Data model acquisition method and apparatus, and product

Abstract

A data model acquisition method and apparatus, and a product, which relate to the technical field of big data analysis, and are used for achieving capabilities such as autonomously completing model acquisition or optimization on the basis of different target mining intentions. The method comprises: receiving an input first target mining intention, and acquiring first training data from a sample database on the basis of the input first target mining intention; on the basis of a plurality of data model training algorithms, training the first training data to obtain a plurality of first data models, and displaying the plurality of first data models; and determining first data models sequentially selected from the plurality of displayed first data models, and verifying whether the sequentially selected first data models are first target data models, wherein a first target user list generated on the basis of the first target data models meets user requirements, the first target user list is obtained from a plurality of pieces of output data, which is from the first target data models after respectively inputting a plurality of pieces of data in a historical database into the first target data models, and the historical database stores all attribute features of a user using a network.

Description

A method, apparatus and product for acquiring a data model

[0001] Cross-references to related applications

[0002] This application claims priority to Chinese Patent Application No. 202410396194.5, filed on March 30, 2024, entitled “A Data Model Acquisition Method, Apparatus and Product”, the entire contents of which are incorporated herein by reference. Technical Field

[0003] This application relates to the field of big data analytics technology, and in particular to a data model acquisition method, apparatus, and product. Background Technology

[0004] Traditional operators often need to analyze users' network usage during network management and sales to identify target users. These target users can include high-value potential customers and potential churned customers with poor service experience, in order to better plan their operations.

[0005] Currently, operators typically obtain target user lists through manual collaboration with network operations and maintenance (O&M) teams. These teams analyze user network usage data and develop relevant algorithms and code to create corresponding data models, enabling customers to obtain target user lists based on these models.

[0006] However, this method requires multiple back-and-forth interactions between operators and network maintenance providers, resulting in a lengthy data model design process, high development costs, and a relatively simple data model that cannot be adjusted in a timely manner according to the current needs of customers, exhibiting poor adaptability and flexibility. Summary of the Invention

[0007] This application provides a data model acquisition method, apparatus, and product to enable operators to autonomously acquire or adjust models based on different target mining intentions, and to acquire target lists based on determined models.

[0008] Firstly, embodiments of this application provide a method for obtaining a data model. The method includes:

[0009] The system receives a first target mining intent, retrieves first training data from a sample database based on this intent (the first target mining intent is used to mine target users), and then trains corresponding models on the first training data using preset multiple data model training algorithms to obtain multiple first data models, which are then displayed. Next, the system determines the first data model selected sequentially from the displayed multiple first data models and verifies whether the sequentially selected first data model is the first target data model. The first target user list generated based on the first target data model satisfies user needs. This first target user list is obtained by inputting multiple data entries from a historical database into the first target data model and then outputting multiple data entries from the first target data model. The historical database stores attribute features of all users using the network, and the sample database stores data labeled with the target users for model training.

[0010] Based on the above method, this application embodiment can automatically generate multiple possible data models with corresponding mining intentions by combining historical data with the target mining intent input by the operator user. This enables efficient mining and expansion based on mining intent. Operators can select models from the generated multiple data models according to actual needs. Furthermore, by adopting an automated modeling approach, no manual intervention is required, eliminating the need for data analysis and engineering development departments. This effectively reduces modeling time and costs, and allows for timely model optimization based on the operator's needs, making it more flexible, convenient, and practical. In addition, this application embodiment can also generate a corresponding user list in real time based on the selected data model after the operator selects it, allowing the operator to determine whether model optimization is needed based on the displayed user list, further enhancing flexibility.

[0011] In one possible implementation, the method further includes:

[0012] The first target data model and the first target mining intent are stored in the model database. When the subsequent input target mining intent is the first target mining intent, the first target data model is obtained from the model database based on the first target mining intent. Then, a second target user list is generated based on the first target data model. The second target user list is obtained by inputting multiple data from the historical database into the first target data model and then outputting multiple data from the first target data model. The historical database is a database obtained by updating the historical database.

[0013] It is understood that the historical database described in this application embodiment can be updated according to actual conditions or operator needs.

[0014] As an example, the user list generated by the data model in this application embodiment can be generated by the data model based on the data information in the current historical database; or, the user list generated by the data model can be generated by the data model based on the data information in the historical database corresponding to a certain period or time range specified by the operator.

[0015] Based on the above method, after determining the target data model, the embodiments of this application can store the target data model in the corresponding model database. Thus, when the operator wants to search for previously used mining intentions to mine user information again, it can directly retrieve and apply the target data model corresponding to the previously used mining intentions from the model database, which can effectively save system overhead and make mining efficiency more efficient.

[0016] In one possible implementation, the method further includes:

[0017] When the first target data model is not present among the multiple data models, a second target mining intent is received. Then, based on the second target mining intent, second training data is obtained from the sample database. The second target mining intent is used to mine the target user. Then, based on a preset set of multiple data model training algorithms, corresponding model training is performed on the second training data to obtain multiple second data models, and the multiple second data models are displayed. The second data model selected sequentially from the displayed multiple second data models is determined, and it is verified whether the sequentially selected second data model is the second target data model. The third target user list generated based on the second target data model satisfies the user requirements. The third target user list is obtained by inputting multiple data from the historical database into the second target data model and then outputting multiple data from the second target data model. The second target mining intent is obtained by modifying the first target mining intent.

[0018] Based on the above method, this application provides a model optimization scheme. When the required target data model cannot be obtained from multiple displayed data models, the operator can appropriately adjust or modify the input target mining intent, and then search for the target data model corresponding to the modified mining intent again. In this way, the final target data model can be found based on similar mining intents, which is more flexible.

[0019] In one possible implementation, obtaining the first training data from the sample database based on the first target mining intent includes: determining the tag of the target user corresponding to the first target mining intent, and / or, the target user using the attribute features of the network;

[0020] The first training data is obtained by retrieving data from the sample database containing the tags of the identified target users and / or the attribute features.

[0021] In one possible implementation, the number of data model training algorithms is related to the number of attribute features contained in the multiple sample data stored in the sample database.

[0022] In one possible implementation, the method further includes: displaying the first target user list generated based on the first target data model; and / or, displaying analysis information obtained by comparing the first target user list with the original user list; wherein the original user list is obtained by inputting multiple data from the historical database into the original data model and then outputting multiple data from the original data model; wherein the original data model is obtained from data models already stored in the data model library based on the first target mining intent.

[0023] Based on the above method, a user list generated based on the data model can be displayed, as well as analytical information obtained after analyzing and processing the user list. This analytical information can be obtained by analyzing the user list, and may include relevant attribute characteristics of the user list, the proportion of users in the user list, etc. Alternatively, the analytical information can be obtained by comparing the user list with the original user list, thereby effectively enriching the content displayed on the interface and enabling operators to more conveniently and quickly understand relevant information about the selected data model, resulting in a high level of user experience.

[0024] In one possible implementation, the first target user list satisfying the user requirements can be achieved by receiving an instruction to confirm that the displayed first target user list and / or the displayed analysis information meet the user requirements.

[0025] In one possible implementation, the target mining intent is described using natural language. Based on this approach, users with mining needs can easily and conveniently input their mining intent without needing to understand code compilation, making it more widely applicable and suitable for more scenarios.

[0026] Secondly, a data model acquisition device is provided. The device includes:

[0027] The input / output module is used to receive a first target mining intent, which is used to mine target users;

[0028] The processing module is used to obtain first training data from a sample database based on the first target mining intent; the sample database stores data labeled with the target user for model training; the first training data is used to perform corresponding model training based on a variety of preset data model training algorithms to obtain multiple first data models; the input / output module is also used to display the multiple first data models.

[0029] The processing module is further configured to determine the first data model selected sequentially from the plurality of first data models displayed, and verify whether the sequentially selected first data model is the first target data model; wherein, the first target user list generated based on the first target data model meets the user requirements, and the first target user list is obtained by inputting multiple data from the historical database into the first target data model and then outputting multiple data from the first target data model; the historical database stores the attribute characteristics of all users using the network.

[0030] In one possible implementation, the processing module is further configured to: store the first target data model and the first target mining intent in a model database; when the target mining intent received by the input / output module is the first target mining intent, obtain the first target data model from the model database based on the first target mining intent; generate a second target user list based on the first target data model, wherein the second target user list is obtained by inputting multiple data from the historical database into the first target data model and then outputting multiple output data based on the first target data model.

[0031] In one possible implementation, the processing module is further configured to:

[0032] When the first target data model is not present among the multiple data models, the second target mining intention is received through the input / output module; the second target mining intention is obtained by modifying the first target mining intention; second training data is obtained from the sample database based on the second target mining intention, and the second target mining intention is used to mine the target user; the second training data is trained according to the corresponding model based on the preset multiple data model training algorithms to obtain multiple second data models, and the multiple second data models are displayed through the input / output module; the second data model selected sequentially among the displayed multiple second data models is determined, and it is verified whether the sequentially selected second data model is the second target data model; wherein, the third target user list generated based on the second target data model satisfies the user requirements, and the third target user list is obtained by inputting multiple data from the historical database into the second target data model and then outputting multiple output data based on the second target data model.

[0033] In one possible implementation, when the processing module obtains the first training data from the sample database based on the first target mining intent, it is specifically used for:

[0034] Determine the tag of the target user corresponding to the first target mining intent, and / or, the target user uses the attribute features of the network;

[0035] The first training data is obtained by retrieving data from the sample database containing the tags of the identified target users and / or the attribute features.

[0036] In one possible implementation, the number of data model training algorithms is related to the number of attribute features contained in the multiple sample data stored in the sample database.

[0037] In one possible implementation, the input / output module is further configured to: display the first target user list generated based on the first target data model; and / or, display analysis information obtained by comparing the first target user list with the original user list; wherein the original user list is obtained by inputting multiple data entries from the historical database into the original data model and then outputting multiple data entries based on the original data model; wherein the original data model is obtained from data models already stored in the data model library based on the first target mining intent.

[0038] In one possible implementation, the processing module determines that the first target user list meets the user requirements by receiving an instruction from the input / output module to confirm that the displayed first target user list and / or the displayed analysis information meet the user requirements.

[0039] In one possible implementation, the target mining intent is described based on natural language.

[0040] Thirdly, embodiments of this application provide a data model acquisition apparatus, the apparatus including a processor and a memory. The memory is used to store program code. The processor is used to read and execute the program code stored in the memory to implement the method as described in the first aspect or any design of the first aspect.

[0041] Fourthly, this application also provides a computer storage medium. The storage medium stores a software program, which, when read and executed by one or more processors, can implement any of the methods provided in the first aspect.

[0042] Fifthly, embodiments of this application provide a computer program product containing instructions that, when run on a computer, cause the computer to perform the method provided in any of the designs of the first aspect described above.

[0043] Sixthly, embodiments of this application provide a chip including a processor. The processor is configured to perform the methods provided in any of the designs of the first aspect.

[0044] In one possible design, the chip also includes a communication interface that is coupled to the processor.

[0045] In one possible design, the chip is connected to a memory for reading and executing software programs stored in the memory to implement the method provided by any of the designs in the first aspect.

[0046] Based on the implementations provided in the foregoing aspects, this application can be further combined to provide more implementations. Furthermore, the technical effects achievable by any of the second to sixth aspects of this application can be referred to the description of the technical effects achievable by the first aspect and its various implementations. Attached Figure Description

[0047] Figure 1 is a schematic diagram of an existing data model acquisition method;

[0048] Figure 2 is a schematic diagram of the first data model acquisition system architecture provided in this application embodiment;

[0049] Figure 3 is a schematic diagram of the second data model acquisition system architecture provided in the embodiments of this application;

[0050] Figure 4 is a schematic diagram of the third data model acquisition system architecture provided in the embodiments of this application;

[0051] Figure 5 is a schematic diagram of the fourth data model acquisition system architecture provided in the embodiments of this application;

[0052] Figure 6 is a schematic diagram of the scenario architecture for the application of the data model acquisition system provided in the embodiments of this application;

[0053] Figure 7 is a schematic diagram of a data model acquisition method provided in an embodiment of this application;

[0054] Figure 8 is a schematic diagram of an intent-aware and translation engine training and construction process provided in an embodiment of this application;

[0055] Figure 9 is a schematic diagram of a data model generation and management engine training and construction process provided in an embodiment of this application;

[0056] Figure 10 is a flowchart illustrating an operator preparation phase according to an embodiment of this application;

[0057] Figure 11 is a flowchart illustrating the operation phase of an operator application according to an embodiment of this application;

[0058] Figure 12 is a schematic diagram of a user list display provided in an embodiment of this application;

[0059] Figure 13 is a schematic diagram of another user list display provided in an embodiment of this application;

[0060] Figure 14 is a schematic diagram of a business data import and storage process provided in an embodiment of this application;

[0061] Figure 15 is a schematic diagram of a process for obtaining training data based on target mining intent provided in an embodiment of this application;

[0062] Figure 16 is a schematic diagram of a process for generating multiple data models according to an embodiment of this application;

[0063] Figure 17 is a schematic diagram of a user list generation process provided in an embodiment of this application;

[0064] Figure 18 is a schematic diagram of a data model storage process provided in an embodiment of this application;

[0065] Figure 19 is a schematic diagram of a data model acquisition device provided in an embodiment of this application;

[0066] Figure 20 is a schematic diagram of another data model acquisition device provided in an embodiment of this application. Detailed Implementation

[0067] The embodiments of this application can be applied to various data model mining scenarios. For example, this application can be used to build intelligent, local adaptive potential customer mining and marketing scenarios.

[0068] To facilitate understanding of the embodiments of this application, the concepts of the technical terms involved are first introduced below:

[0069] (1) Potential customers refer to a group of users with specific data usage habits who have a tendency to purchase new consumption packages or upgrade packages based on their historical consumption records and network traffic usage data. Such users are called potential customers.

[0070] (2) Poor quality: In the field of network communication, based on user network traffic usage data and profiles, some traffic usage data, such as uplink and downlink speeds, differ from the quality promised in the user's purchased service plan. For example, the maximum uplink speed is lower than the speed promised in the service plan. This difference is referred to as poor quality.

[0071] (3) Operator / Customer In this application, operator and customer have the same meaning, which are users / consumers of the goods.

[0072] (4) Large Language Models (LLMs) are artificial intelligence models designed to understand and generate human language. They are trained on large amounts of text data to learn the ability to understand and generate human language and can perform a wide range of tasks, including text summarization, translation, sentiment analysis, and more.

[0073] The core idea of ​​LLM is to learn patterns and language structures in natural language through large-scale unsupervised training, which can simulate human language cognition and generation processes to some extent. Compared with traditional NLP models, LLM can better understand and generate natural text, while also demonstrating a certain degree of logical thinking and reasoning ability.

[0074] In the description of this application, unless otherwise stated, "multiple" refers to two or more. Additionally, " / " indicates that the related objects are in an "or" relationship; for example, A / B can represent A or B. "And / or" in this application merely describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone, where A and B can be singular or plural. Furthermore, to facilitate a clear description of the technical solutions of the embodiments of this application, the terms "first" and "second" are used in the embodiments to distinguish identical or similar items with essentially the same function and effect. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and that "first" and "second" are not necessarily different. It should also be noted that, unless specifically stated, the specific description of some technical features in one embodiment can also be used to explain the corresponding technical features mentioned in other embodiments.

[0075] The following description, in conjunction with the accompanying drawings, illustrates exemplary embodiments of this application, including various details to aid understanding. These embodiments should be considered merely exemplary. Therefore, those skilled in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope of this application. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description. It should be noted that certain software, components, models, and other existing industry solutions may be mentioned in the embodiments of this application. These should be considered exemplary, intended only to illustrate the feasibility of implementing the technical solutions of this application, and do not imply that the applicant has already used or necessarily used the aforementioned solutions.

[0076] The acquisition, transmission, storage, and use of data in this application all comply with the requirements of relevant national laws and regulations.

[0077] Currently, operators typically obtain target user lists through manual collaboration with network operations and maintenance (O&M) teams. These O&M teams analyze user network usage data and develop corresponding algorithms and code to create data models, allowing customers to obtain target user lists based on these models. For example, the current target user list acquisition process is shown in Figure 1. First, the operator proposes target user analysis requirements based on business characteristics and communicates these requirements and related business features to the business department manually. Second, the development and data analysis department designs a data model based on the requirements and business description. This step may involve expert-centric approaches, customizing target user matching templates based on the operator's requirements and expert experience; or data analyst-centric approaches, analyzing data based on the operator's user business needs to obtain a data model corresponding to those needs. Third, the development and engineering department develops and trains the data model. This step requires the development and engineering department to develop and train the data model based on the model established in the previous step, resulting in a trained data model. Finally, the user list is obtained based on the trained data model and returned to the user. Users can then provide feedback based on their satisfaction with the results or modify their requirements for a re-analysis and iteration of the first three steps.

[0078] As can be seen from the above introduction, the data model obtained based on this method has a long design process, high development cost, and the final data model can only handle the analysis of the specific type of target user. It cannot be adjusted in a timely manner according to the current needs of the customer, and its adaptability and flexibility are poor.

[0079] To address the aforementioned issues, this application provides a data model acquisition method, enabling operators to autonomously acquire or adjust models based on different target mining intentions, and to acquire target lists based on determined models.

[0080] The technical solutions in the embodiments of this application will now be described with reference to the accompanying drawings.

[0081] Referring to Figure 2, this is a schematic diagram of the system architecture for a data model acquisition method provided in this application embodiment. The system architecture may include a UI operation layer 100, a data analysis layer 200, and a data storage layer 300. Through inter-layer interaction, an intelligent data model mining and dynamic optimization system is constructed.

[0082] UI operation layer 100 is used to receive mining requests from the user, such as a first target mining intent input by an operator; display multiple data models generated based on the first target mining intent; and display a corresponding user list generated from a data model selected from the multiple data models, wherein the user list is obtained by inputting multiple data entries from a historical database into the selected data model and then outputting multiple data entries based on the selected data model.

[0083] As an example, the target users described in this application embodiment may include potential customer information (also known as potential customers) and / or, poor-quality customer information (also known as poor-quality customers), which can be simply referred to as potential customer users and poor-quality users. For ease of introduction and understanding, the following description will use potential customer users and / or poor-quality users as target users.

[0084] As an example, as shown in Figure 3, the UI operation layer 100 in this embodiment may include an intent setting module 110, a data model setting module 120, and a review module 130.

[0085] The intent setting module 110 is used to receive a first target mining intent input by the mining demander; it is also used to receive a second target mining intent input, the second target mining intent being obtained by modifying the first target mining intent, the first target mining intent being used to mine target users.

[0086] For example, the operator can input the target mining intent in the operation box corresponding to the system display interface in the intent setting module 110, or modify the input target mining intent.

[0087] As an example, the target mining intent input in the embodiments of this application can be based on a natural language description.

[0088] The data model setting module 120 is used to display the plurality of data models, which are obtained by training corresponding models on the acquired training data based on a plurality of preset data model training algorithms; it is also used to select a data model from the plurality of data models or modify the selected data model based on the operator's selection instruction.

[0089] For example, the operation box corresponding to the data model setting module in the system's display interface can display multiple data models. Operators can select a data model from the multiple displayed data models or modify the input target mining intent within the operation box corresponding to the data model setting module.

[0090] The review module 130 is used to display the corresponding user list generated based on the data model selected by the operator. The user list is obtained by inputting multiple data from the historical database into the selected data model and then outputting multiple data based on the selected data model. Alternatively, it can also display analysis information obtained by comparing the user list generated based on the selected data model with the original user list.

[0091] The original user list is obtained by inputting multiple data entries from the historical database into the original data model and then outputting multiple data entries based on the original data model. The original data model is obtained from the data models already stored in the data model library based on the first target mining intent.

[0092] The data reports generated in this application embodiment can be visual data tables or data charts, and are not limited thereto.

[0093] As an example, in this embodiment of the application, after determining that the user list or analysis information displayed in the review module 130 meets the operator's requirements, the operator can store the corresponding data model that generated the user list in the data model library in the data storage layer. For example, a first target data model is stored, wherein a first target user list generated based on the first target data model meets the user's requirements. The first target user list is obtained by inputting multiple data entries from the historical database into the first target data model and then outputting multiple data entries from the first target data model.

[0094] The data analysis layer 200 is used to obtain first training data from the sample database based on the first target mining intent; to perform corresponding model training on the first training data based on preset multiple data model training algorithms to obtain multiple data models; and to generate a corresponding user list or the analysis information based on the data model selected from the multiple data models.

[0095] As an example, as shown in Figure 4, the data analysis layer 200 in this embodiment may include two sub-layers: an intent perception and data model generation layer 210 and a data information generation layer 220.

[0096] The intent-aware and data model generation layer 210 may include an intent-aware and translation engine 211 and a data model management and generation engine 212.

[0097] The intent perception and translation engine 211 is used to perceive and translate the input target mining intent, extract keywords of the target mining intent, such as the name of the data model to be generated, or the relevant attribute features expected to be used for algorithm training, and then obtain corresponding training data from the sample database based on the content obtained by perceiving and translating the target mining intent.

[0098] The data model management and generation engine 212 is used to train the corresponding models on the training data corresponding to the input target mining intention based on a variety of preset data model training algorithms, so as to obtain multiple data models.

[0099] The data information generation layer 220 is used to obtain the corresponding user list from the historical database in the data storage layer based on the data model selected by the user; it can also be used to compare and analyze the user list generated based on the selected data model with the original user list to obtain analysis information.

[0100] The data storage layer is used to store historical data, sample data, data attribute field knowledge, and data models, etc. The sample database stores data with the target user's tag for model training, and the historical database stores the attribute features of all users using the network.

[0101] As an example, as shown in Figure 5, the data storage layer in this embodiment may include multiple databases, such as a data attribute field knowledge base, a sample database, a historical database, and an existing data model library.

[0102] For example, if the operator confirms that the selected data model will be included in the database, the corresponding data model will be stored in the data model library in the data storage layer.

[0103] In practical applications, the system architecture described in Figures 2-5 can be used as a whole for site operation and maintenance, that is, the operation components of the overall network equipment designed in Figures 2-5 can be packaged together for application. Alternatively, the embodiments of this application can also package some of the components in Figures 2-5 according to the needs of operators and use them as an independent operation engine adapted to network equipment.

[0104] Referring to Figure 6, this is a schematic diagram of a scenario architecture based on the above system architecture, which is used as an independent application of the operation engine of the network device provided by the user, according to an embodiment of this application. The scenario architecture includes an upper-layer intelligent BI framework and a lower-layer data storage platform.

[0105] The upper-layer intelligent BI framework is an independent application in the system architecture provided in this application embodiment for data model acquisition and optimization, serving as an adaptation network device operation engine. Corresponding to the user layer and data analysis layer in the above system architecture, it mainly involves receiving the target mining intent input by the operator, translating the target mining intent and generating multiple corresponding data models, and interactively generating a target user list based on the data model selected by the operator from the multiple data models and the database under the operator's adaptive operation scenario, and presenting the generated user list and the generated data model to the user.

[0106] The lower-level data storage platform interacts with the upper-level intelligent BI framework to return corresponding training data, new potential customer / poor quality query data, etc., and records and stores the potential customer / poor quality data model confirmed by the user.

[0107] The system described in this application embodiment can run on a cloud computing device system (which may include at least one cloud computing device, such as a server), or on a physical server, or on an edge computing device system (which may include at least one edge computing device, such as a server, desktop computer, etc.), or on various terminal computing devices, such as laptops, personal desktop computers, mobile phones, tablets, ultra-mobile personal computers (UMPCs), netbooks, personal digital assistants (PDAs), etc.

[0108] For example, the system described in this application embodiment can logically be composed of various parts, and each component of the system can run on different systems or servers. For instance, each part of the system can run on any two of a cloud computing device system, an edge computing device system, and a terminal computing device. The cloud computing device system, the edge computing device system, and the terminal computing device are connected by a communication path, enabling them to communicate and transmit data with each other. Based on this, the data model acquisition method provided in this application embodiment, by adopting an automated modeling approach, eliminates the need for manual intervention, effectively reducing modeling time and costs, and allows for timely model optimization based on the operator's needs, making it more flexible, convenient, and practical.

[0109] Referring to Figure 7, which is a schematic diagram of a data model acquisition method provided in an embodiment of this application, it can be applied to the system or architecture shown in Figures 2 to 6. The executing entity can be the data model acquisition system shown in Figures 2-5, and the specific steps include:

[0110] S701: Receive the first target mining intent input. As an example, in this application embodiment, the first target mining intent is used to mine target users.

[0111] The target users described in this application embodiment include, but are not limited to, potential high-value customer information and / or, poor-quality customer information.

[0112] As an example, the target mining intent described in this application embodiment is based on natural language. Therefore, it allows users with mining needs to easily and conveniently input their mining intent without needing to understand code compilation, making it more widely applicable and suitable for more scenarios.

[0113] S702: Obtain first training data from the sample database based on the first target mining intent. The sample database stores data labeled with the target user for model training.

[0114] As an example, the first training data can be obtained in the following manner in an embodiment of this application:

[0115] For example, embodiments of this application may determine the tag of the target user corresponding to the first target mining intent, and / or the attribute features of the first target user using the network, and then obtain the first training data containing the determined tag of the target user and / or the attribute features from the sample database.

[0116] S703: Based on a preset set of multiple data model training algorithms, perform corresponding model training on the first training data to obtain multiple first data models, and display the multiple first data models.

[0117] As an example, the various data model training algorithms described in this application embodiment are preset based on the number of attribute features contained in the multiple sample data stored in the sample database.

[0118] For example, the attribute features described in the embodiments of this application can be network traffic data information used by the user, such as including but not limited to peak uplink rate (peak_rx_rate), uplink traffic, cumulative uplink traffic, cumulative downlink traffic, total live streaming duration, total live streaming traffic, etc.

[0119] As an example, when displaying the multiple data models on the display interface, this application embodiment can also display the accuracy rate of obtaining the corresponding user list based on each data model, so that users can make model selection better, faster and more accurately based on the accuracy rate corresponding to each data model.

[0120] As an example, in the embodiment of this application, when displaying the multiple data models on the display interface, the multiple data models can be displayed sequentially in descending order of priority. It is understood that the data models displayed earlier have higher priority than the data models displayed later, so that users can select models more quickly and accurately based on the priority of each data model.

[0121] In this application, there are multiple ways to determine the priority of data models, and it is not limited to determining the priority based on the accuracy of user list acquisition for each data model.

[0122] S704: Determine the first data model selected sequentially from the plurality of first data models displayed, and verify whether the sequentially selected first data model is the first target data model.

[0123] For example, assuming the plurality of data models include data models 1 to 5, the operator selects data model 1 from data models 1 to 5 for the first time. The system verifies whether the data model 1 selected by the operator this time is the target data model, that is, whether it meets the operator's needs. For example, the verification method can be that the system generates a corresponding user list 1 based on the data model 1 and displays the user list 1 so that the operator can determine whether the data model 1 is the target data model by viewing the user list 1.

[0124] When the operator finds that the user list 1 obtained based on data model 1 does not meet the requirements, for example, if the system receives an instruction to confirm that the displayed user list 1 does not meet the user requirements, it can select a new data model from the remaining data models 2 to 5. For example, if the operator selects data model 2 the second time, the system generates a corresponding user list 2 based on data model 1 and displays the user list 2. The operator can then continue to check the user list 2 to determine whether data model 2 is the target data model. If data model 2 still does not meet the requirements, the process of selecting a model from the remaining data models and generating and confirming the user list can continue until a suitable data model that meets the requirements is found.

[0125] As an example, in practical applications, operators can select a data model from multiple data models displayed on the interface in the following way:

[0126] Selection Method 1: Operators select data models by clicking the data model icon displayed on the interface.

[0127] Selection Method 2: The operator gives a command to select a data model via voice input.

[0128] Specifically, the first target user list generated based on the first target data model satisfies user needs. This first target user list is obtained by inputting multiple data entries from a historical database into the first target data model, and then generating multiple output data entries based on the first target data model. The historical database stores the attribute characteristics of all users' network usage.

[0129] In this embodiment of the application, the first target user list can be determined to meet user needs in the following manner:

[0130] The system receives an instruction that the displayed list of the first target users and / or the displayed analysis information meet the user's requirements.

[0131] As an example, embodiments of this application may store the first target data model in a model database for subsequent applications based on the first target data model. The specific application methods are not limited to the following three:

[0132] Application Method 1: Based on the same target, the corresponding stored data model is called to extract the intent.

[0133] As an example, in this embodiment of the application, the first target data model and the first target mining intent can be stored in a model database. When the subsequent input target mining intent is the first target mining intent (for example, when the similarity between the subsequent input target mining intent and the first target mining intent is not less than a first threshold, the subsequent input target mining intent can be defaulted to the first target mining intent, wherein the first threshold can be set according to the actual application, such as the first threshold can be 90%), the first target data model is obtained from the model database based on the first target mining intent.

[0134] Then, a second target user list is generated based on the first target data model. This second target user list is obtained by inputting multiple data entries from the historical database into the first target data model and then generating multiple output data entries from the first target data model. Therefore, when the operator needs to search for previously used mining intentions to mine user information again, it can directly retrieve and apply the target data model corresponding to the previously used mining intentions from the model database, effectively saving system overhead and improving mining efficiency.

[0135] Application Method 2: Based on the user's invocation command, retrieve the data model indicated by the invocation command from the model database.

[0136] As an example, in this embodiment of the application, when receiving an instruction from an operator to generate a target user list using the first target data model, the first target data model can be retrieved from the model database. Then, a corresponding target user list is generated based on the first target data model. The target user list is obtained by inputting multiple data entries from the current historical database into the first target data model and then generating multiple output data based on the first target data model.

[0137] As an example, in the scenario of application mode 2, the target user list in this embodiment can also be obtained by inputting multiple data entries of a specified range (or specified interval, specified part) from the current historical database into the first target database model.

[0138] Application Method 3: Based on a fixed call cycle, retrieve the data model corresponding to the current cycle from the model database.

[0139] For example, the data model indicated for each call cycle is the first target data model.

[0140] As an example, in the scenario of application mode 3, this application embodiment can input multiple newly added data entries in the historical database within the corresponding period into the first target data model and obtain multiple output data. Based on the multiple output data, the corresponding target user list generated based on the first target data model in the current period can be obtained.

[0141] As an example, to better help operators determine whether the selected data model meets their actual needs, embodiments of this application can display the user list obtained based on the data model selected by the operator on the display interface, and / or can display the analysis information obtained by comparing and analyzing the user list obtained based on the data model selected by the operator with the original user list on the display interface. The original user list can be obtained by inputting multiple data from the historical database into the original data model and then outputting multiple data based on the original data model. The original data model is obtained from the data models already stored in the data model library based on the first target mining intent.

[0142] Based on this, operators can determine whether the selected data model meets the requirements by viewing and displaying the user list and / or the analysis information. If it does, the selected data model can be stored in the model database as the target model; otherwise, model optimization can be performed.

[0143] In this application embodiment, there are multiple ways to optimize the model when it is determined that the selected data model does not meet the operator's requirements, and these methods are not limited to the following two:

[0144] Optimization method 1: Select a new data model from the remaining data models among the multiple data models, and determine whether the newly selected data model is the first target data model.

[0145] Optimization Method 2: Receive the input of the second target mining intent, and perform the steps of data model acquisition and user list generation based on the second target mining intent.

[0146] The second target mining intention is obtained by modifying the first target mining intention.

[0147] As an example, when the first target data model is not present among the multiple data models, a second target mining intention is received. Based on the second target mining intention, second training data is obtained from the sample database. The second target mining intention is used to mine target users. Based on multiple preset data model training algorithms, the second training data is used to train corresponding models to obtain multiple data models, which are then displayed. Then, a second target data model is determined from the multiple data models. A third target user list generated based on the second target data model satisfies user needs. The third target user list is obtained by inputting multiple data entries from the historical database into the second target data model and then outputting multiple data entries from the second target data model. The second target data model is the data model selected from the displayed multiple data models. The second target mining intention is obtained by modifying the first target mining intention.

[0148] Furthermore, to better understand the embodiments of this application, the different stages of the application based on the data model acquisition system are described in detail below:

[0149] Phase 1: Pre-setting Phase

[0150] In this embodiment of the application, before obtaining a product or system based on the data model acquisition method described above, the two engines (intent perception and translation engine and data model management and generation engine) introduced in Figures 2 to 5 above are trained, thereby enabling the two engines to possess the corresponding capabilities described above. The training process, overall interaction, and inference process of these two engines are briefly described below through specific embodiments.

[0151] Engine Training 1. Train the intent perception and translation engine's attribute feature recognition and translation capabilities.

[0152] Referring to Figure 8, this is a schematic diagram of a preparation phase intention perception and translation engine training and construction process provided in an embodiment of this application.

[0153] S801: The trainer sends the training dataset to the intent-aware and translation engine.

[0154] As an example, the training dataset described in this application embodiment is obtained by using a large amount of general corpus data of words such as "potential customer", "potential customer type", "attribute / field", and "influencing factors" as the basis for constructing a basic large language model (LLM).

[0155] S802: Train the training dataset based on AI algorithms to learn a general natural language grammatical probability distribution model.

[0156] S803: Generates a basic data model with natural language understanding capabilities.

[0157] As an example, the natural language understanding capabilities obtained by the basic data model in this application embodiment include the ability to identify keywords such as "potential customer", "potential customer type", "attribute / field", and "influencing factors" as well as all synonyms.

[0158] S804: The trainer inputs the attribute fields of the data and the corresponding knowledge into the basic data model.

[0159] For example, the attribute fields and corresponding knowledge of the input data in the embodiments of this application are typically carried in the form of YAML files or XSLX tables. Each attribute field knowledge includes the field type, field value description, field meaning, and data example.

[0160] For example, the data attribute fields and corresponding knowledge format examples provided in the embodiments of this application are shown in Table 1 below:

[0161] Table 1 shows the format of attribute fields and their corresponding knowledge.

[0162] S805: Vectorize the input attribute fields and corresponding knowledge.

[0163] S806: Train the basic data model based on the vectorized attribute fields and knowledge.

[0164] Specifically, the process of training the basic data model based on the attribute fields and knowledge in S806 is mainly aimed at training the basic data model to learn the probability distribution model between each Chinese character in the data attribute fields, as well as the correspondence between the data attribute fields and the knowledge content, so as to obtain the intent perception and translation engine based on the basic data model.

[0165] S807: After the basic data model completes training based on the attribute fields and corresponding knowledge, it generates an intent-aware and translation engine with attribute field recognition and translation capabilities.

[0166] S808: Send the intent-aware and translation engine to the model trainer.

[0167] The capabilities include model training, training data model generation, and management engine model generation and construction.

[0168] Referring to Figure 9, this is a schematic diagram of a data model generation and management engine training and construction process provided in an embodiment of this application.

[0169] S901: Model trainers obtain N attribute features of user network usage based on historical databases.

[0170] The historical database mentioned in S901 can be an existing conventional database, that is, a database frequently used in the current operating scenario.

[0171] S902: The model trainer pre-sets multiple data model AI training algorithms in the data model generation and management engine based on the N attribute features.

[0172] For example, in this application embodiment, based on the N attribute features, N*N data model AI training algorithms are preset. The N*N data model AI training algorithms can be i-order j-ary AI classification training algorithms, where i is 1, ..., N and j is 1, ..., N.

[0173] S903: Based on the known M types of potential customers, the data model generation and management engine queries the historical database to obtain the data of each attribute field corresponding to each type of potential customer as training data.

[0174] S904: Based on the existing AI training algorithm, train the data model generation and management engine using the training data.

[0175] S905: Generate the optimal potential customer data model corresponding to each of the M potential customer types.

[0176] For example, there are M types of potential customers in this application embodiment. Therefore, based on the M types of potential customers, this application embodiment can obtain an optimal potential customer identification model for each type, that is, a total of M potential customer identification models are obtained.

[0177] S906: After training is completed, it generates a data model generation and management engine with M known potential customer data models and multiple data model AI training algorithms.

[0178] S907: Send the data model generation and management engine to the model trainer.

[0179] As an example, when operators apply the data model acquisition system provided in the embodiments of this application, they can directly select the corresponding data model from the M known potential customer data models that have been pre-stored based on their own potential customer type requirements.

[0180] Phase Two: Operator Preparation Phase

[0181] Referring to Figure 10, this is a flowchart illustrating an operator preparation phase provided in an embodiment of this application.

[0182] S1001: Operators obtain corresponding sample data and historical data according to business needs.

[0183] S1002: The operator inputs the sample data and historical data into the data model acquisition system.

[0184] S1003: The data model acquisition system stores the sample data and the historical data.

[0185] Phase Three: Application and Operation Phase

[0186] Referring to Figure 11, this is a flowchart illustrating an application operation stage provided in an embodiment of this application.

[0187] S1101: Receives the first target mining intent from the input and sends it to the data analysis layer.

[0188] For example, the target mining intent can be understood as the target user to be mined. For instance, the first target mining intent could be "to mine the characteristics of heavy live streaming users and identify the corresponding potential users".

[0189] S1102: The data analysis layer identifies the mining intent of the first target.

[0190] For example, in this application embodiment, the intent perception and translation engine in the above system can identify the potential customer name and attribute fields corresponding to the first target mining intent from the intent. For example, the obtained potential customer name is "heavy live streaming user", and the obtained attribute fields are: unspecified.

[0191] S1103: The data analysis layer obtains the first training data from the sample database based on the recognition results.

[0192] For example, the sample database stores labeled data with the target user for model training.

[0193] Specifically, the data model generation and management engine in the data analysis layer acquires the first training data.

[0194] S1104: In the data analysis layer, the first training data is trained using various data model training algorithms to obtain multiple data models.

[0195] S1105: The UI operation layer displays the multiple data models.

[0196] S1106: The operator selects one data model from the plurality of data models.

[0197] S1107: The data analysis layer obtains the selected data model.

[0198] S1108: The data analysis layer generates a corresponding user list based on the selected data model.

[0199] For example, the system described in this application embodiment can execute the script of the data model selected by the operator (such as a Python script + SQL query) to obtain the corresponding user list from the historical database.

[0200] For example, in the embodiments of this application, the user list generated in S1108 above can be displayed in a data format, or in the form of a line chart, pie chart, etc.

[0201] For example, taking the target user list that needs to be obtained as a list of potential customers who need 10 Gigabit network speed as an example, see Figure 12, which is a schematic diagram of the display interface of the potential customer list provided by an embodiment of this application. In this embodiment of the application, the display interface can display a list of the potential customer list, the percentage of potential customers obtained after analysis based on the potential customer list, the acquisition of multiple related attribute features corresponding to the potential customer list, and the display of the most important attribute feature among the multiple related attribute features, etc.

[0202] For example, taking the target user list to be obtained as a list of potential customers who need Fiber to the Room (FTTR) technology as an example, see Figure 13, which is a schematic diagram of the display interface of the potential customer list provided by an embodiment of this application. In this embodiment of the application, the display interface can display the numerical values ​​of the attribute characteristics corresponding to each user in the potential customer list, the changing trend of the potential customer list over time, and the proportion of potential customers obtained after analysis based on the potential customer list, etc.

[0203] As an example, this application embodiment can also compare and analyze the old user list with the user list generated in S1108 to obtain analysis information, and then display the analysis information to the operator.

[0204] S1109: Display the user list on the display interface.

[0205] S1110: The operator determines whether the selected data model meets the requirements based on the user list. If yes, execute S1111; otherwise, execute S1112.

[0206] S1111: The selected data model is determined as the target data model and stored in the model database.

[0207] S1112: Perform data model optimization.

[0208] For example, the embodiments of this application can perform model optimization in various ways, and are not limited to the following two:

[0209] Optimization Method 1: The operator selects a new data model from the remaining data models among the multiple data models according to the requirements, and generates the corresponding user list based on the newly selected data model.

[0210] It is understood that the optimization method 1 is to iteratively execute the steps described in S1107 to S1111.

[0211] Optimization Method 2: Re-enter the target to mine intent.

[0212] It is understandable that the optimization method 2 involves iteratively executing the steps described in S1101 to S1111.

[0213] When operators re-describe the target mining intent, they can add corresponding attribute field restrictions to better obtain a suitable target data model.

[0214] To better understand the solutions described in the embodiments of this application, this application selects a target mining intent scenario for introduction. It should be noted that the following examples are for illustrative purposes only and do not constitute a limitation on the embodiments of this application.

[0215] Assuming that when the operator uses the system provided in this application to acquire data models, the target mining intent input through the UI display interface of the system is "Please mine the characteristics of heavy live streaming users and identify the corresponding target users".

[0216] After the system obtains the target mining intent, it extracts keywords from the statement "Please mine the characteristics of heavy live streaming users and identify the corresponding potential users". For example, if the keyword "heavy live streaming users" is obtained, the name of the user list to be generated can be determined as the live streaming potential user list. In this case, no attribute fields are specified in the target mining intent.

[0217] Then, based on the recognition result of the target mining intent, the corresponding training data is obtained from the sample database, and the training data is trained according to the corresponding model based on the preset multiple data model training algorithms to obtain multiple data models, and the multiple data models are displayed.

[0218] Understandably, if the target mining intent indicates attribute fields, for example, if the attribute fields corresponding to the live streaming type include peak uplink rate (peak_rx_rate), uplink traffic, cumulative uplink traffic, cumulative downlink traffic, total live streaming duration, and total live streaming traffic, then the embodiments of this application can train the database based on various AI training algorithms corresponding to these attributes to obtain various models.

[0219] For example, in embodiments of this application, the generated data models 1 to 3 can be displayed on the UI of the system.

[0220] Assume the function expressions for each of the data models 1 to 3 are as follows:

[0221] Data Model 1: a bandwith-type + b up_peak_Tx_rate - c Data Model 2: a1 bandwith-type + b1 up_peak_Tx_rate > c1, ont_count > x1 Data Model 3: a2 bandwith-type + b2 up_peak_Tx_rate > c2, x1 > ont_count > x2

[0222] Where a, b, c, a1, b1, c1, x1, ... are the data model coefficients in the training results of the aforementioned stage.

[0223] Then, the user can select one of the data models 1 to 3 displayed in the system's UI as the pending data model, and retrieve the corresponding user list from the historical database based on the selected data model.

[0224] In this embodiment of the application, the corresponding model accuracy rate can be displayed after each of the data models 1 to 3, thereby enabling operators to better select data models based on the accuracy rate.

[0225] At this point, the user can determine whether the pending data model meets their needs based on the user list. If it does, the pending data model is designated as the target data model and stored. Subsequently, potential customer mining is performed on the acquired user data based on the target data model. If it does not meet the needs, another data model can be selected from the remaining multiple data models displayed in the UI as the pending model, and the subsequent process can be re-executed until the target data model is obtained. Alternatively, a more detailed target mining intent can be re-entered from the UI, and the subsequent data model selection and target data model determination processes can be re-executed based on the newly entered target mining intent until the target data model is obtained.

[0226] The following examples illustrate the solutions provided in this application's embodiments, taking into account the inter-layer interaction processes within the system under different operational scenarios.

[0227] Scenario 1: Based on the data import operation of the operator, store the imported data.

[0228] Referring to Figure 14, this is a schematic diagram of a data import and storage process provided in an embodiment of this application.

[0229] S1401: Operators import service-related data through the intent setting module in the UI operation layer.

[0230] The business-related data may include, but is not limited to, sample data and historical data.

[0231] As an example, the sample data may include potential customer / poor quality customer sample data.

[0232] S1402: The intent setting module transmits the acquired business-related data to the data storage layer.

[0233] S1403: After the data storage layer obtains the business-related data, it stores the business-related data in the corresponding database.

[0234] Scenario 2: Obtain training data by mining the intent based on the target input from the operator.

[0235] Referring to Figure 13, this is a schematic diagram of a process for obtaining training data based on target mining intent provided in an embodiment of this application.

[0236] S1501: Operators input target intents through the intent setting module in the UI operation layer.

[0237] As an example, in this embodiment of the application, the operator can input a potential customer / poor quality type mining intent based on natural language description in the intent setting module.

[0238] The input method can be text input or voice input. For example, users can type using a keyboard to describe their intent to find potential customers using the text or symbols they input. Alternatively, users can use a microphone or other voice device to input voice information describing their intent to find potential customers. There are no restrictions on this method.

[0239] S1502: The intent setting module will obtain the intent perception and translation engine in the target mining intent input data analysis layer.

[0240] S1503: The intent perception and translation engine perceives and translates the target mining intent to obtain the recognition result.

[0241] S1504: Send the recognition results to the data model management and generation engine.

[0242] S1505: The data model management and generation engine obtains the corresponding training data from the sample database in the data storage layer based on the recognition results.

[0243] As an example, the training data is obtained from a sample database in the data storage layer.

[0244] Scenario 3: Generate multiple data models based on the target intent mined by the operator.

[0245] Referring to Figure 16, this is a schematic diagram of a process for generating multiple data models provided in an embodiment of this application.

[0246] S1601: The data model management and generation engine in the data analysis layer obtains training data corresponding to the target mining intent from the data storage layer.

[0247] S1602: The data model management and generation engine performs corresponding model training on the first training data based on preset multiple data model training algorithms to obtain multiple data models.

[0248] S1603: The data model management and generation engine returns the generated multiple data models to the data model setting module in the UI operation layer for display.

[0249] S1604: The operator selects a data model from the multiple data models displayed by the data model setting module in the corresponding display area of ​​the system display interface.

[0250] Scenario 4: Generate the corresponding user list based on the data model selected by the operator.

[0251] Referring to Figure 17, this is a schematic diagram of a user list generation process provided in an embodiment of this application.

[0252] S1701: Data model management and generation engine operation in the system data analysis layer, which selects the data model chosen by the operator.

[0253] As an example, the data model management and generation engine described in this application embodiment can run the script of the data model selected by the operator, query the historical database in the data storage layer, and thereby obtain the corresponding potential customer / poor quality data from the historical database.

[0254] S1702: The data storage layer responds to the operation of the selected data model and generates the corresponding user list.

[0255] S1703: The data storage layer returns the generated user list to the data information generation module in the data analysis layer.

[0256] S1704: The data information generation module displays the user list at the display position corresponding to the UI operation layer review module.

[0257] As an example, in step S1704, the data information generation module can also perform analysis and processing based on the user list to obtain and display the corresponding analysis information.

[0258] As an example, embodiments of this application can determine the content that the review module wants to display based on the operator's needs. For example, it can only display the user list obtained by the operator based on the selected data model, based on the operator's needs. Or, it can display the user lists obtained by multiple data models respectively, so as to better determine the data model suitable for actual application based on the content of the user list corresponding to each data model. Or, it can display the comparative analysis results of the user lists between the newly selected data model and the previous data model, so that the operator can better see the differences between the newly selected data model and the original data model.

[0259] The analytical information may include data display dimensions, strengths and weaknesses, etc., for each user list.

[0260] The analysis information in this application embodiment can be displayed in the form of a data chart or a visual data table, and there is no limitation on this.

[0261] Scenario 5: Store the target data model determined by the operator.

[0262] The target data model is used to indicate that the first target user list generated based on the target data model meets the operator's requirements. The first target user list is obtained by inputting multiple data from the historical database into the first target data model and then outputting multiple data from the first target data model.

[0263] Referring to Figure 18, it is a schematic diagram of a target data model storage process provided in an embodiment of this application.

[0264] S1801: The operator inputs confirmation of the data entry operation for the selected data model.

[0265] S1802: After receiving the confirmation of data entry, store the current data model into the existing model database in the data storage layer.

[0266] Furthermore, when operators want to adjust or change the data model, they can re-enter the target mining intent (e.g., enter a second target mining intent) based on the content of Scenario 2 above, obtain the corresponding training data based on the second target mining intent, and then re-obtain multiple data models based on the training data, etc.; or they can re-select a data model from the remaining data models among the multiple data models displayed in the display area corresponding to the system display interface by the data model setting module based on the content of Scenario 3 above, and then proceed with the subsequent steps.

[0267] This application embodiment can automatically generate multiple possible data models with corresponding mining intentions based on the target mining intent input by the operator user and by fitting mining patterns to historical data. This enables efficient mining and expansion based on mining intent. Operators can select a model from the generated multiple data models according to their actual needs. Furthermore, by adopting an automated modeling method, no manual intervention is required, which can effectively reduce modeling time and costs. The model can also be optimized in a timely manner according to the operator's needs, making it more flexible, convenient, and practical. In addition, this application embodiment can also generate a corresponding user list in real time based on the selected data model after the operator selects a data model, allowing the operator to determine whether model optimization is needed based on the displayed user list, thus enhancing flexibility.

[0268] Furthermore, in this embodiment, the operator can also use the fault location and repair functions provided by the data mining component to perform log analysis after a fault occurs. The operation and maintenance component can also perform anomaly analysis and location on the logs, generating fault classification and repair guidance based on abnormal log fragments. This allows the operator to operate according to the repair guidance or send back logs, completing a self-closing loop for device repair on the operator's end or backend repair.

[0269] Figure 19 is a structural diagram of the device for data model acquisition provided in an embodiment of this application. The device can be implemented by software and may be part or all of the device itself. The device provided in this embodiment can implement the process described in any one of Figures 7 to 18 of this embodiment. The device includes: an input / output module 1910 and a processing module 1920, wherein:

[0270] Input / output module 1910 is used to receive a first target mining intent, which is used to mine target users;

[0271] The processing module 1920 is used to obtain first training data from a sample database based on the first target mining intent; the sample database stores data labeled with the target user for model training; and performs corresponding model training on the first training data based on a variety of preset data model training algorithms to obtain multiple first data models.

[0272] The input / output module 1910 is also used to display the plurality of first data models;

[0273] The processing module 1920 is further configured to determine the first data model selected sequentially from the plurality of first data models displayed, and verify whether the sequentially selected first data model is the first target data model; wherein, the first target user list generated based on the first target data model meets the user requirements, and the first target user list is obtained by inputting multiple data from the historical database into the first target data model and then outputting multiple data from the first target data model; the historical database stores the attribute characteristics of all users using the network.

[0274] In one possible implementation, the processing module 1920 is further configured to:

[0275] The first target data model and the first target mining intent are stored in the model database. When the target mining intent received by the input / output module is the first target mining intent, the first target data model is obtained from the model database based on the first target mining intent. A second target user list is generated based on the first target data model. The second target user list is obtained by inputting multiple data from the historical database into the first target data model and then outputting multiple output data based on the first target data model.

[0276] In one possible implementation, the processing module 1920 is further configured to:

[0277] When the first target data model is not present among the multiple data models, the second target mining intention is received through the input / output module; the second target mining intention is obtained by modifying the first target mining intention; second training data is obtained from the sample database based on the second target mining intention, and the second target mining intention is used to mine the target user; the second training data is trained according to the corresponding model based on the preset multiple data model training algorithms to obtain multiple second data models, and the multiple second data models are displayed through the input / output module; the second data model selected in sequence from the displayed multiple second data models is determined, and it is verified whether the selected second data model is the second target data model; wherein, the third target user list generated based on the second target data model satisfies the user requirements, and the third target user list is obtained by inputting multiple data from the historical database into the second target data model and then outputting multiple output data based on the second target data model.

[0278] In one possible implementation, when the processing module 1920 obtains the first training data from the sample database based on the first target mining intent, it is specifically used for:

[0279] Determine the tag of the target user corresponding to the first target mining intent, and / or, the target user uses the attribute features of the network;

[0280] The first training data is obtained by retrieving data from the sample database containing the tags of the identified target users and / or the attribute features.

[0281] In one possible implementation, the number of data model training algorithms is related to the number of attribute features contained in the multiple sample data stored in the sample database.

[0282] In one possible implementation, the input / output module 1910 is further configured to:

[0283] Display the first target user list generated based on the first target data model; and / or display the analysis information obtained by comparing the first target user list with the original user list; the original user list is obtained by inputting multiple data from the historical database into the original data model and then outputting multiple data based on the original data model; the original data model is obtained from the data models already stored in the data model library based on the first target mining intention.

[0284] In one possible implementation, the processing module determines that the first target user list meets the user requirements by means of the following method:

[0285] The input / output module receives instructions to confirm that the displayed first target user list and / or the displayed analysis information meet the user's needs.

[0286] In one possible implementation, the target mining intent is described based on natural language.

[0287] The module division in this embodiment is illustrative and represents only one logical functional division. In actual implementation, other division methods are possible. Furthermore, the functional modules in each embodiment of this application can be integrated into a single processor, exist as separate physical entities, or be integrated into a single module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0288] If the integrated module is implemented as a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a terminal device (which may be a personal computer, network device, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0289] This application also provides another data model acquisition device in its embodiments. Figure 20 provides an exemplary possible architecture diagram of the data model acquisition device.

[0290] The data model acquisition device includes a memory 2001, a processor 2002, a communication interface 2003, and a bus 2004. The memory 2001, processor 2002, and communication interface 2003 are interconnected via the bus 2004.

[0291] The memory 2001 can be a ROM, static storage device, dynamic storage device, or RAM. The memory 2001 can store programs. When the program stored in the memory 2001 is executed by the processor 2002, the processor 2002 and the communication interface 2003 are used to execute the data model acquisition method shown in any of Figures 7 to 16. The memory 2001 can also store information such as multiple databases.

[0292] The processor 2002 can be a general-purpose central processing unit (CPU), a microprocessor, an application-specific integrated circuit (ASIC), a graphics processing unit (GPU), or one or more integrated circuits.

[0293] The processor 2002 can also be an integrated circuit chip with signal processing capabilities. In implementation, some or all of the functions of the modeling device of this application can be completed by the integrated logic circuits in the hardware of the processor 2002 or by instructions in software form. The aforementioned processor 2002 can also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the above embodiments of this application. The general-purpose processor can be a microprocessor, or the processor can be any conventional processor, etc. The steps of the methods disclosed in the embodiments of this application can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules can be located in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art.

[0294] Communication interface 2003 uses transceiver modules, such as, but not limited to, transceivers, to enable communication between the modeling device and other devices or communication networks. For example, point cloud data can be acquired through communication interface 2003.

[0295] Bus 2004 may include a pathway for transmitting information between various components of the modeling apparatus (e.g., memory 2001, processor 2002, communication interface 2003).

[0296] The descriptions of the processes corresponding to the above-mentioned figures each have their own emphasis. For parts of a process that are not described in detail, please refer to the relevant descriptions of other processes.

[0297] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions. When these computer program instructions are loaded and executed on a server or terminal, they generate all or part of the processes or functions described in the embodiments of this application. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic cable, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to the server or terminal, or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, and magnetic tape), an optical medium (e.g., digital video disk (DVD), etc.), or a semiconductor medium (e.g., solid-state drive).

[0298] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in one or more flowchart illustrations and / or one or more block diagrams.

[0299] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for obtaining a data model, characterized in that, include: The system receives a first target mining intent as input, and obtains first training data from a sample database based on the first target mining intent. The first target mining intent is used to mine target users. The sample database stores labeled data of the target user for model training; Based on multiple preset data model training algorithms, the first training data is trained with corresponding models to obtain multiple first data models, and the multiple first data models are displayed. The first data model selected sequentially from the displayed plurality of first data models is determined, and it is verified whether the sequentially selected first data model is the first target data model; wherein, the first target user list generated based on the first target data model meets the user requirements, and the first target user list is obtained by inputting multiple data from the historical database into the first target data model and then outputting multiple data from the first target data model; the historical database stores the attribute characteristics of all users using the network.

2. The method as described in claim 1, characterized in that, The method further includes: The first target data model and the first target mining intent are stored in the model database. When the subsequent input target mining intent is the first target mining intent, the first target data model is obtained from the model database based on the first target mining intent; A second target user list is generated based on the first target data model. The second target user list is obtained by inputting multiple data entries from the historical database into the first target data model and then outputting multiple data entries from the first target data model.

3. The method as described in claim 1 or 2, characterized in that, The method further includes: When the first target data model is not present among the multiple data models, a second target mining intention is received; the second target mining intention is obtained by modifying the first target mining intention. Based on the second target mining intent, second training data is obtained from the sample database, and the second target mining intent is used to mine the target user; Based on a variety of preset data model training algorithms, the second training data is trained with corresponding models to obtain multiple second data models, and the multiple second data models are displayed. The second data model selected sequentially from the displayed plurality of second data models is determined, and it is verified whether the sequentially selected second data model is the second target data model; wherein, the third target user list generated based on the second target data model satisfies the user requirements, and the third target user list is obtained by inputting multiple data from the historical database into the second target data model and then outputting multiple data from the second target data model.

4. The method according to any one of claims 1 to 3, characterized in that, The step of obtaining the first training data from the sample database based on the first target mining intent includes: Determine the tag of the target user corresponding to the first target mining intent, and / or, the target user uses the attribute features of the network; The first training data is obtained by retrieving data from the sample database containing the tags of the identified target users and / or the attribute features.

5. The method according to any one of claims 1 to 4, characterized in that, The number of data model training algorithms is related to the number of attribute features contained in the multiple sample data stored in the sample database.

6. The method according to any one of claims 1 to 5, characterized in that, The method further includes: Display the list of the first target users generated based on the first target data model; and / or, The display shows the analysis information obtained by comparing the first target user list with the original user list; the original user list is obtained by inputting multiple data from the historical database into the original data model and then outputting multiple data based on the original data model; the original data model is obtained from the data model stored in the data model library based on the first target mining intention.

7. The method as described in claim 6, characterized in that, The first list of target users meets the user requirements and is determined in the following ways: Receive an instruction to confirm that the displayed first target user list and / or the displayed analysis information meet the user's needs.

8. The method according to any one of claims 1 to 7, characterized in that, The target mining intent is described using natural language.

9. A data model acquisition device, characterized in that, include: The input / output module is used to receive a first target mining intent, which is used to mine target users; The processing module is used to obtain first training data from the sample database based on the first target mining intent; The sample database stores labeled data of the target user for model training; Based on multiple preset data model training algorithms, the first training data is used to train corresponding models to obtain multiple first data models. The input / output module is also used to display the plurality of first data models; The processing module is further configured to determine the first data model selected sequentially from the plurality of first data models displayed, and verify whether the sequentially selected first data model is the first target data model; wherein, the first target user list generated based on the first target data model meets the user requirements, and the first target user list is obtained by inputting multiple data from the historical database into the first target data model and then outputting multiple data from the first target data model; the historical database stores the attribute characteristics of all users using the network.

10. The apparatus as claimed in claim 9, characterized in that, The processing module is also used for: The first target data model and the first target mining intent are stored in the model database. When the target mining intent received by the input / output module is the first target mining intent, the first target data model is obtained from the model database based on the first target mining intent; A second target user list is generated based on the first target data model. The second target user list is obtained by inputting multiple data entries from the historical database into the first target data model and then outputting multiple data entries from the first target data model.

11. The apparatus as claimed in claim 9 or 10, characterized in that, The processing module is also used for: When the first target data model is not present among the multiple data models, the second target mining intention is received through the input / output module. The second target mining intent is obtained by modifying the first target mining intent; Based on the second target mining intent, second training data is obtained from the sample database, and the second target mining intent is used to mine the target user; Based on multiple preset data model training algorithms, the second training data is trained with corresponding models to obtain multiple second data models, and the multiple second data models are displayed through the input / output module. The second data model selected sequentially from the displayed plurality of second data models is determined, and it is verified whether the sequentially selected second data model is the second target data model; wherein, the third target user list generated based on the second target data model satisfies the user requirements, and the third target user list is obtained by inputting multiple data from the historical database into the second target data model and then outputting multiple data from the second target data model.

12. The apparatus according to any one of claims 9 to 11, characterized in that, When the processing module obtains the first training data from the sample database based on the first target mining intent, it is specifically used for: Determine the tag of the target user corresponding to the first target mining intent, and / or, the target user uses the attribute features of the network; The first training data is obtained by retrieving data from the sample database containing the tags of the identified target users and / or the attribute features.

13. The apparatus according to any one of claims 9 to 12, characterized in that, The number of data model training algorithms is related to the number of attribute features contained in the multiple sample data stored in the sample database.

14. The apparatus according to any one of claims 9 to 13, characterized in that, The input / output module is also used for: Display the list of the first target users generated based on the first target data model; and / or, The display shows the analysis information obtained by comparing the first target user list with the original user list; the original user list is obtained by inputting multiple data from the historical database into the original data model and then outputting multiple data based on the original data model; the original data model is obtained from the data model stored in the data model library based on the first target mining intention.

15. The apparatus as claimed in claim 14, characterized in that, The processing module determines whether the first target user list meets the user requirements through the following methods: The input / output module receives instructions to confirm that the displayed first target user list and / or the displayed analysis information meet the user's needs.

16. The apparatus according to any one of claims 9 to 15, characterized in that, The target mining intent is described using natural language.

17. A data model acquisition device, characterized in that, include: Memory, used to store program instructions; A processor, configured to be coupled to the memory, to invoke program instructions in the memory, and to execute the method as described in any one of claims 1-8.

18. A chip, characterized in that, The chip is connected to a memory and is used to read and execute program code stored in the memory to implement the method as described in any one of claims 1-8.

19. A computer storage medium, characterized in that, The computer storage medium stores a computer program, which, when executed by a computer, causes the computer to perform the method as described in any one of claims 1-8.

20. A computer program product containing instructions, characterized in that, When the computer program product is run on a computer, it causes the computer to perform the method as described in any one of claims 1-8.

Images