Modeling methods, apparatuses, devices, and media

Abstract

The present disclosure relates to a modeling method, device, equipment and medium. The modeling method comprises: displaying a modeling graphical interface; in response to a selection operation of a user, adding a training data node corresponding to a target data set and a data processing node corresponding to a target operator in a second region; in response to an editing operation of the user, displaying a modeling flow directed acyclic graph formed by the training data node and the data processing node in the second region; and performing model training by using the target data set and the target operator according to the modeling flow directed acyclic graph, to obtain a trained model. According to the embodiments of the present disclosure, the efficiency of model training can be improved while ensuring the security of data of each party, thereby improving the user experience.

Description

Technical Field

[0001] This disclosure relates to the field of artificial intelligence technology, and in particular to a modeling method, apparatus, device and medium. Background Technology

[0002] In traditional computing methods, one party typically performs data calculations based on its own data. However, with the development of big data, situations have arisen where multiple parties need to provide data for collaborative calculations.

[0003] However, in related technologies, collaborative computing is often not possible while protecting data privacy, which reduces the security of multi-party data and makes it difficult for multiple parties to cooperate efficiently, thus affecting the user experience. Summary of the Invention

[0004] To address the aforementioned technical problems, this disclosure provides a modeling method, apparatus, device, and medium.

[0005] Firstly, this disclosure provides a modeling method, including:

[0006] The modeling graphical interface is displayed. The modeling graphical interface includes a first area and a second area. The first area is used to provide a dataset list and an operator list. The dataset list includes a first dataset list from the model trainer and a second dataset list authorized by the model collaborator to the model trainer.

[0007] In response to the user's selection operation, the training data node corresponding to the target dataset and the data processing node corresponding to the target operator are added in the second area. The target dataset is the dataset selected in the dataset list by the selection operation, and the target operator is the operator selected in the operator list by the selection operation.

[0008] In response to the user's editing action, a Directed Acyclic Graph (DAG) of the modeling process, consisting of training data nodes and data processing nodes, is displayed in the second area. The training data nodes and data processing nodes in the DAG have logical relationships set by the editing action.

[0009] Based on the directed acyclic graph modeling process, the target dataset and target operator are used to train the model, resulting in the trained model.

[0010] Secondly, this disclosure provides a modeling apparatus, comprising:

[0011] The first display module is used to display the modeling graphical interface. The modeling graphical interface includes a first area and a second area. The first area is used to provide a dataset list and an operator list. The dataset list includes a first dataset list from the model trainer and a second dataset list authorized by the model collaborator to the model trainer.

[0012] The operator selection module is used to respond to the user's selection operation by adding the training data node corresponding to the target dataset and the data processing node corresponding to the target operator in the second area. The target dataset is the dataset selected in the dataset list by the selection operation, and the target operator is the operator selected in the operator list by the selection operation.

[0013] The second display module is used to respond to the user's editing operation and display the directed acyclic graph of the modeling process formed by the training data nodes and the data processing nodes in the second area. The training data nodes and the data processing nodes in the directed acyclic graph of the modeling process have logical relationships set by the editing operation.

[0014] The model training module is used to train the model using the target dataset and target operators based on the directed acyclic graph of the modeling process, and obtain the trained model.

[0015] Thirdly, this disclosure provides a modeling apparatus, including:

[0016] processor;

[0017] Memory, used to store executable instructions;

[0018] The processor is used to read executable instructions from memory and execute the executable instructions to implement the modeling method of the first aspect.

[0019] Fourthly, this disclosure provides a computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to implement the modeling method of the first aspect.

[0020] The technical solution provided in this disclosure has the following advantages compared with the prior art:

[0021] The modeling method, apparatus, device, and medium of this disclosure are capable of displaying a modeling graphical interface. The graphical interface displays a first region and a second region, including a dataset list and an operator list. The dataset list includes a first dataset list for the model trainer and a second dataset list authorized by the model collaborator to the model trainer. Then, in response to a user's selection operation, the second region adds the dataset selected in the dataset list (i.e., the training data node corresponding to the target dataset) and the operator selected in the operator list (i.e., the data processing node corresponding to the target operator). Continuing to respond to a user's editing operation, the second region displays a directed acyclic graph (DAG) of the modeling process formed by the training data nodes and data processing nodes. The training data nodes and data processing nodes in this DAG have logical relationships set by the editing operation. Finally, based on the DAG, the modeling process is used to... The model is trained using the dataset and target operator to obtain the trained model. Based on the user's selections, training data nodes corresponding to the target dataset and data processing nodes corresponding to the target operator are added to the second area of ​​the displayed modeling graphical interface. A directed acyclic graph (DAG) of the modeling process is also displayed in the second area based on the user's editing. This allows for model training using the target dataset and target operator, resulting in the trained model. Furthermore, it enables the fusion and integration of data from multiple sources for model training. The use of DAG modeling facilitates data and algorithm selection and the establishment of logical relationships, effectively lowering the barrier to entry for modeling. Moreover, model collaborators can train the model locally, thus improving training efficiency while ensuring data security for all parties, thereby enhancing the user experience. Attached Figure Description

[0022] 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 the originals and elements are not necessarily drawn to scale.

[0023] Figure 1 A schematic flowchart illustrating a modeling method provided in an embodiment of this disclosure;

[0024] Figure 2 A schematic diagram of a modeling graphical interface provided in an embodiment of this disclosure;

[0025] Figure 3 A schematic diagram of a waiting prompt interface provided in an embodiment of this disclosure;

[0026] Figure 4 A schematic diagram of a data statistics page provided in an embodiment of this disclosure;

[0027] Figure 5 A schematic diagram of an information prompt interface provided in an embodiment of this disclosure;

[0028] Figure 6 A schematic diagram of an authorization management interface provided in an embodiment of this disclosure;

[0029] Figure 7 This is a schematic diagram of the structure of a modeling device provided in an embodiment of the present disclosure;

[0030] Figure 8 This is a schematic diagram of the structure of a modeling device provided in an embodiment of the present disclosure. Detailed Implementation

[0031] 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.

[0032] It should be understood that the steps described in the method embodiments of this disclosure may be performed in different orders and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of this disclosure is not limited in this respect.

[0033] The term "comprising" and its variations as used herein are open-ended inclusions, meaning "including but not limited to". The term "based on" means "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". Definitions of other terms will be given in the description below.

[0034] 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.

[0035] 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".

[0036] 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.

[0037] To address the aforementioned problems, this disclosure provides a modeling method, apparatus, device, and medium. The following is a detailed description... Figures 1 to 6 The modeling method provided in the embodiments of this disclosure will be described in detail.

[0038] Figure 1 A schematic flowchart of a modeling method provided in an embodiment of this disclosure is shown.

[0039] In this embodiment of the disclosure, the modeling method can be executed by an electronic device. The electronic device may include, but is not limited to, mobile terminals such as laptops, PDAs (personal digital assistants), and PADs (tablet computers), as well as fixed terminals such as digital TVs and desktop computers.

[0040] like Figure 1 As shown, the modeling method may include the following steps.

[0041] S110. Display the modeling graphical interface. The modeling graphical interface includes a first area and a second area. The first area is used to provide a dataset list and an operator list. The dataset list includes a first dataset list from the model trainer and a second dataset list authorized by the model collaborator to the model trainer.

[0042] In this embodiment of the disclosure, when a user wants to perform modeling, the modeling operation can be performed through an electronic device. The electronic device can respond to the user's modeling operation and display a modeling graphical interface.

[0043] Optionally, the modeling graphical interface can be a graphical interface used by users to perform modeling operations.

[0044] Optionally, the modeling graphical interface may include a first region and a second region. The first region may be used to provide a list of datasets and a list of operators, while the second region may be used to display a directed acyclic graph of the modeling process.

[0045] Optionally, the dataset list can be a list of multiple datasets. Specifically, the dataset list may include a first dataset list from the model trainer and a second dataset list authorized by the model collaborator to the model trainer.

[0046] Optionally, the list of operators can be a list of multiple operators.

[0047] Optionally, operators can be elements used to compose a directed acyclic graph of the modeling process. These operators can include those related to the dataset, those related to the processing method, and those related to the algorithm.

[0048] Optionally, the model trainer can be a user who performs modeling operations and trains the model.

[0049] Optionally, the model collaborator can be a user who provides the dataset license.

[0050] Optionally, the first dataset list can be a list of datasets provided by the model trainer.

[0051] Optionally, the second dataset list can be a list of datasets that the model collaborator has licensed to the model trainer.

[0052] Optionally, the directed acyclic graph of the modeling process can be a directed acyclic graph used to illustrate the modeling process.

[0053] Figure 2 A schematic diagram of a modeling graphical interface provided in an embodiment of this disclosure is shown.

[0054] like Figure 2 As shown, in the modeling graphical interface, the first area 201 can be used to provide an element list, namely a dataset list and an operator list, and the second area 202 can be used to display the directed acyclic graph 203 of the modeling process.

[0055] Specifically, when a user wants to perform modeling, they can do so through an electronic device. The electronic device responds to the user's modeling actions by displaying a graphical modeling interface comprising a first area and a second area. The first area can provide a list of datasets and operators, while the second area can display a directed acyclic graph (DAG) of the modeling process. The dataset list can include data from both the user (the model trainer) and collaborators (model partners).

[0056] S120. In response to the user's selection operation, add the training data node corresponding to the target dataset and the data processing node corresponding to the target operator in the second area. The target dataset is the dataset selected in the dataset list by the selection operation, and the target operator is the operator selected in the operator list by the selection operation.

[0057] In this embodiment of the disclosure, after the electronic device displays the modeling graphical interface, it can add training data nodes corresponding to the target dataset and data processing nodes corresponding to the target operator in the second area in response to the user's selection operation.

[0058] Optionally, the selection operation can be an operation in which the user selects a dataset from the dataset list and an operator from the operator list within the first region on the modeling graphical interface.

[0059] Optionally, the target dataset can be the dataset selected by the user from the dataset list through a selection action.

[0060] Optionally, the target operator can be an operator selected by the user from the list of operators through a selection operation.

[0061] Alternatively, the training data nodes can be nodes generated by the user selecting the target dataset through a selection operation.

[0062] Optionally, the data processing node can be a node generated by the user selecting the target operator through a selection operation.

[0063] Continue to refer to Figure 2 In the second region 202, the "first dataset" node and the "second dataset" node can be the training data nodes corresponding to the target dataset, that is, the target dataset can be the first dataset and the second dataset. The "data analysis" node, the "federated feature engineering" node, the "longitudinal logistic regression algorithm" node and the "model prediction" node can be the data processing nodes corresponding to the target operators, that is, the target operators can be the data analysis operator, the federated feature engineering operator, the longitudinal logistic regression algorithm operator and the model prediction operator.

[0064] Alternatively, the data analysis operator can be an operator used to analyze and process the dataset.

[0065] Alternatively, the federated feature engineering operator can be an operator that encrypts and performs feature processing on the dataset.

[0066] Optionally, the longitudinal logistic regression algorithm operator can be an operator used to train a model on a dataset.

[0067] Optionally, the model prediction operator can be an operator for predicting the model after training.

[0068] Specifically, after displaying the modeling graphical interface, the electronic device can respond to the user's selection operation by adding the dataset selected in the dataset list (i.e., the training data node corresponding to the target dataset) in the second area, and adding the operator selected in the operator list (i.e., the data processing node corresponding to the target operator) in the second area.

[0069] S130. In response to the user's editing operation, a directed acyclic graph of the modeling process formed by training data nodes and data processing nodes is displayed in the second area. The training data nodes and data processing nodes in the directed acyclic graph of the modeling process have logical relationships set by the editing operation.

[0070] In this embodiment of the disclosure, after the electronic device adds training data nodes and data processing nodes in the second area, it can also respond to the user's editing operation by displaying a directed acyclic graph of the modeling process formed by the training data nodes and data processing nodes in the second area.

[0071] Optionally, the editing operation can be an operation in which the user edits each node in the second area.

[0072] Optionally, the directed acyclic graph of the modeling process can be a directed acyclic graph formed by the logical relationships set between training data nodes and data processing nodes through editing operations.

[0073] Specifically, users can set the logical relationship between training data nodes and data processing nodes through editing operations, thereby enabling the training data nodes and data processing nodes to form a directed acyclic graph of the modeling process based on the logical relationship.

[0074] Continue to refer to Figure 2 In the second area 202, users can edit the training data nodes: the "First Dataset" node and the "Second Dataset" node, and the data processing nodes: the "Data Analysis" node, the "Federated Feature Engineering" node, the "Longitudinal Logistic Regression Algorithm" node, and the "Model Prediction" node. This creates a logical relationship between the training data nodes and the data processing nodes. The electronic device can respond to this logical relationship and display the logical relationship between the training data nodes and the data processing nodes in the second area 202 through directed connections between the nodes. The second area 202 also displays a directed acyclic graph 203 of the modeling process formed by the training data nodes and the data processing nodes.

[0075] Specifically, after the electronic device adds training data nodes and data processing nodes in the second area, the user can set the logical relationship between the training data nodes and data processing nodes through editing operations. The electronic device can receive and respond to the editing operation, display the logical relationship between the training data nodes and data processing nodes in the second area, and display the directed acyclic graph of the modeling process formed by the training data nodes and data processing nodes based on the logical relationship.

[0076] S140. Based on the directed acyclic graph in the modeling process, use the target dataset and target operator to train the model and obtain the trained model.

[0077] In this embodiment of the disclosure, after the electronic device displays the directed acyclic graph of the modeling process in the second area, it can use the target dataset and target operator to train the model based on the directed acyclic graph of the modeling process, and obtain the trained model. That is, it can use the data of multiple partners to train the model and obtain the trained model.

[0078] Specifically, after displaying the directed acyclic graph of the modeling process, the electronic device can train the model according to the logical relationship between the target dataset and the target operator based on the directed acyclic graph of the modeling process, thereby obtaining the trained model.

[0079] Continue to refer to Figure 2The electronic device can first obtain the first dataset and the second dataset according to the nodes included in the directed acyclic graph 203 of the modeling process, perform data analysis and processing on the first dataset, perform federated feature engineering on the processed first dataset and the second dataset, then perform regression calculation on the obtained encrypted data through the longitudinal logistic regression algorithm to obtain the model to be predicted, and finally perform model prediction on the model to be predicted to obtain the trained model.

[0080] Therefore, in this embodiment of the disclosure, a modeling graphical interface can be displayed, and a first area and a second area including a dataset list and an operator list can be displayed in the modeling graphical interface. The dataset list includes a first dataset list for the model trainer and a second dataset list authorized by the model collaborator to the model trainer. Then, in response to the user's selection operation, the dataset selected in the dataset list (i.e., the training data node corresponding to the target dataset) and the operator selected in the operator list (i.e., the data processing node corresponding to the target operator) are added to the second area. Continuing to respond to the user's editing operation, a directed acyclic graph of the modeling process formed by the training data nodes and data processing nodes is displayed in the second area. The training data nodes and data processing nodes in this directed acyclic graph have logical relationships set by the editing operation. Finally, based on the directed acyclic graph of the modeling process, the target dataset and... The target operator is used to train the model, resulting in a trained model. Based on user selections, training data nodes corresponding to the target dataset and data processing nodes corresponding to the target operator are added to the second area of ​​the displayed modeling graphical interface. A directed acyclic graph (DAG) of the modeling process is also displayed in the second area, allowing for model training using the target dataset and target operator. This enables the fusion and integration of data from multiple sources for model training. Furthermore, the use of a DAG modeling approach facilitates data and algorithm selection and the establishment of logical relationships, effectively lowering the barrier to entry for modeling. Moreover, model collaborators can train the model locally, thus improving training efficiency while ensuring data security for all parties, thereby enhancing the user experience.

[0081] The S140 of the present disclosure embodiment will now be described in detail.

[0082] Optionally, the target dataset may include a first dataset selected from a first dataset list and a second dataset selected from a second dataset list.

[0083] Specifically, the target dataset may include a first dataset selected by the selection operation from the first dataset list of the model trainer, and a second dataset selected from the second dataset list authorized by the model collaborator to the model trainer.

[0084] Optionally, S140 may specifically include:

[0085] S1401. Based on the directed acyclic graph of the modeling process, use the first dataset and the target operator to train the model and obtain the first model parameters.

[0086] In this embodiment of the disclosure, after the electronic device displays the directed acyclic graph of the modeling process, it can use the first dataset and the target operator to train the model and obtain the first model parameters.

[0087] Optionally, the first model parameters can be model parameters obtained by training the model using the first dataset.

[0088] Specifically, after displaying the directed acyclic graph of the modeling process, the electronic device can obtain the first dataset selected from the first dataset list of the model training side according to the user's selection operation, and use the first dataset and the target operator to train the model according to the logical relationship between the first dataset and the target operator in the directed acyclic graph of the modeling process, thereby obtaining the first model parameters.

[0089] Continue to refer to Figure 2 The electronic device can train the model using the first dataset and the target operator based on the logical relationship between the "first dataset" node corresponding to the first dataset and the "data analysis" node, "federated feature engineering" node, "longitudinal logistic regression algorithm" node and "model prediction" node corresponding to the target operator in the directed acyclic graph 203 of the modeling process, thereby obtaining the first model parameters.

[0090] S1402. Send a model training request to the model partner to which the second dataset belongs. The model training request carries the model training file formed by the directed acyclic graph of the modeling process. The model training request is used to enable the model partner to which the second dataset belongs to train the model based on the second dataset and the model training file to obtain the second model parameters.

[0091] In this embodiment, a model training request can be sent to the model collaborator to which the second dataset belongs. The model training request can be sent after the electronic device obtains the first model parameters, or it can be sent before the entire training process begins. That is, the local model training and the model collaborator's model training processes can be performed in parallel or sequentially. Provided the modeling objective is achieved, this application does not impose any restrictions on the execution order.

[0092] Alternatively, the model training request can be a request to enable the model collaborator to train the model based on the second dataset and the model training file.

[0093] Optionally, the model training request may carry a model training file formed by the directed acyclic graph of the modeling process.

[0094] Optionally, the model training file can be formed from a directed acyclic graph based on the modeling process or a target directed acyclic subgraph.

[0095] Optionally, the target directed acyclic subgraph can be composed of the target nodes related to the second dataset and the logical relationships between the target nodes in the directed acyclic graph of the modeling process.

[0096] Continue to refer to Figure 2 In the directed acyclic graph 203 of the modeling process, the target nodes related to the second dataset can be the "second dataset" node, the "federated feature engineering" node, the "longitudinal logistic regression algorithm" node, and the "model prediction" node. The electronic device can form a target directed acyclic subgraph based on each target node and the logical relationship between each target node.

[0097] Specifically, the electronic device can construct a corresponding target directed acyclic subgraph based on the target nodes related to the second dataset and the logical relationships between the target nodes in the directed acyclic graph of the modeling process, and form a corresponding model training file. Then, it sends a model training request carrying the model training file to the model partner. After receiving the model training request, the model partner can train the model on the second dataset selected by the user's selection operation in the second dataset list based on the model training file, thereby obtaining the corresponding second model parameters, and feeding back the second model parameters to the electronic device.

[0098] Therefore, in this embodiment of the disclosure, the electronic device can train the model using the directed acyclic graph of the modeling process according to the datasets of each party, and obtain the corresponding first model parameters and second model parameters. Since the model trainer and the model collaborator both conduct model training locally, the efficiency of model training is improved while ensuring the security of the data of each party, thereby enhancing the user experience.

[0099] S1403: Receive the second model parameters fed back by the model collaborator to which the second dataset belongs.

[0100] In this embodiment of the disclosure, after the electronic device sends a model training request to the model collaborator to which the second dataset belongs, it can receive the second model parameters fed back by the model collaborator to which the second dataset belongs.

[0101] S1404. Perform parameter fusion on the first model parameters and the second model parameters to obtain the model parameters of the trained model.

[0102] In this embodiment of the disclosure, after receiving the second model parameters fed back by the model collaborator to which the second dataset belongs, the electronic device can fuse the first model parameters and the second model parameters to obtain the model parameters of the trained model.

[0103] Optionally, the first model parameters and the second model parameters each include at least a portion of the model parameters.

[0104] Specifically, the first model parameters can be at least a portion of the model parameters about the first dataset in the model obtained by training the model based on the first dataset and the directed acyclic graph of the modeling process, and the second type of parameters can be at least a portion of the model parameters about the second dataset in the model obtained by training the model based on the second dataset and the directed acyclic graph of the modeling process.

[0105] In some embodiments, the first model parameter and the second model parameter may each include different parts of the model parameters from all the model parameters. For example, the model parameters may include parameter A, parameter B, and parameter C, the first model parameter may include parameter A and parameter B, and the second model parameter may include parameter C.

[0106] Furthermore, the electronic device can train the model using the first dataset and the target operators in the directed acyclic graph of the modeling process, thereby obtaining the first model parameters. It can also construct a corresponding target directed acyclic subgraph based on the target nodes related to the second dataset and the logical relationships between them in the directed acyclic graph of the modeling process, and form a corresponding model training file. Then, it sends a model training request carrying the model training file to the model partner, so that after receiving the model training request, the model partner can train the model on the second dataset selected by the user's selection operation in the second dataset list based on the model training file, thereby obtaining the corresponding second model parameters.

[0107] In other embodiments, the first model parameter and the second model parameter may each include overlapping model parameters among all model parameters. For example, the model parameters may include parameter A, parameter B, and parameter C, the first model parameter may include parameter A and parameter B, and the second model parameter may include parameter B and parameter C.

[0108] Furthermore, the electronic device can train the model using the first dataset and the target operators in the directed acyclic graph of the modeling process, thereby obtaining the first model parameters. It can also construct a corresponding target directed acyclic subgraph based on the target nodes related to the second dataset and the logical relationships between them in the directed acyclic graph of the modeling process, and form a corresponding model training file. Then, it sends a model training request carrying the model training file to the model partner, so that after receiving the model training request, the model partner can train the model on the second dataset selected by the user's selection operation in the second dataset list based on the model training file, thereby obtaining the corresponding second model parameters.

[0109] In some other embodiments, the first model parameter and the second model parameter may each include all model parameters present in the list of all model parameters. For example, the model parameters may include parameter A, parameter B, and parameter C; the first model parameter may include parameter A1, parameter B1, and parameter C1; and the second model parameter may include parameter A2, parameter B2, and parameter C2.

[0110] Furthermore, the electronic device can train the model based on the first dataset and the directed acyclic graph of the modeling process to obtain the first model parameters; it can also generate a corresponding model training file based on the directed acyclic graph and the second dataset of the modeling process, and then send a model training request carrying the model training file to the model partner. After receiving the model training request, the model partner can train the model on the second dataset selected by the user's selection operation in the second dataset list based on the model training file to obtain the corresponding second model parameters.

[0111] Optionally, parameter fusion can be performed on the first model parameters and the second model parameters. For example, when the first model parameters and the second model parameters can each include different parts of the model parameters, the first model parameters and the second model parameters can be added to the corresponding model parameter positions in the model to obtain the corresponding model; when the first model parameters and the second model parameters can each include overlapping parts of the model parameters, the overlapping parts of the first model parameters and the second model parameters can be weighted and summed, averaged, etc., to obtain the corresponding partial model parameters, and the parts of the partial model parameters that do not overlap with the first model parameters and the second model parameters can be added to the corresponding model parameter positions in the model; when the first model parameters and the second model parameters each include all the model parameters, the first model parameters and the second model parameters can each be weighted and summed, averaged, etc., to obtain the corresponding all model parameters, and all model parameters can be added to the corresponding model parameter positions in the model.

[0112] Specifically, after receiving the second model parameters from the model partner to which the second dataset belongs, the electronic device can fuse the first model parameters and the second model parameters according to a pre-determined parameter fusion method to obtain the model parameters of the trained model.

[0113] Therefore, in this embodiment of the disclosure, the model can be trained separately based on the datasets of each party to obtain the model parameters corresponding to the datasets of each party, and then the model parameters of the trained model can be determined. This improves the efficiency of model training while ensuring the security of the data of each party, thereby enhancing the user experience.

[0114] In some embodiments, optionally, after the electronic device sends a model training request to the model partner to which the second dataset belongs, it is also necessary to wait for the model partner to train the second dataset.

[0115] Optionally, after sending a model training request to the model collaborator to which the second dataset belongs, the modeling method further includes: displaying a waiting prompt window, the waiting prompt window containing waiting prompt information and a return trigger control, the return trigger control being used to trigger the cancellation of displaying the waiting prompt window.

[0116] In this embodiment of the disclosure, after the electronic device sends a model training request to the model collaborator to which the second dataset belongs, the electronic device may display a waiting prompt window, wherein the display waiting prompt window may include waiting prompt information and a return trigger control.

[0117] Figure 3 A schematic diagram of a waiting prompt interface provided in an embodiment of this disclosure is shown.

[0118] like Figure 3 As shown, the electronic device can display a waiting prompt interface, which can display a waiting prompt window 301. The waiting prompt window 301 can include a waiting prompt message 302 and a return trigger control 303.

[0119] Optionally, the waiting prompt window can be a prompt window used to display waiting prompt information to the user and return to the trigger control.

[0120] Optionally, the waiting prompt message can be information used to inform the user and model partners about the progress of model training on the second dataset and model training files. For example, Figure 3 The 302 error message can be interpreted as "Processing, please wait".

[0121] Optionally, the return trigger control can be used to trigger the cancellation of the display of the waiting prompt window.

[0122] See also Figure 3After the electronic device sends a model training request to the model partner, a waiting prompt window 301 can be superimposed on the directed acyclic graph of the modeling process. The waiting prompt window 301 can display a waiting prompt message 302 "Processing, please wait" and a return trigger control 303.

[0123] Furthermore, the user can also trigger the return trigger control 303 displayed in the waiting prompt window 301. The electronic device can receive and respond to the trigger operation, cancel the display of the waiting prompt window 301, and continue to display the directed acyclic graph of the modeling process.

[0124] Optionally, the triggering action can be an action that the user triggers to return to the trigger control.

[0125] Therefore, in this embodiment of the disclosure, after the electronic device sends a model training request to the model partner, a waiting prompt window can be displayed to the user, so that the user can keep track of the model training progress at all times and improve the user experience.

[0126] In other embodiments, users can also view the data used by the trained model.

[0127] Optionally, the modeling method may further include: in response to a user's data viewing operation, displaying a data statistics page, the data statistics page including partner statistics information corresponding to at least one model partner; in response to a user's viewing operation of target partner statistics information in at least one partner statistics information, displaying authorized data statistics information corresponding to the target model partner to which the target partner statistics information belongs, the authorized data statistics information being used to display the historical data usage quantity authorized by the target model partner to the model training party, the historical data usage quantity being obtained based on statistics from the historical model training process.

[0128] In this embodiment of the disclosure, the electronic device can receive and respond to the user's data viewing operation and display the corresponding data statistics page.

[0129] Optionally, the data viewing operation can be an operation used by users to view the data statistics page.

[0130] Optionally, the data statistics page can be a page that includes statistical information about at least one model partner.

[0131] Optionally, the partner statistics information can be the data statistics information corresponding to the model partner. For example, the partner statistics information may include the model partner identifier (ID), the model partner name, the dataset, and statistical data controls, etc., which are not limited here.

[0132] Specifically, when a user wants to view the data used after model training, they can perform the data viewing operation through an electronic device. The electronic device can receive and respond to the data viewing operation and display the corresponding data statistics page, which may include statistical information of at least one model partner.

[0133] This statistical data can be used in scenarios such as billing.

[0134] Figure 4 This illustration shows a schematic diagram of a data statistics page provided in an embodiment of the present disclosure.

[0135] like Figure 4 As shown, the data statistics page 401 may include the partner statistics information 402 corresponding to the three model partners. The partner statistics information 402 may include the corresponding model partner ID (Identity document, ID), model partner name, dataset, and statistical data control 403, etc.

[0136] Furthermore, after displaying the data statistics page, the electronic device can respond to the user's viewing operation of the target partner statistics in at least one partner statistics information by displaying the authorized data statistics information corresponding to the target model partner to which the target partner statistics information belongs.

[0137] Optionally, the target partner statistics can be the partner statistics that the user wants to view.

[0138] Optionally, the view operation can be an operation for users to view statistical information of the target partner.

[0139] Optionally, the target model partner can be the model partner corresponding to the target partner's statistical information.

[0140] Optionally, the authorized data statistics can be statistics on the number of historical data uses that the target model partner has authorized to the model trainer.

[0141] Optionally, the amount of historical data used can be the number of datasets used during the training of historical models.

[0142] Specifically, during the historical model training process, the electronic device can count the number of all datasets used in the historical model training process or the number of data in the dataset, thereby obtaining the corresponding historical data usage.

[0143] For example, before training a model, a user can select the dataset to be trained through an electronic device. The electronic device can obtain and count the number of datasets selected by the user and / or the number of data contained in the dataset. During the model training process, it can also obtain and count the number of training datasets / data used to train the model and the number of training sessions. Thus, after the model training is completed, the number of datasets / data used by the model in the historical training process can be obtained, that is, the number of historical data used.

[0144] Continue to refer to Figure 4 When a user wants to view the target partner's statistics, they can trigger the statistics data control 403 in the partner statistics information 402. The electronic device can respond to this viewing operation and display the authorized data statistics information corresponding to the target model partner to which the target partner statistics information belongs.

[0145] Specifically, users can view the statistical data control through electronic devices. In response to this viewing operation, the electronic device can display the authorized data statistics information corresponding to the target model partner. This authorized data statistics information can be used to show the historical data usage of the target model partner to the model training party.

[0146] Therefore, in this embodiment of the disclosure, the electronic device can display the authorized data statistics information corresponding to the model partner, so that the user can understand in detail the amount of data used during the model training process, which facilitates the subsequent settlement process and improves the user experience.

[0147] In some other embodiments, the electronic device can also filter the dataset list, which will be described in detail below.

[0148] Optionally, prior to S120, the modeling method may further include: in response to a user's data filtering operation on the dataset list, filtering the dataset list according to the filtering method corresponding to the data filtering operation.

[0149] In this embodiment of the disclosure, the electronic device can receive and respond to a user's data filtering operation on the dataset list, and filter the dataset list according to the filtering method corresponding to the data filtering operation, thereby obtaining a filtered dataset list.

[0150] Alternatively, the data filtering operation can be an operation for users to filter a list of datasets.

[0151] Optionally, the filtering method can be a pre-defined method for filtering different datasets. For example, filtering by the number of records or by data type.

[0152] Specifically, before model training, users can filter the dataset list using an electronic device. The electronic device can receive and respond to the user's data filtering operation, filtering the dataset list according to the filtering method corresponding to the data filtering operation, and obtaining the filtered dataset list. For example, the user can filter the first dataset list of the model training party and the second dataset list authorized by the model partner to the model training party. The electronic device can obtain the data volume of the dataset in the dataset list after the data filtering operation, and then use this data volume as the historical data usage quantity for statistics.

[0153] See also Figure 2 Users can select a first dataset list from the dataset list, choosing a filtering method for this list. For example, they can filter by the number of records, the data type, or a range of records. This allows the electronic device to receive and respond to the filtering operation, selecting a dataset with 100 records and filtering it according to the specified method. The result is a filtered list of the first dataset containing 100 records, along with the total data volume. Similarly, filtering can be performed on a second dataset list authorized by the model partner for model training; this will not be elaborated upon here.

[0154] Optionally, after a user performs a data filtering operation on the dataset list, the electronic device can respond to the data filtering operation by displaying all the datasets filtered out in the first area, so that the user can select the target dataset for model training in the first area and then perform model training.

[0155] Therefore, in this embodiment of the disclosure, the dataset list can be filtered according to the filtering method, which can effectively select the required dataset, reduce the number of datasets used to train the model, thereby reducing the model training cost and improving the user experience.

[0156] Optionally, the selection operation may include dataset drag-and-drop operations and operator drag-and-drop operations.

[0157] Optionally, the dataset drag-and-drop operation may include dragging the target dataset from the first region to the second region.

[0158] Optionally, the operator drag operation may include dragging the target operator from the first region to the second region.

[0159] Continue to refer to Figure 2 Users can select a target dataset in the first area 201 and drag it to the second area, which is called a dataset drag-and-drop operation. Similarly, users can select a target operator in the first area 201 and drag it to the second area, which is called an operator drag-and-drop operation. The electronic device can respond to the dataset drag-and-drop operation and the operator drag-and-drop operation by adding the training data node corresponding to the target dataset and the data processing node corresponding to the target operator in the second area 202.

[0160] Therefore, in this embodiment of the disclosure, users can select the target dataset and target operator by dragging and dropping, which can facilitate and accurately construct the directed acyclic graph of the modeling process and improve the user experience.

[0161] Optionally, S140 may specifically include: responding to the user's model training trigger operation, training the target operator using the target dataset according to the directed acyclic graph of the modeling process, and obtaining the trained model.

[0162] In this embodiment of the disclosure, the electronic device can receive a user's model training trigger operation and, in response to the model training trigger operation, train the target operator using the target dataset according to the directed acyclic graph of the modeling process to obtain the trained model.

[0163] Optionally, the model training trigger operation can be a user-initiated operation to train the model to be trained.

[0164] Optionally, the model training trigger operation can be a trigger operation on the model training trigger control.

[0165] Specifically, users can trigger the model training by activating the model training trigger control displayed on the electronic device. The electronic device can receive and respond to the model training trigger operation and train the model according to the logical relationship between the target dataset and the target operator in the directed acyclic graph of the modeling process, thereby obtaining the trained model. The specific implementation method is described above and will not be repeated here.

[0166] Therefore, in this embodiment of the disclosure, model training can be performed in response to a model training trigger operation, thereby obtaining a trained model, improving the efficiency of model training, and thus enhancing the user experience.

[0167] In some other embodiments, the resource data required for model training needs to be confirmed before model training can be performed, which will be described in detail below.

[0168] Optionally, prior to S140, the modeling method further includes: in response to the model confirmation operation, displaying an information prompt window, the information prompt window including at least one of the following: dataset information of the target dataset, dataset source of the target dataset, resource usage information of the target dataset for the dataset source, and a model training trigger control, wherein the resource usage information is determined by the resource usage parameters carried by the target dataset; wherein the model training trigger operation is a trigger operation on the model training trigger control, the model training trigger control is used to trigger the application for authorization to train the model from the model partner.

[0169] In this embodiment of the disclosure, before performing model training, the electronic device can also receive and respond to the user's model confirmation operation and display a corresponding information prompt window.

[0170] Optionally, the model confirmation operation can be an operation for the user to confirm the model information.

[0171] Optionally, the model confirmation operation can be an operation in which the user confirms that model training is to be carried out and prepares to apply for authorization to train the model from the partner;

[0172] For example, the model confirmation action could be clicking the button to confirm and start training.

[0173] Optionally, the information prompt window can be a prompt window used to display information about the model to be trained to the user.

[0174] Optionally, the information prompt window may include at least one of the following: dataset information of the target dataset, dataset source of the target dataset, resource usage information of the target dataset for the dataset source, and model training trigger control.

[0175] Optionally, resource usage information can be determined from the resource usage parameters carried by the target dataset.

[0176] Optionally, the dataset information of the target dataset may include the number of datasets, the type of dataset, etc., which are not limited here.

[0177] Optionally, the source of the target dataset can be the provider of the target dataset. For example, the source of the target dataset can be the model trainer, model collaborator, etc., without limitation here.

[0178] Optionally, the model training trigger control can be a control used to trigger model training. Specifically, the model training trigger control is used to trigger the request for authorization to train the model from the model partner.

[0179] Specifically, after receiving the user's model confirmation operation, the electronic device can respond to the model confirmation operation by displaying a corresponding information prompt window on the information prompt interface. The information prompt window may include the dataset information of the target dataset, the dataset source of the target dataset, the resource usage information of the target dataset relative to the dataset source, and the model training trigger control.

[0180] Furthermore, the electronic device can determine the corresponding resource usage information based on the resource usage parameters carried by the target dataset, and display it on the information prompt interface, based on the dataset source of the target dataset.

[0181] For example, for a dataset A, which may contain 200 data entries, an electronic device needs to use one unit of central processing unit (CPU) and two units of graphics processing unit (GPU) to process these 200 data entries. Therefore, the resource usage parameters carried by the first dataset identifier A can be "CPU=1, GPU=2".

[0182] Figure 5 A schematic diagram of an information prompt interface provided in an embodiment of this disclosure is shown.

[0183] like Figure 5 As shown, after the electronic device responds to the user's model confirmation operation, a corresponding information prompt window 501 can be displayed. The information prompt window 501 includes area 502, area 503, and a model training trigger control 504. Area 502 can display dataset information of the target dataset, such as the number of dataset entries, dataset type, etc., and the dataset source of the target dataset, such as "My Data," i.e., the model training provider and "Partner 1," the model partner. Area 503 can display resource usage information of the target dataset relative to the dataset source. For example, the resource usage information of dataset B provided by the model training provider could be "CPU=1, GPU=2," and the resource usage information of dataset C provided by the model training provider could be "CPU=2, GPU=3."

[0184] Specifically, the user can trigger the model training trigger control 504. The electronic device can obtain and respond to the model training trigger operation, apply for authorization for model training from the model partner, and after receiving the application, the model partner can choose to authorize the application or not authorize the application, and provide the corresponding authorization information to the electronic device.

[0185] For example, after a user triggers the model training trigger control, the electronic device can respond to the model training trigger operation by requesting authorization from the model partner. This includes sending the model partner information such as the target dataset, the source of the target dataset, the resource usage information of the target dataset relative to the source of the dataset, and authorization / disagreement controls. Upon receiving the request, the model partner can view the basic information of the model to be trained through the target dataset information, the source of the target dataset, and the resource usage information of the target dataset relative to the source of the dataset. The model partner can also confirm whether to agree to the model training by triggering the authorization / disagreement control. The electronic device can then determine whether to train the model based on the model partner's authorization / disagreement control.

[0186] Optionally, S140 may specifically include: after the model collaborator authorizes model training, using the target dataset and target operator to train the model according to the directed acyclic graph of the modeling process, and obtain the trained model.

[0187] In this embodiment, after the electronic device requests authorization for model training from the model partner through a model training trigger operation, it can, upon receiving the authorization request returned by the model partner, perform model training using the first dataset and the target operator according to the directed acyclic graph of the modeling process, obtain the first model parameters, and send a model training request to the model partner to which the second dataset belongs, obtaining the second model parameters fed back by the model partner. The first and second model parameters are then fused to obtain the model parameters of the trained model. Simultaneously, the electronic device can display a waiting prompt window during the model training process, which can show the user information about the model training progress, allowing the user to understand the model training progress at any time and improving the user experience.

[0188] Optionally, after the electronic device requests authorization for model training from the model partner through the model training trigger operation, if it receives a response from the model partner that the request is not authorized, it displays the corresponding unauthorized request information to the model training provider. After receiving the unauthorized request information, the model training provider can perform operations such as re-applying for authorization, rebuilding the model, or abandoning model training. Thus, model training can only be carried out if the model partner agrees to authorize the model training request, thereby ensuring the security of data for all parties.

[0189] In some other embodiments, users can also upload datasets, which will be described in detail below.

[0190] Optionally, the modeling method further includes: displaying a data management interface in response to a user's data management operation; and uploading the dataset information of the dataset to be uploaded corresponding to the data upload operation to the server in response to a user's data upload operation within the data management interface. The dataset information of the dataset to be uploaded is used to update the dataset list on the server, and the dataset to be uploaded is a locally stored dataset.

[0191] In this embodiment of the disclosure, the electronic device can display a data management interface in response to a user's data management operation.

[0192] Alternatively, data management operations can be operations performed by users to manage datasets.

[0193] Alternatively, the data management interface can be an interface used to enable users to manage datasets.

[0194] Specifically, users can trigger operations on the data management controls displayed on the electronic device, i.e., perform data management operations. The electronic device can receive and respond to the data management operations and display the data management interface to the user.

[0195] Furthermore, after displaying the data management interface, the electronic device can respond to the user's data upload operation within the data management interface by uploading the dataset information of the dataset to be uploaded to the server.

[0196] Optionally, the data upload operation can be an operation in which a user uploads the dataset information of the dataset to be uploaded.

[0197] Optionally, the dataset information for the dataset to be uploaded can be dataset information used to update the dataset list on the server. This dataset information may include the number of data items in the dataset, the data type of the dataset, and the upload time of the dataset, etc., which are not limited here.

[0198] Optionally, the dataset to be uploaded can be a locally stored dataset.

[0199] Specifically, after the electronic device displays the data management interface, the user can trigger the data upload control corresponding to the dataset information of the dataset to be uploaded through the data management interface, that is, to perform a data upload operation. The electronic device can receive and respond to the data upload operation, and upload the dataset information of the dataset to be uploaded to the server. The server can update the dataset information in the dataset list according to the received dataset information of the dataset to be uploaded.

[0200] Therefore, in this embodiment of the disclosure, users can upload local dataset information of datasets to be uploaded at any time, thereby enabling the server to update the dataset list in a timely manner and thus improving the user experience.

[0201] In some other embodiments, users can also manage the authorization of uploaded datasets, which will be described in detail below.

[0202] Optionally, the modeling method further includes: in response to an authorization management operation, displaying an authorization management interface, the authorization management interface including a list of third datasets and a list of second datasets authorized by the model trainer for the model collaborator;

[0203] In response to a user's authorization modification operation on the third dataset list, adjust the data authorization information corresponding to the third dataset list according to the authorization modification result.

[0204] Upload the data authorization information corresponding to the third dataset list to the server.

[0205] In this embodiment of the disclosure, the user can also perform authorization management operations through an electronic device. The electronic device can receive and respond to the authorization management operation, displaying a corresponding authorization management interface. The authorization management interface may include a third dataset list and a fourth dataset list.

[0206] Optionally, the authorization management operation can be the operation of opening the authorization management interface for the user.

[0207] Optionally, the authorization management interface can be an interface used to display the permissions of the dataset list.

[0208] Specifically, the data in the authorization management interface for the third dataset list can be obtained by the electronic device from the local database of the model trainer; the data in the authorization management interface for the fourth dataset list can be obtained by the electronic device from the dataset information uploaded by the model partner on the server.

[0209] Optionally, the third dataset list can be a list of datasets authorized by the model trainer for the model collaborator.

[0210] Optionally, the fourth dataset list can be a list of datasets authorized by the model collaborator for the model trainer.

[0211] Figure 6 A schematic diagram of an authorization management interface provided in an embodiment of this disclosure is shown.

[0212] like Figure 6 As shown, the authorization management interface 601 may include a "I Authorize" control 602 and a "Authorize Me" control 603. The interface corresponding to the "I Authorize" control 602 may include a third dataset list, and the interface corresponding to the "Authorize Me" control 603 may include a fourth dataset list.

[0213] Specifically, when a user wants to manage the authorization of a dataset list, they can trigger the corresponding authorization management control through an electronic device to perform the authorization management operation. The electronic device can receive and respond to the user's authorization management operation and display the corresponding authorization management interface 601. Among them, the user can trigger the "I Authorize" control 602 to make the electronic device display the authorization management interface 601 that includes a third dataset list; the user can trigger the "Authorize Me" control 603 to make the electronic device display the authorization management interface 601 that includes a fourth dataset list.

[0214] Furthermore, when a user wants to manage the authorization of the local dataset list, that is, to manage the authorization of the third dataset list, they can trigger the modification authorization control corresponding to the third dataset list in the authorization management interface. That is, when the authorization modification operation is performed, the electronic device can receive and respond to the user's authorization modification operation for the third dataset list, and adjust the data authorization information corresponding to the third dataset list according to the authorization modification result.

[0215] See also Figure 6 Users can click the authorization modification control 605 corresponding to the third dataset list 604. The electronic device can receive and respond to the user's authorization modification operation, and adjust the data authorization information corresponding to the third dataset list according to the authorization modification result. For example, users can modify the authorization status of the corresponding third dataset list 604 through the authorization modification control 605. They can revoke the authorization of the third dataset list 604 that is already authorized, or allow authorization for the third dataset list 604 that is not yet authorized. The electronic device can receive the authorization modification result corresponding to this authorization modification operation and adjust the data authorization information 606 corresponding to the third dataset list 604. For example, the number of authorized parties corresponding to the data authorization information 606 can be adjusted.

[0216] Furthermore, the electronic device can upload the data authorization information corresponding to the adjusted third dataset list to the server.

[0217] Therefore, in this embodiment of the disclosure, users can modify the permissions of the third dataset list through electronic devices, ensuring the security of the use of the third dataset list and thus improving the user experience.

[0218] Figure 7 A schematic diagram of a modeling apparatus provided in an embodiment of this disclosure is shown.

[0219] In some embodiments of this disclosure, Figure 7The modeling device shown can be installed in an electronic device, which may include, but is not limited to, mobile terminals such as laptops, PDAs, PADs, etc., and fixed terminals such as digital TVs, desktop computers, etc.

[0220] like Figure 7 As shown, the modeling device 700 may include a first display module 710, an operator selection module 720, a second display module 730, and a model training module 740.

[0221] The first display module 710 can be used to display a modeling graphical interface. The modeling graphical interface includes a first area and a second area. The first area is used to provide a dataset list and an operator list. The dataset list includes a first dataset list from the model trainer and a second dataset list authorized by the model collaborator to the model trainer.

[0222] The operator selection module 720 can be used to add training data nodes corresponding to the target dataset and data processing nodes corresponding to the target operator in the second region in response to the user's selection operation. The target dataset is the dataset selected in the dataset list by the selection operation, and the target operator is the operator selected in the operator list by the selection operation.

[0223] The second display module 730 can be used to display a directed acyclic graph of the modeling process formed by training data nodes and data processing nodes in a second area in response to a user's editing operation. The training data nodes and data processing nodes in the directed acyclic graph of the modeling process have logical relationships set by the editing operation.

[0224] The model training module 740 can be used to train a model based on a directed acyclic graph of the modeling process, using the target dataset and the target operator, to obtain the trained model.

[0225] Therefore, in this embodiment of the disclosure, a modeling graphical interface can be displayed, and a first area and a second area including a dataset list and an operator list can be displayed in the modeling graphical interface. The dataset list includes a first dataset list from the model trainer and a second dataset list authorized by the model collaborator to the model trainer. Then, in response to a user's selection operation, the dataset selected in the dataset list (i.e., the training data node corresponding to the target dataset) and the operator selected in the operator list (i.e., the data processing node corresponding to the target operator) are added to the second area. Continuing to respond to a user's editing operation, a directed acyclic graph of the modeling process formed by the training data nodes and data processing nodes is displayed in the second area. The training data nodes in this directed acyclic graph are... The data nodes and data processing nodes have logical relationships set by editing operations. Finally, based on the directed acyclic graph of the modeling process, the model is trained using the target dataset and the target operator to obtain the trained model. Thus, on the displayed modeling graphical interface, training data nodes corresponding to the target dataset and data processing nodes corresponding to the target operator can be added to the second area according to the user's selection. Based on the user's editing operations, the corresponding directed acyclic graph of the modeling process is displayed in the second area. This allows the model to be trained using the target dataset and the target operator, thereby improving the efficiency of model training while ensuring the security of data from all parties, and thus enhancing the user experience.

[0226] In some embodiments of this disclosure, the target dataset may include a first dataset selected from a first dataset list and a second dataset selected from a second dataset list.

[0227] In some embodiments of this disclosure, the model training module 740 may include a first training unit, a second training unit, a parameter receiving unit, and a parameter fusion unit.

[0228] The first training unit can be used to train the model using the first dataset and the target operator according to the directed acyclic graph of the modeling process, and obtain the first model parameters.

[0229] The second training unit can be used to send a model training request to the model partner to which the second dataset belongs. The model training request carries the model training file formed by the directed acyclic graph of the modeling process. The model training request is used to enable the model partner to which the second dataset belongs to train the model based on the second dataset and the model training file to obtain the second model parameters.

[0230] This parameter receiving unit can be used to receive the second model parameters fed back by the model collaborator to which the second dataset belongs.

[0231] This parameter fusion unit can be used to fuse the first model parameters and the second model parameters to obtain the model parameters of the trained model.

[0232] In some embodiments of this disclosure, the first model parameter and the second model parameter may each include at least a portion of the model parameters.

[0233] In some embodiments of this disclosure, the model training file can be formed based on a target directed acyclic subgraph, which can be composed of target nodes related to the second dataset and the logical relationships between the target nodes in the directed acyclic graph of the modeling process.

[0234] In some embodiments of this disclosure, the model training module 740 may further include a first display unit.

[0235] The first display unit can be used to display a waiting prompt window after sending a model training request to the model partner to which the second dataset belongs. The waiting prompt window can contain waiting prompt information and a return trigger control, which can be used to trigger the cancellation of the display of the waiting prompt window.

[0236] In some embodiments of this disclosure, the modeling apparatus 700 may further include a third display module and a fourth display module.

[0237] The third display module can be used to display a data statistics page in response to the user's data viewing operation. The data statistics page includes statistical information of at least one model partner.

[0238] The fourth display module can be used to respond to a user's viewing operation of the target partner statistics in at least one partner statistics information, and display the authorized data statistics of the target model partner to which the target partner statistics belong. The authorized data statistics are used to show the historical data usage of the target model partner to the model training party. The historical data usage is obtained based on the statistics of the historical model training process.

[0239] In some embodiments of this disclosure, the modeling apparatus 700 may further include a data filtering module.

[0240] This data filtering module can be used to filter the dataset list according to the filtering method corresponding to the data filtering operation, in response to the user's data filtering operation on the dataset list, before adding the training data node corresponding to the target dataset and the data processing node corresponding to the target operator in the second area.

[0241] In some embodiments of this disclosure, the selection operation may include a dataset drag operation and an operator drag operation. The dataset drag operation may include the operation of dragging a target dataset from a first region to a second region, and the operator drag operation may include the operation of dragging a target operator from a first region to a second region.

[0242] In some embodiments of this disclosure, the model training module 740 may further include a third training unit.

[0243] This third training unit can be used to respond to the user's model training trigger operation, and to train the model using the target dataset and target operator according to the directed acyclic graph of the modeling process, so as to obtain the trained model.

[0244] In some embodiments of this disclosure, the model training module 740 may further include a second display unit.

[0245] The second display unit can be used to display an information prompt window in response to the model confirmation operation. The information prompt window includes at least one of the following: dataset information of the target dataset, dataset source of the target dataset, resource usage information of the target dataset for the dataset source, and a model training trigger control. The resource usage information is determined by the resource usage parameters carried by the target dataset. The model training trigger operation is a trigger operation on the model training trigger control, which is used to trigger the application for authorization to train the model from the model partner.

[0246] In some embodiments of this disclosure, the model training module 740 can also be specifically used to train the model using the target dataset and target operator according to the directed acyclic graph of the modeling process after the model collaborator authorizes the model training, so as to obtain the trained model.

[0247] In some embodiments of this disclosure, the modeling apparatus 700 may further include a fifth display module and a first upload module.

[0248] The fifth display module can be used to display the data management interface in response to user data management operations.

[0249] The first upload module can be used to respond to the user's data upload operation in the data management interface, and upload the dataset information of the dataset to be uploaded corresponding to the data upload operation to the server. The dataset information of the dataset to be uploaded is used to update the dataset list on the server. The dataset to be uploaded is a dataset stored locally.

[0250] In some embodiments of this disclosure, the modeling apparatus 700 may further include a sixth display module, an authorization modification module, and a second upload module.

[0251] The sixth display module can be used to display the authorization management interface in response to authorization management operations. The authorization management interface includes a list of third datasets and a list of second datasets that the model trainer has authorized for the model collaborator.

[0252] This authorization modification module can be used to respond to user authorization modification operations on the third dataset list, and adjust the data authorization information corresponding to the third dataset list according to the authorization modification result.

[0253] The second upload module can be used to upload data authorization information corresponding to the third dataset list to the server.

[0254] It should be noted that, Figure 7 The modeling device 700 shown can perform Figures 1 to 6 The various steps in the method embodiment shown are implemented. Figures 1 to 6 The various processes and effects in the method embodiments shown are not elaborated here.

[0255] Figure 8 A schematic diagram of the structure of a modeling device provided in an embodiment of this disclosure is shown.

[0256] In some embodiments of this disclosure, Figure 8 The modeling device shown can be any electronic device that the user wants to use for modeling operations. This electronic device can include, but is not limited to, mobile terminals such as laptops, PDAs (Personal Digital Assistants), and PADs (Tablet PCs), as well as fixed terminals such as digital TVs and desktop computers.

[0257] like Figure 8 As shown, the modeling device may include a processor 801 and a memory 802 storing computer program instructions.

[0258] Specifically, the processor 801 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0259] Memory 802 may include a mass storage device for information or instructions. For example, and not limitingly, memory 802 may include a hard disk drive (HDD), a floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or a Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 802 may include removable or non-removable (or fixed) media. Where appropriate, memory 802 may be internal or external to the integrated gateway device. In a particular embodiment, memory 802 is a non-volatile solid-state memory. In a particular embodiment, memory 802 includes read-only memory (ROM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (Electrically Programmable ROM, EPROM), an electrically erasable programmable PROM (EEPROM), an electrically alterable ROM (EAROM), or flash memory, or a combination of two or more of these.

[0260] The processor 801 performs the steps of the modeling method provided in the embodiments of this disclosure by reading and executing computer program instructions stored in the memory 802.

[0261] In one example, the modeling device may also include a transceiver 803 and a bus 804. Wherein, as... Figure 8 As shown, the processor 801, memory 802 and transceiver 803 are connected via bus 804 and communicate with each other.

[0262] Bus 804 includes hardware, software, or both. For example, and not limitingly, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industrial Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a MicroChannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local Bus (VLB) bus, or other suitable buses, or a combination of two or more of these. Where appropriate, bus 804 may include one or more buses. Although specific buses are described and illustrated in the embodiments of this application, this application considers any suitable bus or interconnection.

[0263] This disclosure also provides a computer-readable storage medium that can store a computer program that, when executed by a processor, causes the processor to implement the modeling method provided in this disclosure.

[0264] The aforementioned storage medium may, for example, include a memory 802 containing computer program instructions, which can be executed by a processor 801 of the modeling device to complete the modeling method provided in the embodiments of this disclosure. Optionally, the storage medium may be a non-transitory computer-readable storage medium, such as a ROM, random access memory (RAM), compact disc ROM (CD-ROM), magnetic tape, floppy disk, and optical data storage device.

[0265] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the term "comprising" is intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus.

[0266] The above description is merely a specific embodiment of this disclosure, enabling those skilled in the art to understand or implement it. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this disclosure. Therefore, this disclosure is not to be limited to the embodiments described herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A modeling method, characterized in that, include: The modeling graphical interface is displayed, which includes a first area and a second area. The first area is used to provide a dataset list and an operator list. The dataset list includes a first dataset list of the model trainer and a second dataset list authorized by the model collaborator to the model trainer. In response to the user's selection operation, training data nodes corresponding to the target dataset and data processing nodes corresponding to the target operator are added in the second area. The target dataset is the dataset selected by the selection operation in the dataset list, and the target operator is the operator selected by the selection operation in the operator list. In response to the user's editing operation, a directed acyclic graph of the modeling process formed by the training data nodes and the data processing nodes is displayed in the second area, wherein the training data nodes and the data processing nodes in the directed acyclic graph of the modeling process have a logical relationship set by the editing operation; Based on the directed acyclic graph of the modeling process, the target dataset and the target operator are used to train the model locally to obtain the trained model.

2. The method according to claim 1, characterized in that, The target dataset includes a first dataset selected from the first dataset list and a second dataset selected from the second dataset list; The step of training the model using the target dataset and the target operator based on the directed acyclic graph of the modeling process to obtain the trained model includes: Based on the directed acyclic graph of the modeling process, the first model parameters are obtained by training the model using the first dataset and the target operator. A model training request is sent to the model partner to which the second dataset belongs. The model training request carries the model training file formed by the directed acyclic graph of the modeling process. The model training request is used to enable the model partner to which the second dataset belongs to train the model based on the second dataset and the model training file to obtain the second model parameters. Receive the second model parameters from the model collaborator to which the second dataset belongs; The first model parameters and the second model parameters are fused to obtain the model parameters of the trained model.

3. The method according to claim 2, characterized in that, The first model parameter and the second model parameter each include at least a portion of the model parameters; The model training file is formed based on a target directed acyclic subgraph, which consists of the target nodes related to the second dataset and the logical relationships between the target nodes in the directed acyclic graph of the modeling process.

4. The method according to claim 2, characterized in that, After sending the model training request to the model collaborator to which the second dataset belongs, the method further includes: A waiting prompt window is displayed, which includes a waiting prompt message and a return trigger control. The return trigger control is used to cancel the display of the waiting prompt window.

5. The method according to claim 1, characterized in that, The method further includes: In response to the user's data viewing operation, a data statistics page is displayed, which includes at least one of the model's partner's partner statistics information; In response to a user's viewing operation of target partner statistics in at least one of the partner statistics, the authorized data statistics corresponding to the target model partner to which the target partner statistics belong are displayed. The authorized data statistics are used to show the historical data usage quantity authorized by the target model partner to the model training party.

6. The method according to claim 1 or 5, characterized in that, Before adding the training data node corresponding to the target dataset and the data processing node corresponding to the target operator in the second region in response to the user's selection operation, the method further includes: In response to a user's data filtering operation on the dataset list, the dataset list is filtered according to the filtering method corresponding to the data filtering operation.

7. The method according to claim 1, characterized in that, The selection operation includes a dataset drag operation and an operator drag operation. The dataset drag operation includes dragging the target dataset from the first region to the second region, and the operator drag operation includes dragging the target operator from the first region to the second region.

8. The method according to claim 1, characterized in that, The step of training a model using the target dataset and the target operator based on the directed acyclic graph in the modeling process to obtain the trained model includes: In response to the user's model training trigger operation, the model is trained using the target dataset and the target operator according to the directed acyclic graph of the modeling process, and the trained model is obtained.

9. The method according to claim 8, characterized in that, Before the model training trigger operation in response to the user, and before training the model using the target dataset and the target operator according to the directed acyclic graph of the modeling process to obtain the trained model, the method further includes: In response to the model confirmation operation, an information prompt window is displayed. The information prompt window includes at least one of the following: dataset information of the target dataset, dataset source of the target dataset, resource consumption information of the target dataset for the dataset source, and model training trigger control. The model training triggering operation is a triggering operation of the model training triggering control, which is used to trigger the application for authorization to train the model from the model partner; The step of training the model using the target dataset and the target operator based on the directed acyclic graph of the modeling process to obtain the trained model includes: After the model partner authorizes model training, the model is trained using the target dataset and the target operator according to the directed acyclic graph of the modeling process, and the trained model is obtained.

10. The method according to claim 1, characterized in that, The method further includes: The data management interface is displayed in response to user data management operations. In response to a user's data upload operation within the data management interface, the dataset information of the dataset to be uploaded corresponding to the data upload operation is uploaded to the server. The dataset information of the dataset to be uploaded is used to enable the server to update the dataset list. The dataset to be uploaded is a locally stored dataset.

11. The method according to claim 1, characterized in that, The method further includes: In response to the authorization management operation, the authorization management interface is displayed, which includes a list of third datasets authorized by the model trainer for the model collaborator and a list of second datasets. In response to a user's authorization modification operation on the third dataset list, the data authorization information corresponding to the third dataset list is adjusted according to the authorization modification result corresponding to the authorization modification operation; Upload the data authorization information corresponding to the third dataset list to the server.

12. A modeling apparatus, characterized in that, include: The first display module is used to display the modeling graphical interface. The modeling graphical interface includes a first area and a second area. The first area is used to provide a dataset list and an operator list. The dataset list includes a first dataset list of the model trainer and a second dataset list authorized by the model collaborator to the model trainer. The operator selection module is used to add training data nodes corresponding to the target dataset and data processing nodes corresponding to the target operator in the second region in response to the user's selection operation. The target dataset is the dataset selected by the selection operation in the dataset list, and the target operator is the operator selected by the selection operation in the operator list. The second display module is used to respond to the user's editing operation and display a directed acyclic graph of the modeling process formed by the training data nodes and the data processing nodes in the second area, wherein the training data nodes and the data processing nodes in the directed acyclic graph of the modeling process have a logical relationship set by the editing operation. The model training module is used to train the model locally using the target dataset and the target operator according to the directed acyclic graph of the modeling process, so as to obtain the trained model.

13. A modeling device, characterized in that, include: processor; Memory, used to store executable instructions; The processor is configured to read the executable instructions from the memory and execute the executable instructions to implement the modeling method according to any one of claims 1-11.

14. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, causes the processor to implement the modeling method described in any one of claims 1-11.

Images