons is also increasing. This increasing number of cancellations results in a significant waste of resources. Summary of the Invention

[0003] The summary section of this disclosure provides a brief overview of concepts that will be described in detail in the subsequent detailed description section. This summary section is not intended to identify key or essential features of the claimed technical solutions, nor is it intended to limit the scope of the claimed technical solutions. Some embodiments of this disclosure provide methods, apparatus, electronic devices, and computer-readable media for generating liability models to address one or more of the technical problems mentioned in the background section above.

[0004] In a first aspect, some embodiments of this disclosure provide a method for generating a liability determination model. The method includes: selecting at least one meta-model from a pre-trained meta-model set based on at least one sub-business scenario information corresponding to business scenario information, wherein each meta-model corresponds to one sub-business scenario information; performing transfer on the at least one meta-model to obtain an initial liability determination model; and training the initial liability determination model based on a training sample set to obtain a liability determination model for use in determining liability between the parties corresponding to an order corresponding to the business scenario information, thereby obtaining a liability determination result. The training samples in the training sample set include order features and order annotation information.

[0005] Secondly, some embodiments of this disclosure provide a responsibility judgment model generation apparatus, the apparatus comprising: a selection unit configured to select at least one meta-model from a pre-trained meta-model set based on at least one sub-business scenario information corresponding to business scenario information, wherein each meta-model corresponds to one sub-business scenario information; a transfer unit configured to transfer the at least one meta-model to obtain an initial responsibility judgment model; and a training unit configured to train the initial responsibility judgment model based on a training sample set to obtain a responsibility judgment model for use in judging the responsibilities of both parties corresponding to an order corresponding to the business scenario information, thereby obtaining a responsibility judgment result, wherein the training samples in the training sample set include order features of the order and annotation information of the order.

[0006] Thirdly, some embodiments of this disclosure provide a method for outputting a judgment result, including: extracting order features of a target order; inputting the order features into a judgment model to obtain a judgment result, wherein the judgment model is generated according to the method described in any of the implementations of the first aspect above.

[0007] Fourthly, some embodiments of this disclosure provide a judgment result generation apparatus, including: an extraction unit configured to extract order features of a target order; and a judgment result generation unit configured to input the order features into a judgment model to obtain a judgment result, wherein the judgment model is generated according to the method described in any implementation of the first aspect above.

[0008] Fifthly, some embodiments of this disclosure provide an electronic device, including: one or more processors; and a storage device having one or more programs stored thereon, wherein 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 implementation of the first aspect above.

[0009] Sixthly, some embodiments of this disclosure provide a computer-readable medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the method described in any of the implementations of the first aspect above.

[0010] The embodiments disclosed above have the following beneficial effects: This disclosure provides a method for generating a responsibility assessment model, which can achieve rapid generation of the model. Specifically, this disclosure decomposes a business scenario into multiple sub-business scenarios and selects at least one meta-model based on the sub-business scenario information, thereby realizing model building based on business scenario decomposition. On this basis, the responsibility assessment model is finally obtained through meta-model transfer and training. Since different business scenarios often have overlapping sub-business scenarios, the meta-model can be reused, avoiding the long model training time caused by independently building models for different business scenarios. Furthermore, since the meta-model is pre-trained, the training time can be further shortened. Attached Figure Description

[0011] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and elements are not necessarily drawn to scale.

[0012] Figure 1 This is a flowchart of some embodiments of the liability determination model generation method according to this disclosure;

[0013] Figure 2 This is an exemplary generation flowchart of the metamodel in the liability assessment model generation method disclosed herein;

[0014] Figure 3 This is an exemplary structural diagram of the metamodel in the liability assessment model generation method disclosed herein;

[0015] Figure 4 This is a flowchart of some embodiments of the method for generating judgment results according to this disclosure;

[0016] Figure 5 These are schematic diagrams illustrating the structure of some embodiments of the liability model generation apparatus according to this disclosure;

[0017] Figure 6 These are schematic diagrams of some embodiments of the apparatus for generating judgment results according to this disclosure;

[0018] Figure 7 This is a schematic diagram of the structure of an electronic device suitable for implementing some embodiments of the present disclosure. Detailed Implementation

[0019] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0020] It should also be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings. Unless otherwise specified, the embodiments and features described in this disclosure can be combined with each other.

[0021] It should be noted that the concepts of "first" and "second" mentioned in this disclosure are used only to distinguish different devices, modules or units, and are not used to limit the order of functions performed by these devices, modules or units or their interdependencies.

[0022] It should be noted that the terms "a" and "a plurality of" used in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0023] The names of messages or information exchanged between multiple devices in the embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.

[0024] The method for generating a liability assessment model or a liability assessment result provided in this disclosure can be applied to the business scenario of ride-hailing services. In addition, the method for generating a liability assessment model or a liability assessment result provided in this disclosure can also be applied to other existing or future business scenarios, and this disclosure does not limit such applications.

[0025] To address one or more of the problems described in the background section, when an order is cancelled...

[0026] First, the platform needs to determine who is responsible for the order cancellation and then take appropriate action to minimize the number or proportion of order cancellations and reduce waste of resources.

[0027] When determining liability, rule-based or machine learning models are typically used. However, these methods often present the following technical challenges: First, since an order may fall under multiple scenarios, multi-class classification is unsuitable. Machine learning-based liability determination requires building a model for each scenario, resulting in a large number of models and long training times. Second, rule-based methods cannot fully cover all scenarios, leading to a decrease in recall.

[0028] Secondly, research has found that accurately informing both parties of the reasons for being held liable can help them avoid similar situations in the future, thereby reducing the number or proportion of order cancellations.

[0029] Based on this, on the one hand, this disclosure provides a method for generating a liability assessment model to shorten the model training time and achieve rapid model deployment.

[0030] On the other hand, this disclosure provides a method for generating liability assessment results to accurately interpret the results, thereby reducing the number or proportion of order cancellations.

[0031] This disclosure will now be described in detail with reference to the accompanying drawings and embodiments.

[0032] First refer to Figure 1 The diagram illustrates a flow 100 of some embodiments of the accountability model generation method according to this disclosure. The accountability model generation method includes the following steps:

[0033] Step 101: Select at least one meta-model from the pre-trained meta-model set based on at least one sub-business scenario information corresponding to the business scenario information.

[0034] In some embodiments, the executing entity of the accountability model generation method may first determine at least one sub-business scenario information corresponding to the business scenario information. Then, for each sub-business scenario information, a meta-model corresponding to that sub-business scenario information is selected from a pre-trained meta-model set. This results in at least one meta-model. The executing entity of the accountability model generation method can be various electronic devices. In practice, the aforementioned executing entity can determine at least one sub-business scenario information corresponding to the business scenario information by receiving manually input split configuration information or by querying relevant correspondence tables. The business scenario information is used to represent a business scenario, which may be accountability for canceled orders, accountability for complained orders, etc.

[0035] In practice, for a business scenario represented by business scenario information, the business scenario can be broken down into at least one sub-business scenario based on its key points. The judgment result of each sub-business scenario will affect the judgment result of the original business scenario. In this case, the business scenario information can correspond to at least one sub-business scenario information used to represent the at least one sub-business scenario obtained from the breakdown.

[0036] In some embodiments, multiple meta-models can be pre-trained. Each meta-model corresponds to a sub-business scenario information and is used to determine the judgment result under the sub-business scenario represented by that sub-business scenario information. Furthermore, if a meta-model corresponds to a sub-business scenario information, then it can be understood that the meta-model and the sub-business scenario represented by that sub-business scenario information also correspond.

[0037] For example, the business scenario represented by the business scenario information could be a ride-hailing cancellation liability assessment scenario, that is, a scenario where liability is determined for the driver and passenger when an order is cancelled. In practice, the ride-hailing cancellation liability assessment scenario can be broken down into multiple sub-business scenarios as needed, including but not limited to: driver-induced cancellation, driver overcharging, passenger overloading, passenger carrying pets, etc. The sub-business scenario of driver-induced cancellation can correspond to a meta-model, which can be used to determine whether a driver-induced cancellation has occurred, i.e., the liability assessment result under this sub-business scenario. Similarly, other sub-business scenarios also correspond to a meta-model. These sub-business scenarios may all affect the liability assessment result for ride-hailing cancellations. In practice, these meta-models can be various forms of models such as trained artificial classification networks and decision trees. As an example, the meta-model can be an artificial neural network used for classification, including but not limited to DNN (Deep Neural Networks), TextCNN (Text Convolutional Network), HAN (Heterogeneous Graph Attention Network), BERT, etc. The inputs to these meta-models can be order features, and the outputs can be the judgment results under sub-business scenarios.

[0038] It's important to note that the pre-trained meta-models are not designed for a specific business scenario. Generally, these meta-models are applicable to a variety of different business scenarios. For example, the meta-model for driver overcharging can appear in both order cancellation and complaint handling scenarios. Therefore, once these meta-models are generated, they can be reused in multiple business scenarios, allowing for the rapid addition of new business scenarios.

[0039] Step 102: Transfer at least one meta-model to obtain an initial liability judgment model.

[0040] In some embodiments, the aforementioned execution entity can transfer at least one meta-model, including but not limited to splitting, reorganizing, or introducing other network structures into the at least one meta-model to obtain an initial accountability model. For example, as needed, partial network structures can be extracted from each meta-model and assembled to obtain the initial accountability model. As an example, each assembled network structure receives input order features and then obtains its own output. It should be noted that since the extracted partial network structures may be intermediate layers of the meta-model, their outputs are generally the outputs of intermediate layers and often differ from the outputs of the meta-model. Based on this, the outputs of each network structure can be fused as needed, and the fused results can be classified to obtain the accountability result. In other words, the initial accountability model may also include a fusion layer, an activation function layer (e.g., sigmoid), etc., as needed.

[0041] In some optional implementations of certain embodiments, step 102 may include the following two sub-steps:

[0042] Sub-step one: Extract the target network structure from each meta-model in at least one meta-model. This results in at least one target network structure.

[0043] As an example, the remaining structure outside the output layer of each meta-model can be extracted as the target network structure, thus obtaining at least one target network structure.

[0044] Sub-step two involves assembling at least one target network structure, a feature cross layer, and a logistic regression layer to obtain an initial liability assessment model.

[0045] In these implementations, the feature cross layer (cross) can be used for feature crossing, thereby capturing higher-order correlations between sparse features and improving the prediction accuracy of the accountability network. The accountability result is then obtained through the logistic regression layer. In practice, the outputs of the various target network structures and the feature cross layer can be concatenated and then input into the logistic regression layer to obtain the accountability result. In practice, the specific structures of the feature cross layer and the logistic regression layer can be flexibly configured according to actual needs. As an example, Figure 3 The diagram illustrates an exemplary structure of the meta-model in the accountability model generation method of this disclosure. At least one target network structure 301, a feature cross-layer 302, and a logistic regression layer 303 together constitute the initial accountability model.

[0046] Step 103: Based on the training sample set, train the initial accountability model to obtain the accountability model used to determine the accountability of both parties in the order corresponding to the business scenario information, and obtain the accountability result.

[0047] In some embodiments, the aforementioned execution entity can train the obtained initial accountability model based on the training sample set, and terminate the training and obtain the accountability model when the training stopping condition is met. In practice, the training stopping condition may be that the number of iterations reaches a preset threshold, the loss value is less than a preset loss value threshold, etc.

[0048] In some embodiments, the training samples used during training may include order features and annotation information. The annotation information can characterize the liability determination between the two parties involved in the order (i.e., the product / service provider and the consumer). For example, in the scenario of determining liability for canceled ride-hailing orders, the annotation information could be "driver is liable and passenger is not liable," "driver is not liable and passenger is liable," "driver is liable and passenger is liable," etc. It should be noted that in practice, either party can also be annotated. For example, the annotation information could also be "driver is liable," "driver is not liable," etc. The order features of a canceled order can be data from multiple dimensions, including but not limited to spatiotemporal features, consumer (passenger) features, text features generated during order execution, image features generated during order execution, etc. Each dimension will have different metrics.

[0049] Continuing with the ride-hailing application scenario as an example, spatiotemporal characteristics can include response time, estimated pick-up time / distance, actual pick-up time / distance, driver arrival time / distance, driver cancellation time / distance, etc. Consumer characteristics can include passenger credit score, number of completed rides / cancellations / negative reviews / complaints / being complained about in the past week / month / three months, driver service score, and driver's completed rides / cancellations / negative reviews / complaints / being complained about in the past week / month / three months, etc. Text features generated during order execution can include transcripts of driver / passenger conversations, instant messages sent by drivers / passengers, and transcripts of in-car recordings.

[0050] Based on this, the aforementioned executing entity can use the order features of canceled orders in the training samples as input and the annotation information of canceled orders as the expected output to train the initial accountability model. In practice, backpropagation, stochastic gradient descent, and other methods can be used to adjust the parameters in the initial accountability model until the training stopping condition is met, at which point training ends and the accountability model is obtained. The accountability model is used to determine the accountability of both parties corresponding to orders corresponding to business scenario information and generate accountability results. The training samples in the training sample set include the order features of the orders and the annotation information of the orders. Among them, the orders corresponding to business scenario information can be orders whose order attributes are related to the business scenario. For example, if the business scenario information is the accountability of canceled orders, then the corresponding orders are canceled orders.

[0051] Some embodiments of this disclosure provide a method for generating a responsibility assessment model, enabling rapid generation of such models. Specifically, this disclosure breaks down a business scenario into multiple sub-business scenarios and selects at least one meta-model based on each sub-business scenario, thereby achieving model building based on the business scenario decomposition. On this basis, the responsibility assessment model is finally obtained through meta-model transfer and training. Since different business scenarios often have overlapping meta-models, meta-model reuse can be achieved, avoiding the long model training time caused by independently building models for different business scenarios. Furthermore, since the meta-model is pre-trained, the training time can be further shortened.

[0052] Although Figure 1 The aforementioned embodiments can produce the described technical effects. However, since they are derived from a meta-model, the accuracy of the accountability model is closely related to the accuracy of the meta-model. Therefore, the meta-model in the accountability model generation method provided in this disclosure can be trained in the following manner.

[0053] Continue to refer to Figure 2 This illustrates an exemplary generation process 200 of the metamodel in the accountability model generation method of this disclosure. The metamodel is obtained through the following steps:

[0054] Step 201: Obtain the order sample set and the corresponding annotation information set. The annotation information includes the judgment result, judgment feature, and the category to which the judgment feature belongs. There is a one-to-one correspondence between the order samples and the annotation information.

[0055] In some embodiments, the execution entity for generating the meta-model may be the same as or different from the execution entity for generating the judgment model described above. This application does not impose any limitations on this.

[0056] Based on this, the execution entity generated by the meta-model can first obtain a set of order samples and a set of annotation information. The annotation information can include the judgment result, judgment features, and the category to which the judgment features belong. In practice, the judgment result for each order sample can be determined manually. The judgment result could be that the driver is responsible and the passenger is not responsible, the driver is not responsible and the passenger is responsible, the driver is responsible and the passenger is responsible, etc. Furthermore, judgment features can be annotated; that is, it needs to be labeled which features led to the judgment result. For example, if it is found that the driver increases the price during a call, or the driver stays at the order-grabbing point after accepting an order, then it is very likely that these features led to the order cancellation. In addition, the category of these judgment features needs to be labeled. The judgment features can be divided into different categories according to actual needs. As an example, the judgment features can be divided into different categories based on the type of order information they originate from (e.g., from order trajectory, from order call, from order video, from order message, etc.).

[0057] Step 202: Determine the target order sample set corresponding to the metamodel from the order sample set based on the category to which the judgment feature that matches the sub-business scenario information corresponding to the metamodel belongs.

[0058] In some embodiments, in practice, different order samples are often associated with different data types. For example, some order samples may only contain call data, while others may only contain trajectory data. Similarly, there may be order message data and order video data. Of course, some order samples may also contain two or more types of data.

[0059] Based on this, the aforementioned execution entity can first determine the accountability features that match the sub-business scenario information corresponding to the meta-model, and then determine the target order sample set corresponding to the meta-model from the order sample set according to the category to which the matched accountability features belong. For example, the sub-business scenario information corresponding to the aforementioned meta-model information is "driver overcharging." Therefore, the accountability feature matching the sub-business scenario information could be "driver overcharging." Then, since the category to which the accountability feature "driver overcharging" belongs is "from calls," order samples containing only call data can be selected, thus obtaining the target order sample set. This allows for accurate matching of order samples to each meta-model, thereby improving training efficiency.

[0060] Step 203: Extract the order features of each order sample in the target order sample set to obtain the order feature set.

[0061] In some embodiments, various feature extraction algorithms can be used to extract order features from each order sample, thereby obtaining an order feature set. For example, a bag-of-words model can be used for message data, while convolutional neural networks (CNNs) can be used to extract features for video data. It can be understood that the order features in the order feature set generally correspond one-to-one with the order samples in the target order sample set.

[0062] In some optional implementations of certain embodiments, order features of the target category for each order sample in the target order sample set can also be extracted. In practice, features can be categorized into different types based on the channel through which they are generated and the business process, such as basic order features, spatiotemporal features, user profile features, and text features. Text features include transcripts of driver / passenger conversations, instant messages sent by drivers / passengers, and in-vehicle transcripts. Tabular features include city, order type, vehicle type, and order channel. In practice, it has been found that different categories of features affect the prediction performance of the cloud model. Therefore, after determining the target order sample set for a meta-model, the target category of the features used as input to the meta-model can be determined based on the model's performance. For example, model performance can be determined through metrics such as accuracy and speed. For instance, testing has shown that using both text features and basic order features as input can improve model accuracy. Therefore, order features from these two categories can be used as input features for the model. This allows for the extraction of order features of the target category, thereby improving the prediction performance of each meta-model.

[0063] Optionally, the network structure of the initial meta-model can be determined based on the target category. Since each meta-model uses different feature categories, the network structure of each meta-model will also differ accordingly, thus better utilizing the output features and improving the predictive performance of the meta-model. For example, basic order features, spatiotemporal features, and user profile features can use DNN models, while text features can use TextCNN, HAN, BERT, etc. It should be noted that the categories here differ from those in step 201 in terms of classification method and classification purpose.

[0064] Step 204: Based on the target order sample set and the order feature set, train the initial meta-model to obtain the meta-model.

[0065] In some embodiments, the aforementioned execution entity can train an initial meta-model based on a target order sample set and an order feature set to obtain a meta-model. Specifically, order features can be used as input, and the judgment results from the corresponding order sample set can be used as the expected output to train the initial meta-model until the training termination condition is met, thereby obtaining the meta-model.

[0066] It is understandable that steps 201-204 can be repeated to obtain multiple meta-models. However, due to the large number of meta-models in practice, training multiple meta-models is time-consuming. Therefore, in some optional implementations of certain embodiments, meta-models in the meta-model set that use order features of the same category as input, such as those that all use text features as input, can be trained through joint training. This reduces the number of model parameters and speeds up the training process.

[0067] In some embodiments, the order samples of the meta-model are determined according to the category to which the liability assessment feature belongs, thereby ensuring the accuracy of the meta-model and helping to improve the accuracy of the liability assessment model in the future.

[0068] Further reference Figure 4 The document illustrates a flow 400 of some embodiments of a method for generating adjudication results. Flow 400 of this method includes the following steps:

[0069] Step 401: Extract the order features of the target order.

[0070] In some embodiments, the entity executing the accountability result generation method can extract order features of the target order using various feature extraction methods. It is understood that the entity executing the accountability result generation method and the entity executing the accountability model generation method may be the same or different, and this disclosure does not impose any limitations on this.

[0071] Step 402: Input the order features into the accountability model to obtain the accountability result.

[0072] In some embodiments, the aforementioned executing entity can input order characteristics into the liability determination model to obtain a liability determination result. For example, the liability determination result could be that the driver is liable and the passenger is not liable. The liability determination model is based on, for example, Figure 2 The corresponding methods of those embodiments generated

[0073] In some optional implementations of certain embodiments, the above method may further include the following steps:

[0074] Step one involves determining the contribution value of the output results of each target network structure and feature cross layer in at least one target network structure within the accountability model. In these implementations, the importance of the output results of each target network structure and feature cross layer can be interpreted using the SHAP (Shapley Additive Explanations) method, thereby obtaining the contribution value of each network. Here, the networks include the aforementioned target network structures and feature cross layers.

[0075] Step 2: Based on the judgment results and contribution values, select the target network from each target network structure and feature cross-layer.

[0076] In these implementations, the aforementioned executing entity can determine the responsibility result and contribution value, and select the target network from various target network structures and feature cross layers. For example, if the responsibility result is found to be faulty, k networks can be selected in descending order of contribution value. Conversely, if the responsibility result is found to be unfaulty, k networks can be selected in ascending order of contribution value. The value of k is specified according to the actual situation, for example, it can be 3.

[0077] Step 3: Output the sub-service scenario information corresponding to the target network as the reason for liability determination.

[0078] In some embodiments, the aforementioned executing entity can output the sub-business scenario information corresponding to the target network as the reason for liability determination. It is understood that since the target network is selected from various target network structures and feature cross layers, the sub-business scenario information corresponding to the target network can be the sub-business scenario information corresponding to the corresponding target network structure or feature cross layer. The target network structure is extracted from the meta-model. Therefore, the sub-business scenario information corresponding to the target network structure can be the sub-business scenario information corresponding to the corresponding meta-model. Furthermore, the sub-business scenario information corresponding to the feature cross layer can be determined based on its corresponding input features. For example, if the input feature corresponding to the feature cross layer is a passenger changing their boarding point, then the relevant information of the input feature can be used as the sub-business scenario information.

[0079] In some embodiments, a method for generating liability assessment results is provided, thereby generating such results. Optionally, the liability assessment results can also be accurately interpreted to reduce the number or proportion of order cancellations.

[0080] Further reference Figure 5 As an implementation of the methods shown in the above figures, this disclosure provides some embodiments of a responsibility assessment model generation apparatus, which are similar to... Figure 2 Corresponding to the method embodiments shown, the device can be specifically applied to various electronic devices.

[0081] like Figure 5As shown, the accountability model generation apparatus 500 in some embodiments includes a selection unit 501, a transfer unit 502, and a training unit 503. The selection unit 501 is configured to select at least one meta-model from a pre-trained meta-model set based on business scenario information. The transfer unit 502 is configured to transfer the at least one meta-model to obtain an initial accountability model. The training unit 503 is configured to train the initial accountability model based on a training sample set to obtain an accountability model used to determine the accountability of both parties in an order corresponding to the business scenario information, thereby obtaining an accountability result. The training samples in the training sample set include order features and order annotation information.

[0082] In some optional implementations of embodiments, the migration unit 502 may be further configured to: extract the target network structure of each metamodel in at least one metamodel to obtain at least one target network structure; and assemble the at least one target network structure, the feature cross layer, and the logistic regression layer to obtain an initial judgment model.

[0083] In some optional implementations of the embodiments, the metamodel in the metamodel set is trained through the following steps: obtaining an order sample set and a set of annotation information corresponding to the order sample set, wherein the annotation information includes the judgment result, judgment feature, and the category to which the judgment feature belongs; determining the target order sample set corresponding to the metamodel from the order sample set according to the category to which the judgment feature belongs that matches the sub-business scenario information corresponding to the metamodel; extracting the order features of each order sample in the target order sample set to obtain an order feature set; and training the initial metamodel based on the target order sample set and the order feature set to obtain the metamodel.

[0084] In some alternative implementations of the embodiments, the metamodels in the metamodel set that take order features of the same category as input are trained through joint training.

[0085] It is understandable that the units described in the device 500 are related to the reference. Figure 2 The steps in the described method correspond accordingly. Therefore, the operations, features, and beneficial effects described above for the method also apply to the device 500 and the units contained therein, and will not be repeated here.

[0086] The following is for reference. Figure 6 As an implementation of the methods shown in the above figures, this disclosure provides some embodiments of a responsibility assessment model generation apparatus, which are similar to... Figure 4 Corresponding to the method embodiments shown, the device can be specifically applied to various electronic devices.

[0087] like Figure 6As shown, the accountability result generation device 600 in some embodiments includes an extraction unit 601 and an accountability result generation unit 602. The extraction unit 601 is configured to extract order features of the target order. The accountability result generation unit 602 is configured to input the order features into an accountability model to obtain an accountability result.

[0088] In some optional implementations of certain embodiments, the apparatus may further include a determining unit, a network selecting unit, and an output unit. The determining unit is configured to determine the contribution value of the output results of each target network structure and feature cross-layer in the at least one target network structure. The network selecting unit is configured to select a target network from the respective target network structures and feature cross-layers based on the judgment result and the contribution value. The output unit is configured to output the sub-service scenario information corresponding to the target network.

[0089] It is understandable that the units described in the device 600 are related to the reference. Figure 4 The steps in the described method correspond accordingly. Therefore, the operations, features, and beneficial effects described above for the method also apply to device 400 and the units contained therein, and will not be repeated here.

[0090] Continue to refer to Figure 7 It illustrates electronic devices suitable for implementing some embodiments of this disclosure (e.g., Figure 1 A schematic diagram of the structure of the electronic device (700) in the device. Figure 7 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of this disclosure.

[0091] like Figure 7 As shown, the electronic device 700 may include a processing unit (e.g., a central processing unit, a graphics processor, etc.) 701, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 702 or a program loaded from a storage device 708 into a random access memory (RAM) 703. The RAM 703 also stores various programs and data required for the operation of the electronic device 700. The processing unit 701, ROM 702, and RAM 703 are interconnected via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0092] Typically, the following devices can be connected to I / O interface 705: input devices 706 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 707 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 708 including, for example, magnetic tapes, hard disks, etc.; and communication devices 709. Communication device 709 allows electronic device 700 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 7 An electronic device 700 with various devices is shown; however, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively. Figure 7 Each box shown can represent a device or multiple devices as needed.

[0093] In particular, according to some embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, some embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication device 709, or installed from storage device 708, or installed from ROM 702. When the computer program is executed by processing device 701, it performs the functions defined in the methods of some embodiments of this disclosure.

[0094] It should be noted that, in some embodiments of this disclosure, the computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In some embodiments of this disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In some embodiments of this disclosure, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0095] In some implementations, clients and servers can communicate using any currently known or future-developed network protocol such as HTTP (Hypertext Transfer Protocol) and can interconnect with digital data communication (e.g., communication networks) of any form or medium. Examples of communication networks include local area networks (“LANs”), wide area networks (“WANs”), the Internet (e.g., the Internet of Things), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future-developed networks.

[0096] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device. The aforementioned computer-readable medium carries one or more programs that, when executed by the electronic device, cause the electronic device to: select at least one meta-model from a pre-trained meta-model set based on business scenario information; perform transfer analysis on the at least one meta-model to obtain an initial accountability model; and train the initial accountability model based on a training sample set to obtain an accountability model for judging the responsibilities of both parties corresponding to the order corresponding to the business scenario information, thereby obtaining an accountability result. The training samples in the training sample set include order features and order annotation information; or

[0097] Extract the order features of the target order; input the order features into the accountability model to obtain the accountability result, wherein the accountability model is generated according to the method described in any of the implementation methods of the first aspect above; determine the contribution value of the output results of each target network structure and feature cross layer in at least one target network structure; select the target network from each target network structure and feature cross layer according to the accountability result and contribution value; output the sub-business scenario information corresponding to the target network.

[0098] Computer program code for performing operations of some embodiments of this disclosure can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone 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 remote computers, the remote computer can 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 can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0099] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0100] The units described in some embodiments of this disclosure can be implemented in software or hardware. The described units can also be housed in a processor; for example, a processor may be described as including a selection unit, a migration unit, and a training unit. The names of these units do not necessarily limit the unit itself; for example, a selection unit may be described as "a unit configured to select at least one meta-model from a pre-trained set of meta-models based on business scenario information."

[0101] The functions described above in this document can be performed, at least in part, by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: Field Programmable Gate Arrays (FPGAs), Application-Specific Integrated Circuits (ASICs), Application Standard Products (ASSPs), System-on-Chip (SoCs), Complex Programmable Logic Devices (CPLDs), and so on.

[0102] The above description is merely a selection of preferred embodiments of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in the embodiments of this disclosure.

Claims

1. A method for generating a liability assessment model, comprising: Based on at least one sub-business scenario information corresponding to the business scenario information, at least one meta-model is selected from the pre-trained meta-model set, wherein each meta-model corresponds to one sub-business scenario information. The at least one meta-model is transferred to obtain an initial accountability model; including: extracting the target network structure of each meta-model in the at least one meta-model to obtain at least one target network structure; assembling the at least one target network structure, a feature cross layer, and a logistic regression layer to obtain the initial accountability model; the feature cross layer is used for feature cross to capture higher-order correlations between sparse features; Based on the training sample set, the initial accountability model is trained to obtain an accountability model for judging the responsibilities of both parties corresponding to the order corresponding to the business scenario information, and to obtain the accountability result. The training samples in the training sample set include the order features of the order and the labeling information of the order.

2. The method according to claim 1, wherein, Assemble the at least one target network structure, feature cross layer, and logistic regression layer to obtain the initial liability judgment model, including: The outputs of the at least one target network structure and the feature cross layer are concatenated and then input into the logistic regression layer to obtain the judgment result output by the logistic regression layer.

3. The method according to claim 1, wherein, The metamodels in the metamodel set are obtained through the following steps: Obtain an order sample set and a set of annotation information corresponding to the order sample set, wherein the annotation information includes the judgment result, the judgment feature, and the category to which the judgment feature belongs; Based on the category of the judgment feature that matches the sub-business scenario information corresponding to the meta-model, determine the target order sample set corresponding to the meta-model from the order sample set; Extract the order features of each order sample from the target order sample set to obtain the order feature set; Based on the target order sample set and the order feature set, the initial meta-model is trained to obtain the meta-model.

4. The method according to claim 3, wherein, The step of extracting the order features of each order sample in the target order sample set includes: Extract the target category order features for each order sample in the target order sample set.

5. A method for generating a liability assessment result, comprising: Extract the order characteristics of the target order; The order features are input into the accountability model to obtain the accountability result. The accountability model is generated by the method described in any one of claims 1-4.

6. The method according to claim 5, wherein, The method further includes: Determine the contribution value of the output results of each target network structure and feature cross layer in at least one target network structure in the accountability model; Based on the judgment result and the contribution value, a target network is selected from each target network structure and feature cross layer; The sub-service scenario information corresponding to the target network is output as the reason for liability determination.

7. A liability assessment model generation device, comprising: The selection unit is configured to select at least one meta-model from a pre-trained meta-model set based on at least one sub-business scenario information corresponding to the business scenario information, wherein each meta-model corresponds to one sub-business scenario information. A transfer unit is configured to transfer the at least one meta-model to obtain an initial accountability model; the transfer unit is specifically configured to extract the target network structure of each meta-model in the at least one meta-model to obtain at least one target network structure; and assemble the at least one target network structure, a feature cross layer, and a logistic regression layer to obtain the initial accountability model; wherein, the feature cross layer is used for feature cross to capture higher-order correlations between sparse features. The training unit is configured to train the initial accountability model based on the training sample set to obtain an accountability model for judging the accountability of both parties corresponding to the order corresponding to the business scenario information, and to obtain an accountability result. The training samples in the training sample set include the order features of the order and the annotation information of the order.

8. A device for generating a judgment result, comprising: The extraction unit is configured to extract the order features of the target order; The accountability result generation unit is configured to input the order features into the accountability model to obtain the accountability result, wherein the accountability model is generated by the method according to any one of claims 1-4.

9. An electronic device, comprising: One or more processors; Storage device, on which one or more programs are stored, When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1-4 or 5-6.

10. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method described in any one of claims 1-4 or 5-6.

Images