Cloud card distribution method and device, equipment and medium

Abstract

The invention is suitable for the technical field of communication, and particularly relates to a cloud card distribution method, device, equipment and medium, the method comprises the steps that a card distribution request is received, the card distribution request is used for requesting to distribute a cloud card required for accessing an operator network for a target terminal, and the card distribution request comprises position information when the target terminal initiates the card distribution request; according to network quality information of each operator in an operator list corresponding to the position information, a target cloud card is determined from a cloud card set corresponding to the operator list, and the network quality information comprises at least one of a network type and a network quality level; and sending the target cloud card to the target terminal. The network quality information of the operator at the position can be determined in real time according to the geographic position when the terminal initiates the card distribution request, and the optimal cloud card is dynamically allocated to the target terminal according to the network quality information, so that the target terminal surfs the Internet through the cloud card.

Description

Technical Field

[0001] This application belongs to the field of communication technology, and in particular relates to a method, apparatus, device, and medium for allocating cloud cards. Background Technology

[0002] In mobile communications, the cloud SIM card serves as a user's digital identity credential, enabling the terminal to access the operator's network identified by the PLMN (Public Land Mobile Network). Its core objective is to allocate cloud SIM cards to user terminals, allowing them to register and access the corresponding operator's network, ensuring users receive the best possible network quality experience.

[0003] However, in practical applications, common SIM card allocation methods rely on limited information, often leading to inaccurate allocation results. This can prevent terminals from accessing the optimal network quality for their current geographical location, impacting user experience and connection reliability. Summary of the Invention

[0004] This application provides a method, apparatus, device, and medium for allocating cloud cards, which can determine the network quality information of operators at different locations in real time. The network quality information can represent the quality of the network services provided by the operator, and thus the optimal cloud card can be dynamically selected and allocated to the terminal for Internet access based on the operator's network quality information.

[0005] In a first aspect, embodiments of this application provide a method for allocating cloud cards, the method comprising: Receive a card allocation request. The card allocation request is used to request the allocation of a cloud card required for the target terminal to access the operator's network. The card allocation request includes the location information of the target terminal when it initiates the card allocation request. Based on the network quality information of each operator in the operator list corresponding to the location information, the target cloud card is determined from the cloud card set corresponding to the operator list. The network quality information includes at least one of network standard type and network quality level. Send the target cloud card to the target terminal.

[0006] This technical solution receives a SIM card allocation request containing terminal location information and selects a target cloud SIM card based on the network quality information (network type, quality level) of each operator corresponding to that location. At the time of SIM card allocation decision-making, the overall service level of each available network at the current location is known in advance, thus achieving optimal cloud SIM card allocation. This not only avoids increased power consumption and prolonged service unavailability caused by prolonged or frequent network searches by the terminal, but also proactively matches the terminal with a cloud SIM card from an operator with better network quality at the current geographical location, significantly improving the user's network service experience and connection reliability, and realizing intelligent and precise scheduling of cloud SIM card resources.

[0007] In conjunction with the first aspect, in one possible implementation of the first aspect, the card splitting request also includes time information and device information of the target terminal, and the method further includes: The trained network quality prediction model is used to process the first data sequence corresponding to the SIM card request to obtain the network quality information of each operator in the operator list corresponding to the location information. The first data sequence includes time information, the regional grid information corresponding to the location information, and the device information of the target terminal. The regional grid information corresponding to the location information is obtained by dividing the location information into regions.

[0008] In this implementation, a pre-trained network quality prediction model is introduced, and predictions are made using a first data sequence containing time, regional grid, and device information. This upgrades the acquisition of network quality information from static statistics to dynamic intelligent reasoning. The model can learn and integrate the complex nonlinear relationships between multi-dimensional spatiotemporal and device characteristics and network quality, thereby achieving accurate and real-time predictions of the network quality (standard and level) of various operators at any location and time point. This provides an intelligent and high-precision decision-making basis for subsequent optimal cloud card selection.

[0009] In conjunction with the first aspect, in one possible implementation of the first aspect, when the network quality information includes network standard type and network quality level, the target cloud card is determined from the set of cloud cards corresponding to the operator list based on the network quality information of each operator in the operator list corresponding to the location information, including: Based on the network standard type of each operator in the operator list corresponding to the location information, a first cloud card set is determined from the cloud card set corresponding to the operator list. The first cloud card set includes at least one cloud card, and the network standard type of each cloud card in the first cloud card set is the network standard type supported by the target terminal. The target cloud card is determined from the first set of cloud cards based on the network quality level of each operator in the operator list corresponding to the location information.

[0010] In this implementation, a step-by-step selection mechanism is used to determine the target cloud SIM card. First, based on the predicted network standard type, a first set of cloud SIM cards matching the terminal capabilities is selected, ensuring that the network standard of the allocated cloud SIM cards meets the terminal capability requirements. Then, further optimization is performed based on network quality level, ensuring that the final allocation result simultaneously meets the requirements of usability and high quality, making the allocation logic more rigorous and efficient.

[0011] In conjunction with the first aspect, in one possible implementation of the first aspect, the target cloud card is determined from the first set of cloud cards based on the network quality level of each operator in the operator list corresponding to the location information, including: Based on the network quality level of each operator in the operator list corresponding to the location information, the cloud card with the highest network quality level of the corresponding operator in the first cloud card set is determined as the target cloud card; Alternatively, the target cloud card can be determined from the first set of cloud cards based on the network quality level and load of each operator in the operator list corresponding to the location information.

[0012] This implementation offers two more specific optimization strategies. The first is to directly select the cloud SIM card with the highest predicted quality level, aiming for the optimal network experience for a single connection. The second is to comprehensively consider both network quality level and cloud SIM card resource load, ensuring high quality while also balancing the overall load of the cloud SIM card resource pool. This helps avoid the concentrated consumption of high-quality SIM card resources from a single operator, preventing other location devices from needing these resources from being unable to allocate them, thus achieving a balance between user experience and resource supply.

[0013] In conjunction with the first aspect, in one possible implementation of the first aspect, the target cloud card is determined from the first set of cloud cards based on the network quality level and load of each operator in the operator list corresponding to the location information, including: Based on the network quality level of each operator in the operator list corresponding to the location information, a second cloud card set is determined from the first cloud card set, and the second cloud card set includes at least one cloud card; Based on the load of each operator in the operator list corresponding to the location information, the target cloud card is determined from the second cloud card set, and the operator corresponding to the target cloud card has the minimum load.

[0014] In this implementation, the selection process is further refined by first defining a high-quality candidate set (the second cloud SIM card set) based on quality levels, and then selecting the operator's cloud SIM card with the lightest load from this set. This strategy ensures that all candidate cloud SIM cards are at a high quality level, and then achieves a reasonable distribution of traffic among high-quality resources through load balancing. This maintains high user connection quality while preventing localized overload and optimizing resource utilization efficiency.

[0015] In conjunction with the first aspect, in one possible implementation of the first aspect, a second cloud SIM card set is determined from a first cloud SIM card set based on the network quality level of each operator in the operator list corresponding to the location information. The second cloud SIM card set includes at least one cloud SIM card, including: The cloud cards in the first cloud card set whose network quality level of the corresponding operator is higher than the first level threshold are merged into the second cloud card set. The first level threshold is a preset fixed value. Alternatively, the cloud cards of the n operators with the highest network quality rankings in the first cloud card set can be merged into a second cloud card set, where n is a positive integer and is less than or equal to the total number of operators corresponding to the cloud cards in the first cloud card set.

[0016] This implementation method clarifies two specific quantification methods for constructing the second cloud SIM card set (high-quality candidate set): a fixed threshold method (quality level above a fixed value) and a dynamic ranking method (selecting the top n operators). The fixed threshold method has a unified standard and is easy to implement and manage; the dynamic ranking method can adapt to the overall distribution of network quality in different regions, ensuring that several relatively optimal choices are always selected. Both methods make the selection process for the high-quality candidate set clear and the standards explicit, enhancing the operability and adaptability of the solution.

[0017] In conjunction with the first aspect, in one possible implementation of the first aspect, when the network quality information includes network standard type and network quality level, a trained network quality prediction model is used to process the first data sequence corresponding to the SIM card request to obtain the network quality information of each operator in the operator list corresponding to the location information, including: The trained network quality prediction model is used to process the regional grid information corresponding to the location information and the device information of the target terminal in the first data sequence to obtain the network standard type of each operator in the operator list; The first data sequence is processed using a trained network quality prediction model to obtain the network quality level of each operator in the operator list.

[0018] In this implementation, the network quality prediction model is explicitly defined to output network type and network quality level separately. This separate prediction design allows the model to be specifically optimized for the two prediction tasks of "network type identification" and "service quality evaluation," which have different characteristics and patterns. Different model structures or feature processing methods may be used, thereby potentially improving the accuracy of each prediction task and the overall prediction performance of the model.

[0019] In conjunction with the first aspect, in one possible implementation of the first aspect, the method further includes: Acquire historical network access data, which is source data related to the network generated when multiple terminals accessed different operators' networks within a historical period; Determine the regional grid information corresponding to the historical location information based on the historical location information in the source data; By associating the historical time information, historical location information, corresponding regional grid information, and device information of each set of source data in the historical online data, a set of training data corresponding to each set of source data is constructed. The information related to network quality in each set of source data in the historical network data is quantified to obtain the network quality label corresponding to each set of source data; The training data corresponding to each set of source data and the network quality label corresponding to each set of source data are combined into a training sample, thus obtaining the training sample set. The network quality prediction model is trained using the training sample set to obtain a well-trained network quality prediction model.

[0020] This implementation details the data preparation and methods for training the network quality prediction model. A high-quality training sample set is constructed by spatially gridding, spatiotemporally correlating, and quantifying labels from massive, multi-source historical network access data of terminals. This process transforms chaotic and heterogeneous business data into a structured, feature-rich machine learning dataset with clear learning objectives (labels). This lays a solid data foundation for training a prediction model capable of accurately capturing the spatiotemporal variations in network quality, and is a core prerequisite for the realization and effectiveness of the entire AI prediction system.

[0021] In conjunction with the first aspect, in one possible implementation of the first aspect, the network quality label corresponding to each set of source data includes a network standard label and a network quality level label; the network standard label is extracted from the network standard information in each set of source data, and the network quality level label is obtained by quantifying the network quality-related information in each set of source data in historical network data.

[0022] This implementation further clarifies the composition of training data labels, including network standard labels and network quality level labels. Network standard labels are extracted directly from the raw data, ensuring label accuracy; network quality level labels are obtained through comprehensive quantitative scoring of multi-dimensional network-related information (such as signal strength, traffic volume, and anomalies), objectively and comprehensively reflecting the quality of a single network experience. This label construction method enables the model to simultaneously learn and predict specific access technologies (standards) and abstract service levels (levels), supporting the separate processing of prediction tasks.

[0023] Secondly, this application also provides an apparatus for allocating cloud SIM cards. The apparatus includes a request receiving module and an allocation module. The request receiving module is used to receive a SIM card allocation request, which requests the allocation of a cloud SIM card required for a target terminal to access the operator's network. The SIM card allocation request includes the location information of the target terminal when it initiates the SIM card allocation request. The allocation module is used to determine the target cloud SIM card from the set of cloud SIM cards corresponding to the operator list based on the network quality information of each operator in the operator list corresponding to the location information. The network quality information includes at least one of network standard type and network quality level. The allocation module then sends the target cloud SIM card to the target terminal.

[0024] Thirdly, this application also provides an electronic device. The electronic device includes a memory, one or more processors, and a computer program stored in the memory and executable on the processor. The electronic device executes the computer program to implement any of the implementations of the first aspect described above.

[0025] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the method of any of the implementations of the first aspect described above.

[0026] Fifthly, this application also provides a computer program product that, when run on an electronic device, causes the electronic device to execute any of the implementation methods of the first aspect described above.

[0027] It is understood that the beneficial effects of the second to fifth aspects mentioned above can be found in the relevant descriptions in the first aspect mentioned above, and will not be repeated here. Attached Figure Description

[0028] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 This is a structural block diagram of a cloud card allocation system provided in one embodiment of this application; Figure 2 This application provides an embodiment based on, as shown in, Figure 1 The diagram shown illustrates the process of cloud card allocation in the cloud card allocation system. Figure 3 This is a block diagram of the module structure of an AI prediction system provided in an embodiment of this application; Figure 4 This is a flowchart of a method for allocating cloud cards according to an embodiment of this application; Figure 5 This is a flowchart of a training method for a network quality prediction model provided in an embodiment of this application; Figure 6 This is a structural block diagram of a device for allocating cloud cards according to an embodiment of this application; Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0030] In the field of mobile communications, cloud SIM card technology, as a new user identity management solution, virtualizes the functions of the traditional physical SIM (Subscriber Identity Module) card, storing it in the cloud as a digital credential, which becomes the legitimate identity identifier for terminal devices to access the network. By reading the cloud SIM card credential, the terminal device initiates access authentication to a specific operator's network uniquely identified by the PLMN (Public Land Mobile Network) code. After successful authentication, a data connection can be established and communication can commence.

[0031] In this process, network quality (including but not limited to signal strength, data transmission rate, network latency, and stability) directly determines the user's actual service experience. Therefore, the core objective of the cloud SIM card allocation system is to dynamically and intelligently allocate suitable cloud SIM cards to terminals based on their real-time network environment, enabling them to access the optimal PLMN and thus ensuring users receive consistently good network service quality.

[0032] Current cloud SIM card allocation systems typically make allocation decisions based on the following two types of information: (1) real-time signal measurement data reported by the terminal device during a brief network scan when initiating a request, such as the signal reception power of each available PLMN; and (2) empirical data formed by statistical analysis of network indicators generated by historical users in different geographical locations, such as the average signal strength or throughput of each operator in a specific area. The system allocates the corresponding cloud SIM card to the terminal according to preset rules (e.g., prioritizing the PLMN with the strongest signal).

[0033] However, the above scheme has obvious limitations: (1) A short network search cannot reflect the full coverage of the network at the current location, and its decision-making mechanism relies on a single indicator dimension, which overemphasizes signal strength and cannot effectively reflect the actual network load, available bandwidth and end-to-end latency of services and other key quality elements. (2) The historical statistical data it uses has inherent lag and one-sidedness. Each network indicator (signal strength, rate) is viewed in isolation and lacks correlation analysis between multi-dimensional quality indicators, resulting in a deviation between the decision-making basis and the actual user experience.

[0034] The cloud card allocation system in the above scheme does not have the ability to predict and learn the network environment. If the terminal needs to perform deep or frequent network searches in order to obtain more comprehensive network information, it will significantly increase the power consumption burden of the terminal and prolong its service interruption time, which will have a negative impact on the user experience.

[0035] Therefore, this application provides a corresponding solution, including a method for allocating cloud SIM cards and a network quality prediction model used to implement the method. This network quality prediction model is deployed in an AI (Artificial Intelligence) prediction system, capable of predicting the network quality of different PLMNs (Publicly Licensed Mobile Networks) corresponding to the terminal's location based on the cloud SIM card allocation request received from the terminal. When the terminal requests cloud SIM card allocation, the AI ​​prediction system intelligently predicts the network type and service quality of currently available PLMNs based on multi-dimensional features such as device type, geographical location, and time carried in the request information.

[0036] Optionally, the AI ​​prediction system deploys two models capable of predicting network quality: one generates a comprehensive network quality level and network standard type for each PLMN. The network quality level accurately quantifies the expected service quality of each operator's network, while the network standard type indicates the type of network provided by the PLMN to the terminal, ensuring that the terminal can access the most suitable network layer. Based on the predicted network quality level and network standard type, a suitable cloud SIM card is allocated to the terminal, enabling it to access the optimal PLMN network and thus achieve high-quality network connectivity. The solution provided in this application effectively improves the efficiency and accuracy of network quality prediction, enhances the accuracy of cloud SIM card allocation, and improves network service quality through intelligent prediction and multi-dimensional evaluation.

[0037] like Figure 1 , Figure 1 This is a structural block diagram of a cloud card allocation system provided in one embodiment of this application. The cloud card allocation system involves a terminal 01, an access system 02, an AI prediction system 03, and a cloud card management and scheduling system 04. Communication connections are established between the access system 02 and the terminal 01, the AI ​​prediction system 03, and the cloud card management and scheduling system 04, respectively.

[0038] Among them, terminal 01 is the device terminal that applies for cloud cards; access system 02 is used to process business message interaction between the terminal; AI prediction system 03 is used to receive prediction requests submitted by access system 02 and predict the network quality level and network standard of each operator; cloud card management and scheduling system 04 is used to manage the cloud card pool and is responsible for card allocation and scheduling.

[0039] Figure 2 Based on such Figure 1 The diagram shown illustrates the process of cloud card allocation in the cloud card allocation system, including the following steps.

[0040] S201, Terminal 01 sends a card splitting request to Access System 02.

[0041] The SIM card splitting request is used to apply for a cloud SIM card so that terminal 01 can access the operator's network. The SIM card splitting request includes the following information: (1) Device IMEI (International Mobile Equipment Identity): The IMEI can accurately identify the specific terminal 01 device that initiated the request and bind it to the subsequently allocated cloud card; (2) Equipment type: mobile communication equipment, Internet of Things equipment, etc., representing the types of terminal 01, such as MIFI devices, shared power banks, car navigation, smart water meters, etc. (3) Device version: indicates the hardware or software specifications of terminal 01, including supported network standards (such as 4G / 5G), device capabilities, etc. (4) Longitude and latitude of the current location: representing the specific physical location (country, region) of terminal 01; (5) Card splitting time: The time when the device requests card splitting.

[0042] S202, Access system 02 sends a prediction request to AI prediction system 03 based on the card sub-request.

[0043] The prediction request is used to call the real-time prediction interface provided by the AI ​​prediction system 03. The prediction request includes information from the SIM card request and a list of PLMNs for the country where the terminal is located. At this time, the AI ​​prediction system 03 uses the information in the prediction request as a prerequisite for prediction, and predicts the network quality level of each operator in the country where the terminal 01 is currently located, as well as the corresponding network standard type, as the prediction result.

[0044] In some embodiments, before performing network quality prediction based on the data sequence corresponding to the card request, the data sequence is first converted to conform to a preset model input format. After obtaining the data sequence that meets the input format requirements, it is input into two models deployed in the AI ​​prediction system 03. The network quality prediction model includes a first model and a second model. The first model is used to predict the network type based on the current data sequence, and the second model is used to predict the network quality level based on the current data sequence.

[0045] S203, AI prediction system 03 returns real-time prediction results.

[0046] The AI ​​prediction system 03, based on the information input in the prediction request, calls the first model and the second model to predict the network standard type and network quality level for each operator, obtaining real-time prediction results. Specifically, the first model outputs the predicted network standard type for each operator, and the second model outputs the predicted network quality level for each operator.

[0047] The AI ​​prediction system 03 displays its real-time prediction results as a PLMN list, which includes operator PLMNs arranged in a preset order. Each PLMN corresponds to a predicted network standard type and a predicted network quality level. The AI ​​prediction system 03 then returns the real-time prediction results to the access system 02.

[0048] For example, the PLMN list includes: {PLMN1: RAT_CATEGORY=4, Score_level=9}; {PLMN2: RAT_CATEGORY=4, Score_level=8}, indicating the existence of two operators, PLMN1 and PLMN2. RAT_CATEGORY refers to the predicted network standard type of the PLMN, and Score_level refers to the predicted network quality level of the PLMN. In the example above, operator PLMN1 has a predicted network standard type of 4G and a network quality level of 9; operator PLMN2 has a predicted network standard type of 4G and a network quality level of 8.

[0049] Among them, RAT_CATEGORY corresponds to four network standard types: 2G, 3G, 4G, and 5G; and ten network quality levels, which are integers from 1 to 10. The higher the value of the network quality level, the better the network service quality provided by the operator.

[0050] S204, Access system 02 requests cloud card allocation from cloud card management and scheduling system 04 based on real-time prediction results.

[0051] At this point, access system 02 sends the real-time prediction results and the data sequence corresponding to the card allocation request to cloud card management and scheduling system 04. Cloud card management and scheduling system 04, based on preset card allocation rules, combined with the resource load within the cloud card pool and the real-time prediction results, selects the optimal operator's cloud card as the target cloud card and returns relevant information about the target cloud card, such as IMSI (International Mobile Subscriber Identity) and basic cloud card information (roaming list, prohibited registration list, etc.), to access system 02.

[0052] S205, the cloud card management and scheduling system 04 allocates cloud cards to the access system 02.

[0053] The cloud card management and scheduling system 04 returns the allocation results to the access system 02. The allocation results include the cloud card ISMI and the cloud card basic information.

[0054] S206, Access system 02 sends the allocation result to terminal device 01 through the card response.

[0055] Terminal 01 applies for a target cloud card based on the allocation result, requests access to the network of the operator corresponding to the target cloud card, and realizes Internet access.

[0056] The aforementioned cloud card allocation process is primarily based on real-time prediction results output by an AI prediction system, such as... Figure 3 As shown, Figure 3 This is a block diagram of the module structure of the AI ​​prediction system in this application.

[0057] The AI ​​prediction system 03 includes the following modules: database 31, data acquisition module 32, feature engineering module 33, model training module 34, real-time prediction module 35, and API interface 36.

[0058] Among them, database 31, data acquisition module 32, feature engineering module 33, and model training module 34 are used to train the network quality prediction model (including the first model and the second model) to obtain the trained network quality prediction model.

[0059] Both the first and second models are deployed in the AI ​​prediction system 03 to implement the tasks of the real-time prediction module 35; the API interface 36 is the interaction interface between the AI ​​prediction system 03 and the access system 02, and the access system 02 calls the API interface 36 to realize the real-time prediction of cloud card allocation.

[0060] The following explains the function of each module in the AI ​​prediction system 03.

[0061] Database 31 is used to store network-related source data reported in real time or periodically recorded by multiple terminals during use, and to save network-related contextual source data recorded by the business system. That is, database 31 stores network-related record data generated by different users accessing the networks of different operators using their terminals within a historical time period.

[0062] It is worth noting that the business data stored in the database has many categories, such as order data related to users, activity data, user traffic data, etc. The AI ​​prediction system in this application can also be extended to real-time business prediction in other scenarios. For different prediction tasks, it is only necessary to adjust the data type of the collected data and the appropriate training model, which will not be described in detail here.

[0063] Optionally, the database 31 may include, but is not limited to, the following initial source data subsets: (1) operator network indicator data reported by the terminal periodically: network signal and network standard; (2) traffic data reported periodically or in real time; (3) cloud card network registration abnormality record, which is the data reported when there is a cloud card network registration abnormality event during the terminal's internet access process; (4) card replacement event record, which is generated by the server after each terminal device actively initiates card replacement, and is used to record the event of the terminal replacing the cloud card during the internet access process.

[0064] The data acquisition module 32 is used to acquire and preprocess multiple initial source data subsets in the database 31, and filter out abnormal and duplicate data in the initial source data subsets.

[0065] By using contextual information, data from multiple different initial source data subsets are correlated to generate historical network access data. Each set of source data in the historical network access data includes device type, cloud card registration operator PLMN, location information, time information, network standard, signal strength, traffic information, and anomaly information.

[0066] The feature engineering module 33 is used to perform feature transformation on historical network data, extract data related to the network quality prediction task, and perform feature transformation on the extracted data to obtain a training sample set for training the network quality prediction model.

[0067] In this process, the feature engineering module 33 first performs format conversion and preprocessing on the historical network data, regenerating new data features. For example, it converts the historical location information (latitude and longitude) in the historical network data into H3 grid IDs to obtain the regional grid information corresponding to the historical location information. It converts the historical time information corresponding to each set of source data into the following format: (year, month, day, hour, minute, second, whether it is a weekend, whether it is a holiday). It converts the network standard type value to 2, 3, 4, 5, representing 2G, 3G, 4G, and 5G respectively. It converts the device type string to an integer value of 0 or 1, indicating whether 5G is not supported or is supported.

[0068] Next, the feature engineering module 33 continues to quantify the network quality-related information in each group of source data in the historical network data based on preset rules. It performs feature transformation on network standard information, signal strength, traffic volume, and other network quality information to obtain network quality labels corresponding to each group of source data: including network standard type labels and network quality level labels (network quality score / value). Each group of source data and its corresponding network quality label constitute a training sample. The above processing is performed on each group of source data in the historical network data to obtain a training sample set consisting of multiple training samples.

[0069] In some embodiments, feature engineering can also aggregate the training samples in the training sample set according to a certain time dimension (e.g., by hour) and specific rules. The aggregated dataset is smaller than the original training sample set. For example, for the training sample set corresponding to the network type prediction task, feature engineering aggregates multiple training samples of a certain PLMN at a certain location grid and appearing on the same day according to the highest type rule to generate a new training sample for predicting the network type. After processing all training samples in the training sample set according to this rule, a first sample set corresponding to the training sample set is obtained, and the first sample set is used to train the first model.

[0070] For example, in a certain location grid, both 4G and 5G network standards appear in multiple PLMN1 training samples on the same day. After aggregating all the training samples on that day, the network standard of the new PLMN1 training sample is 5G.

[0071] For example, for the training sample set corresponding to the network quality level prediction task, feature engineering is used to aggregate multiple training samples with different scores that appear in the same hour for a certain PLMN at a certain location information corresponding to a certain area grid. This is done according to a preset highest score rule (retaining the highest-scoring training sample) or average score rule (taking the average of the network quality scores of all training samples) to generate a new training sample for predicting the network quality level. After processing all training samples in the training sample set according to the above rules, a second sample set corresponding to the training sample set is obtained. The first model is trained using the second sample set.

[0072] For example, in a certain grid location area, among three PLMN1 training samples within one hour, each training sample has a network quality score of 82, 80, and 90. After aggregating according to the average score rule ((82+80+90) / 3=86), a new PLMN1 training sample is obtained, with a corresponding score of 86. The network quality level corresponding to this score is 9.

[0073] The first sample set is used to train the first model, enabling the first model to predict the network type. The second sample set is used to train the second model, enabling the second model to predict the network quality level.

[0074] The model training module 34 trains the first model based on the first sample set and trains the second model based on the second sample set, resulting in a trained first model and a trained second model. The first model and the second model together serve as network quality prediction models and are deployed in the AI ​​prediction system respectively.

[0075] When a target terminal requests the allocation of a cloud card, it interacts with the access system 02 through API interface 36 to obtain the data sequence corresponding to the card allocation request. The real-time prediction module 35 is then called to input the data sequence into the first model and the second model respectively to predict the predicted network standard type and the predicted network quality level. The prediction result is then returned to the access system 02 through API interface 36.

[0076] The method for allocating cloud cards provided in this application will be explained based on the above description, such as... Figure 4 As shown, Figure 4 This is a flowchart of a method for allocating cloud cards according to an embodiment of this application. The method is applied to a network device, wherein the network device can be configured to deploy the above-described method. Figure 1 The devices in the cloud card allocation system shown, including the access system, AI prediction system, and cloud card management and scheduling system, can also be deployed with the aforementioned features. Figure 3 The device for the AI ​​prediction system shown includes the following steps.

[0077] S410 receives card splitting requests.

[0078] The card splitting request is used to request the allocation of a cloud card required for the target terminal to access the operator's network. The card splitting request includes the location information of the target terminal when it initiates the card splitting request.

[0079] For example, when a user arrives at a location and needs to access the internet via a target terminal, the target terminal automatically sends a SIM card request to the access system, requesting a cloud SIM card so that the target terminal can access the network provided by the corresponding operator through the cloud SIM card. The cloud SIM card is the identity credential required for the target terminal to access the operator's network. The target terminal uses the identity information provided by the cloud SIM card to authenticate itself through the operator's network, and after successful authentication, it can access the internet.

[0080] This embodiment uses a network device deploying an AI prediction system as an example. The access system can be a system deployed on other electronic devices, used to realize interaction with the target terminal and forward the target terminal's SIM card allocation request to the network device with the AI ​​prediction system. It calls the API interface provided by the network device and uses the network quality prediction model in the AI ​​prediction system to predict the network quality of multiple operators corresponding to the target terminal's location. Network quality is a key factor in determining cloud SIM card allocation. When allocating a cloud SIM card to the target terminal, the cloud SIM card from the operator with the best network quality indicated in the prediction results will be prioritized for allocation.

[0081] S420 determines the target cloud card from the set of cloud cards corresponding to the operator list based on the network quality information of each operator in the operator list corresponding to the location information.

[0082] Optionally, the trained network quality prediction model is used to process the first data sequence corresponding to the SIM card request to obtain the network quality information of each operator in the operator list corresponding to the location information.

[0083] The network quality information includes at least one of the following: network standard type and network quality level.

[0084] The card splitting request also includes time information and target terminal device information. Based on the information included in the card splitting request, a first data sequence corresponding to the card splitting request is obtained. The first data sequence includes time information, area grid information corresponding to the location information, and target terminal device information; the area grid information corresponding to the location information is obtained by dividing the location information into areas.

[0085] Optionally, the location information included in the card request is the first location of the target terminal recorded in the form of longitude and latitude, and the device information includes the device type and device version information of the target terminal.

[0086] For example, when the network quality information includes network standard type and network quality level, the trained network quality prediction model is used to process the regional grid information corresponding to the location information in the first data sequence and the device information of the target terminal to obtain the network standard type of each operator in the operator list.

[0087] For example, when network quality information includes network standard type and network quality level, a trained network quality prediction model is used to process the first data sequence to obtain the network quality level of each operator in the operator list. That is, the network quality prediction model predicts the network quality of each operator in the operator list corresponding to the location information, and outputs the results from two dimensions: network standard type and network quality level.

[0088] For example, there are a total of 5 operators corresponding to the first position. The network quality information of each operator includes the prediction result of the network standard type of the operator and the prediction result of the network quality level of the operator.

[0089] After obtaining the first data sequence, in order to facilitate the analysis of the network quality prediction model and make the format of the first data sequence more in line with the input requirements of the network quality prediction model, the format of each data in the first data sequence can be preprocessed.

[0090] Optionally, feature analysis is performed on the time information to obtain time series information. The time series information is used to indicate the periodic characteristics of the request time, where the time information refers to the request time when the target terminal initiates the card splitting request. Specifically, periodic time series features such as year, month, day, hour, minute, second, whether it is a weekend, and whether it is a holiday are extracted from the request time information to capture the periodic or trend patterns of the data in the time dimension. For example, the request time information for the card splitting request is 08:25:05 on January 1, 2024. By performing feature analysis on it, the timestamp of the request time information is parsed and matched with the preset calendar rules to extract basic time units such as year (2024), month (1), date (1), hour (8), minute (25), and second (5). At the same time, it is determined whether the timestamp is a weekend (0, no) and whether it is a holiday (1, New Year's Day holiday). Together, they constitute structured time series information.

[0091] Among them, variable 1: whether it is a weekend, variable 2: whether it is a holiday. The values ​​of variable 1 and variable 2 can be 0 or 1, where 0 means "no" and 1 means "yes".

[0092] This feature extraction process transforms precise time points into time-series features with periodic patterns, enabling network quality prediction models to accurately identify whether time information occurs during specific periods such as holidays, rest days, or peak times of the day. This captures the periodic patterns of network usage and provides key time dimension inputs for predicting network load and quality changes for different operators during this time period.

[0093] Optionally, the area grid where the target terminal is located is determined based on the location information to obtain area grid information. The location information is used to indicate the first location of the target terminal when it initiates the card splitting request, and the area grid is a grid obtained by dividing the ground location in a specified way.

[0094] The first location is the longitude and latitude coordinates under the GPS (Global Positioning System). The H3 geospatial indexing system is used as the specified method to convert the original GPS coordinates into grid IDs in the H3 geospatial indexing system, thereby obtaining the regional grid information. The grid ID is used to indicate the regional grid to which the coordinates belong under this method.

[0095] The resolution of the grid ID can be selected according to different accuracy requirements. In this embodiment, the h3_res11 resolution is used as the target resolution for explanation.

[0096] This process transforms continuous coordinate points into discrete spatial units, enabling the model to transcend the limitations of specific coordinates and instead learn regional network quality patterns in grid units.

[0097] For example, the first location coordinates (latitude 39.9163, longitude 116.3970), after transformation, yields the H3 grid ID "8b31aa428826fff" at level 11 resolution. This grid ID represents a hexagonal region with an area of ​​approximately 0.1 square kilometers.

[0098] After obtaining time series information and regional grid information, the network quality prediction model is used to process the time series information, regional grid information and target terminal device information to obtain the network quality information of each operator in the operator list corresponding to the location information.

[0099] The above method incorporates time, location, and device information into a first data sequence, performs feature analysis on time, and grids the location data before submitting it to a network quality prediction model for comprehensive analysis. This process achieves multi-dimensional information fusion, considering temporal periodicity, spatial location characteristics, and differences in device capabilities. This allows the prediction results to more accurately reflect network conditions at different times, locations, and with different devices, significantly improving the accuracy and scenario adaptability of the prediction results and overcoming the shortcomings of related technologies that rely on a single decision-making dimension.

[0100] This application uses a network quality prediction model including a first model and a second model as an example for illustration. The network quality prediction model may also include other models to perform different tasks related to network quality prediction, and this application does not limit this. In some embodiments, a network quality prediction model that can simultaneously perform the above two prediction tasks (predicting network type and network quality level) can also be trained to implement the cloud card allocation method proposed in this application.

[0101] For example, the first data sequence is analyzed using the first model to obtain the network type of each operator in the operator list corresponding to the location information. The value range of the network type is as follows: (1) a value of 2 indicates a 2G network; (2) a value of 3 indicates a 3G network; (3) a value of 4 indicates a 4G network; (4) a value of 5 indicates a 5G network.

[0102] 3 indicates that the first model of the 5G network can be viewed as a mapping function, which includes the correspondence between regional grid information, target terminal device information, and network standard type. By using the regional grid information and target terminal device information as input variables and applying the mapping function, the result is the prediction of the network standard type of the operator corresponding to the first position of the target terminal.

[0103] For example, the second model processes the first data sequence to obtain the network quality level of each operator in the operator list corresponding to the location information. The network quality level is an integer between [1, 10]. The larger the value, the higher the network quality level, the higher the level of network service that the operator can provide, and the better the experience when using the cloud card provided by the operator to access the Internet.

[0104] The second model can be viewed as another mapping function, which includes the correspondence between time series information, regional grid information, target terminal device information, and network quality level. By using the time series information, regional grid information, and target terminal device information as input variables and applying the mapping function, the result is the predicted network quality level of the operator corresponding to the first location of the target terminal.

[0105] After integrating the outputs of the two models, complete network quality information is obtained, which is represented as a list of PLMNs. Each list contains the network type and network quality level of each PLMN.

[0106] It is worth noting that before inputting the first data sequence into the first model and the second model, it is necessary to obtain information on multiple operators corresponding to the first location of the target terminal. The target data sequence, together with the information of multiple operators, is then input into the first model and the second model so that the first model and the second model can make predictions according to the different types of operators.

[0107] After obtaining network quality information, the target cloud card can be determined based on the network standard type and network quality level of multiple operators indicated in the network quality information.

[0108] Optionally, after receiving the prediction results returned by the AI ​​prediction system, the access system forwards the prediction results to the cloud SIM card management and scheduling system, which then makes a decision to determine the target cloud SIM card. Alternatively, the network device can make a decision directly based on the prediction results.

[0109] Optionally, when the network quality information includes network standard type and network quality level, a first cloud card set is determined from the cloud card set corresponding to the operator list based on the network standard type of each operator in the operator list corresponding to the location information. The first cloud card set includes at least one cloud card, and the network standard type of each cloud card in the first cloud card set is the network standard type supported by the target terminal.

[0110] For example, there are five PLMNs available for the current location, with their network standards and predicted quality of service levels as follows: {plmn1, rat_category=5, score_level=9; Plmn2, rat_category=4, score_level=8; Plmn3, rat_category=4, score_level=7; Plmn4, rat_category=3, score_level=1; Plmn5, rat_category=4, score_leve=5}.

[0111] The system first determines whether the target terminal supports 4G / 5G network standards based on the target terminal's device information, thus excluding operator PLMN4, which only supports 3G. The remaining operators PLMN1, PLMN2, PLMN3, and PLMN5, which support either 4G or 5G, constitute the operator range corresponding to the first cloud card set. Cloud cards belonging to PLMN1, PLMN2, PLMN3, and PLMN5 constitute the first cloud card set.

[0112] The target cloud card is determined from the first set of cloud cards based on the network quality level of each operator in the operator list corresponding to the location information.

[0113] The methods for determining the target cloud card based on network quality level include the following two: 1. Based on the network quality level of each operator in the operator list corresponding to the location information, determine the cloud card with the highest network quality level of the corresponding operator in the first cloud card set as the target cloud card; For example, in the above example, if the operator with the highest network quality level is PLMN1, then any cloud card belonging to PLMN1 can be selected as the target cloud card.

[0114] 2. Alternatively, the target cloud card can be determined from the first set of cloud cards based on the network quality level and load of each operator in the operator list corresponding to the location information.

[0115] First, based on the network quality level of each operator in the operator list corresponding to the location information, a second cloud card set is determined from the first cloud card set, and the second cloud card set includes at least one cloud card.

[0116] The second cloud card set can be determined in two ways, and either one can be chosen as the method used when obtaining the second cloud card set: 2.1 For example, cloud cards in the first cloud card set whose network quality level of the corresponding operator is higher than the first level threshold are merged into a second cloud card set, where the first level threshold is a preset fixed value; For example, if the first level threshold is set to 7 (level ≥ 7), then among the operators (PLMN1, PLMN2, PLMN3, PLMN5) corresponding to the first cloud card set, the operators with network quality levels higher than this threshold are PLMN1 (level 9), PLMN2 (level 8), and PLMN3 (level 7). Cloud cards belonging to these operators constitute the second cloud card set.

[0117] 2.2 For example, the cloud cards of the n operators with the highest network quality rankings in the first cloud card set are merged into a second cloud card set, where n is a positive integer and is less than or equal to the total number of operators corresponding to the cloud cards in the first cloud card set.

[0118] For example, let n=3. Among the operators (PLMN1, PLMN2, PLMN3, PLMN5) corresponding to the first cloud SIM card set, sort them from highest to lowest network quality level, and select the top 3 operators: PLMN1 (Level 9), PLMN2 (Level 8), and PLMN3 (Level 7). The cloud SIM cards of these operators constitute the second cloud SIM card set.

[0119] Secondly, based on the load of each operator in the operator list corresponding to the location information, the target cloud card is determined from the second cloud card set, and the operator corresponding to the target cloud card has the minimum load.

[0120] For example, taking method 2.2 above as an example, it is known that the operators corresponding to the second cloud card set are PLMN1, PLMN2, and PLMN3. The real-time load data in the cloud card management and scheduling system is queried to determine the load of these three operators.

[0121] Among them, the load measurement indicators include two types: (1) the number of remaining cloud cards; the more remaining cloud cards, the smaller the load of the operator; (2) the load rate, which is the ratio of the number of cloud cards that have been allocated to the total number of cloud cards; the smaller the load rate, the smaller the load of the operator.

[0122] For example, suppose the load rates of the three operators are as follows: PLMN1 load rate 85%, PLMN2 load rate 45%, and PLMN3 load rate 30%. The system selects the operator with the lowest load, PLMN3, and selects an available cloud card from its cloud card pool as the target cloud card.

[0123] S430 sends the target cloud card to the target terminal.

[0124] The target terminal accesses the network of the operator to which the target cloud card belongs through the target cloud card to achieve internet access.

[0125] In summary, the cloud SIM card allocation method provided in this application receives a SIM card allocation request containing terminal location information and selects a target cloud SIM card based on the network quality information (network type, quality level) of each operator corresponding to that location. This allows for advance knowledge of the overall service level of each available network at the current location at the time of SIM card allocation decision, thereby achieving optimal cloud SIM card allocation. This not only avoids increased power consumption and prolonged service unavailability caused by prolonged or frequent network searches by the terminal, but also proactively matches the terminal with a cloud SIM card from an operator with better network quality at the current geographical location, significantly improving the user's network service experience and connection reliability, and realizing intelligent and precise scheduling of cloud SIM card resources.

[0126] Figure 5 This is a flowchart of a training method for a network quality prediction model provided in an embodiment of this application. The method is applied to a training device, which can be configured to deploy the above-described method. Figure 3 The device for the AI ​​prediction system shown can also be implemented by deploying the above-mentioned... Figure 3 The device shown in the AI ​​prediction system includes some modules (data acquisition module 32, feature engineering module 33, and model training module 34). The method includes the following steps.

[0127] S510, retrieves historical network access data.

[0128] Historical network access data refers to network-related source data generated when multiple terminals access different operators' networks within a historical period. In other words, historical network access data includes multiple sets of source data.

[0129] Each set of source data includes: the terminal's historical location information, device information (device type and device version), historical time information, registered operator, network standard information, signal strength, and data traffic.

[0130] S520 determines the regional grid information corresponding to the historical location information based on the historical location information in the source data.

[0131] Among them, the regional grid is a grid obtained by dividing the land surface location in a specified way, and the regional grid information corresponding to the historical location information refers to the grid ID of that regional grid.

[0132] S530 associates the historical time information, historical location information, corresponding regional grid information, and device information of each set of source data in the historical network data to construct a set of training data corresponding to each set of source data.

[0133] S540 quantifies the network quality-related information in each set of source data in historical network data to obtain the network quality label corresponding to each set of source data.

[0134] The network quality labels for each set of source data include network type labels and network quality level labels.

[0135] Network standard labels are extracted from network standard information in each set of source data, while network quality level labels are obtained by quantifying information related to network quality in each set of source data in historical network data.

[0136] For example, a network standard type label is generated based on network standard information. A network quality score for the registered operator is obtained based on traffic volume, signal strength, and network standard information according to preset scoring rules, and a network quality level label is generated based on the network quality score.

[0137] Optionally, for each set of source data: the signal strength is divided into multiple different signal intervals according to multiple preset signal strength threshold ranges, and each signal interval corresponds to a different preset score; the traffic volume is divided into different traffic intervals according to multiple preset threshold ranges, and each traffic interval corresponds to a different preset score and weight; when the network standard information of the source data is 2G or 3G network, the source data does not consider signal strength and traffic volume, and the original operator network quality score is 0; when the source data has cloud card registration abnormality or card replacement event tags, it indicates that the cloud card is used abnormally and will not generate traffic, so the signal strength and traffic volume are not considered, and the original operator network quality score is 0; when the network standard in the source data is 4G or 5G network, the traffic score F and traffic weight are obtained according to the traffic interval where the traffic volume is located, and the signal score is obtained according to the signal interval where the signal strength is located. The network quality score of the registered operator is calculated based on the following formula: Score = (1-Wf)*S+ Wf*F, where Score is the network quality score, Wf is the traffic weight, S is the signal score, and F is the traffic score.

[0138] The network quality score is converted into a network quality level label based on a preset score range.

[0139] For example, when extracting network standard type labels from network standard information, the following rules can be followed: (1) GSM (second generation (2G) mobile communication technology), network type label is 2; indicating that when the device is connected to the GSM network, its network type label is 2; (2) WCDMA (Wideband Code Division Multiple Access, third generation (3G) mobile communication standard), network type label is 3; indicating that when the device is connected to the WCDMA network, its network type label is 3; (3) FDD LTE (Frequency Division Duplex Long Term Evolution) / TDD LTE (Time Division Duplex Long Term Evolution), network standard type label is 4; where LTE is the fourth generation (4G) communication technology, it means that when the device is connected to the FDD LTE / TDD LTE network, its network standard type label is 4; (4) 5G NSA (Non-Standalone) / SA (Standalone), the network standard type label is 5; NSA and SA are two deployment architectures of 5G (Fifth Generation Mobile Communication Technology) network, indicating that when the device is connected to the 5G NSA / SA network, its network standard type label is 5.

[0140] The network quality-related information in each set of source data includes various network indicators (network type, signal strength, traffic volume, reasons for cloud card network access anomalies, reasons for card replacement), and these various network indicators are evaluated to obtain the network quality level.

[0141] Specifically, first, the score and weight corresponding to each network indicator are calculated. Then, a weighted calculation is performed based on the score and weight to obtain the network quality score. The network quality level to which the score belongs is then divided according to a preset score range to obtain the corresponding network quality level target features. The evaluation rules used in this process, including various network indicator information, are illustrated in the following example.

[0142] I. Regarding network standards and signal strength: (1) If the network standard is 2G / 3G, the network is poor, the signal strength level is not distinguished, the signal strength score is 0, and the signal weight is 100%; (2) If the network standard is 4G / 5G, the signal strength can be divided into different intervals according to the preset threshold (db value), and the corresponding level score of each interval is {S1,S2,...,Sn} (configurable).

[0143] II. Regarding traffic volume: The traffic volume is divided into different intervals according to a preset threshold. The scores of different intervals are {F1, F2, ..., Fn}, and the corresponding weights are {Wf1, Wf2, ..., Wfn}, where Wf∈[0, 100%].

[0144] III. Regarding the reasons for abnormal cloud card network registration: If the terminal experiences network dropout or congestion during cloud card usage, an event report will be triggered. Such an event indicates that the network service is interrupted, no traffic is generated, the traffic score is 0, and the weight is 100%.

[0145] IV. Regarding the reasons for changing the card: If an anomaly occurs during cloud card usage, such as dial-up failure or registration rejection, the terminal will proactively request a card replacement and report the reason for the replacement in the corresponding scenario. In this scenario, no data traffic is generated. The traffic score is 0, and the weight is 100%.

[0146] The final network quality score is: Score = (1-Wf)*S + Wf*F.

[0147] S is determined based on the range of signal strength in each set of source data, while F and Wf are determined based on the range of traffic volume in each set of source data.

[0148] Based on the preset score range, the score is converted into a rating level. The specific relationship between the score range and the rating level is as follows: >=90, Level: 10; [80,90), Level: 9; [70,80), Level: 8; [60,70), Level: 7; [50,60), Level: 6; [40,50), Level: 5; [30,40), Level: 4; [20,30), Level: 3; [10,20), Level: 2; [0,10), Level: 1.

[0149] S550 combines the training data corresponding to each set of source data and the network quality label corresponding to each set of source data into a training sample, thereby obtaining a training sample set.

[0150] For example, a network quality prediction model including a first model and a second model is used to illustrate the problem. The first model is used to predict the network type, and the second model is used to predict the network quality level.

[0151] A first training sample set can be obtained by combining the training data corresponding to each set of source data with the network standard label corresponding to each set of source data. This first sample set is used to train the first model. Simultaneously, a second training sample set can be obtained by combining the training data corresponding to each set of source data with the network quality level label corresponding to each set of source data. This second sample set is used to train the second model. The first and second sample sets together constitute the training sample set.

[0152] S560 uses the training sample set to train the network quality prediction model, resulting in a well-trained network quality prediction model.

[0153] For example, a first model is trained using a first sample set to obtain a trained first model, and a second model is trained using a second sample set to obtain a trained second model.

[0154] The trained first and second models together serve as a trained network quality prediction model to perform the network quality prediction task and output the corresponding network quality information based on the input information.

[0155] Optionally, for each first training sample in the first sample set, the following operations are performed: inputting the regional grid information, device information, and registered operator in the first training sample as training features into the first model, and inputting the network standard type label as the target feature into the first model; the training process of the first model aims to reduce the error between the predicted network standard type output by the first model and the network standard type label in the first sample, and by selecting different machine learning algorithm models, determining the optimal first model, and determining the optimal first model as the trained first model.

[0156] Wherein, when the first model meets the first preset training conditions, the training is considered to be completed. The first preset training conditions include at least one of the following: the first model has completed a preset number of training rounds; the first model has been trained using a preset number of first training samples; all first training samples in the first sample set have been used for training; and the prediction accuracy of the updated first model meets the preset accuracy conditions.

[0157] Optionally, for each second training sample in the second sample set, the following operations are performed: inputting the regional grid information, time information, device information, and registered operator in the second sample as training features into the second model, and inputting the network quality level label as the target feature into the second model; the training process of the second model aims to reduce the error between the predicted network quality level output by the second model and the network quality level label in the second sample, and by selecting different machine learning algorithm models, determining the optimal second model, and determining the optimal second model as the trained second model.

[0158] The training is considered complete when the second model meets the second preset training conditions. The second preset training conditions include at least one of the following: the second model has completed a preset number of training rounds; the second model has been trained using a preset number of second training samples; all second training samples in the second sample set have been used for training; and the prediction accuracy of the updated second model meets the preset accuracy conditions.

[0159] In this application, both the first model for predicting network type and the second model for predicting network quality level use classification algorithm models as the initial model architecture. The first and second models are trained separately using two independent model architectures. After deployment, each trained model provides an API interface. Finally, at the business code level, the two API interfaces are encapsulated in code and integrated into a single API interface for clients accessing the system to call.

[0160] In summary, the training method for the network quality prediction model provided in this application offers an efficient and targeted model training approach by specifying the use of a first sample set with network type target features and a second sample set with network quality level target features for independent training of the first and second models. This technical solution ensures that the first model can accurately learn the distribution patterns of network types and that the second model can deeply understand the evaluation criteria for comprehensive network quality by preparing clean and focused training data for different prediction tasks, thus laying a solid foundation for obtaining a high-precision network quality prediction model.

[0161] Corresponding to the method of allocating cloud cards in the above embodiment, Figure 6 A structural block diagram of an apparatus for predicting network quality provided in an embodiment of this application is shown. For ease of explanation, only the parts related to the embodiments of this application are shown.

[0162] Reference Figure 6 The device 600 includes a request receiving module 610 and an allocation module 620. The request receiving module 610 receives a card allocation request, which requests the allocation of a cloud card required for a target terminal to access the operator's network. The card allocation request includes the location information of the target terminal when it initiates the card allocation request. The allocation module 620 determines the target cloud card from the set of cloud cards corresponding to the operator list based on the network quality information of each operator in the operator list corresponding to the location information. The network quality information includes at least one of network standard type and network quality level. The allocation module 620 then sends the target cloud card to the target terminal. This device 600 can be integrated into an electronic device, including but not limited to servers and terminals with high computing power (meeting a specified standard).

[0163] The device 600 can execute steps S410 to S430 above, and can also execute the entire process of cloud card allocation above.

[0164] It should be noted that the information interaction and execution process between the above-mentioned devices / modules are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, and they will not be repeated here.

[0165] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0166] To implement the above embodiments, this application also proposes an electronic device. Figure 7 This is a schematic diagram of the structure of an electronic device according to an embodiment of this application.

[0167] like Figure 7 As shown, the above-mentioned electronic device 700 includes: The system includes a memory 710 and at least one processor 720, and a bus 730 connecting different components (including the memory 710 and the processor 720). The memory 710 stores a computer program, which, when executed by the processor 720, implements the method for allocating cloud cards and the method for training a network quality prediction model according to the embodiments of this application.

[0168] Bus 730 represents one or more of several bus architectures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus using any of the various bus architectures. For example, these architectures include, but are not limited to, the Industry Standard Architecture (ISA) bus, the Micro Channel Architecture (MAC) bus, the Enhanced ISA bus, the Video Electronics Standards Association (VESA) local bus, and the Peripheral Component Interconnect (PCI) bus.

[0169] Electronic device 700 typically includes a variety of electronic device readable media. These media can be any available media that can be accessed by electronic device 700, including volatile and non-volatile media, removable and non-removable media.

[0170] The memory 710 may also include computer system readable media in the form of volatile memory, such as random access memory (RAM) 740 and / or cache memory 750. The electronic device 700 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, the storage system 760 can be used to read and write non-removable, non-volatile magnetic media (…). Figure 7 Not shown; usually referred to as a "hard drive"). Although Figure 7 As not shown, a disk drive for reading and writing to a removable non-volatile disk (e.g., a "floppy disk") and an optical disk drive for reading and writing to a removable non-volatile optical disk (e.g., a CD-ROM, DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to bus 730 via one or more data media interfaces. Memory 710 may include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of the embodiments of this application.

[0171] A program / utility 780 having a set (at least one) of program modules 770 may be stored in, for example, memory 710. Such program modules 770 include—but are not limited to—an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment. Program modules 770 typically perform the functions and / or methods described in the embodiments of this application.

[0172] Electronic device 700 can also communicate with one or more external devices 790 (e.g., keyboard, pointing device, display 771, etc.), and with one or more devices that enable a user to interact with electronic device 700, and / or with any device that enables electronic device 700 to communicate with one or more other computing devices (e.g., network card, modem, etc.). This communication can be performed via input / output (I / O) interface 777. Furthermore, electronic device 700 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via network adapter 773. As shown, network adapter 773 communicates with other modules of electronic device 700 via bus 730. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with electronic device 700, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0173] The processor 720 performs various functional applications and data processing by running programs stored in the memory 610.

[0174] It should be noted that the implementation process and technical principles of the electronic device in this embodiment are explained in the foregoing description of the method for allocating cloud cards and the training method for the network quality prediction model in the embodiments of this application, and will not be repeated here.

[0175] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps described in the various method embodiments above.

[0176] This application provides a computer program product that, when run on an electronic device, enables the electronic device to perform the steps described in the various method embodiments above.

[0177] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of this application can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. A computer-readable medium can include at least: any entity or device capable of carrying computer program code to a photographic device / electronic device, a recording medium, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks. In some regions, according to legislation and patent practice, computer-readable media cannot be electrical carrier signals or telecommunication signals.

[0178] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0179] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0180] In the embodiments provided in this application, it should be understood that the disclosed devices / electronic devices and methods can be implemented in other ways. For example, the device / electronic device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0181] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0182] In the foregoing, specific details such as particular system architectures and techniques have been set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application can also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted to avoid unnecessary detail from obscuring the description of this application.

[0183] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0184] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0185] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."

[0186] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0187] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0188] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for allocating cloud cards, characterized in that, The method includes: Receive a card allocation request, the card allocation request is used to request the allocation of a cloud card required for the target terminal to access the operator's network, the card allocation request includes the location information of the target terminal when it initiates the card allocation request; Based on the network quality information of each operator in the operator list corresponding to the location information, the target cloud card is determined from the cloud card set corresponding to the operator list. The network quality information includes at least one of network standard type and network quality level. Send the target cloud card to the target terminal.

2. The method according to claim 1, characterized in that, The card splitting request also includes time information and device information of the target terminal, and the method further includes: The first data sequence corresponding to the SIM card request is processed using a trained network quality prediction model to obtain network quality information for each operator in the operator list corresponding to the location information; the first data sequence includes the time information, the regional grid information corresponding to the location information, and the device information of the target terminal; the regional grid information corresponding to the location information is obtained by dividing the location information into regions.

3. The method according to claim 1, characterized in that, When the network quality information includes network standard type and network quality level, determining the target cloud card from the cloud card set corresponding to the operator list based on the network quality information of each operator in the operator list corresponding to the location information includes: Based on the network standard type of each operator in the operator list corresponding to the location information, a first cloud card set is determined from the cloud card set corresponding to the operator list. The first cloud card set includes at least one cloud card, and the network standard type of each cloud card in the first cloud card set is the network standard type supported by the target terminal. The target cloud card is determined from the first cloud card set based on the network quality level of each operator in the operator list corresponding to the location information.

4. The method according to claim 3, characterized in that, The step of determining the target cloud card from the first cloud card set based on the network quality level of each operator in the operator list corresponding to the location information includes: Based on the network quality level of each operator in the operator list corresponding to the location information, the cloud card with the highest network quality level of the corresponding operator in the first cloud card set is determined as the target cloud card. Alternatively, the target cloud card can be determined from the first set of cloud cards based on the network quality level and load of each operator in the operator list corresponding to the location information.

5. The method according to claim 4, characterized in that, The step of determining the target cloud SIM card from the first cloud SIM card set based on the network quality level and load of each operator in the operator list corresponding to the location information includes: Based on the network quality level of each operator in the operator list corresponding to the location information, a second cloud card set is determined from the first cloud card set, and the second cloud card set includes at least one cloud card; The target cloud card is determined from the second cloud card set based on the load of each operator in the operator list corresponding to the location information, and the operator corresponding to the target cloud card has the lowest load.

6. The method according to claim 5, characterized in that, The second cloud SIM card set is determined from the first cloud SIM card set based on the network quality level of each operator in the operator list corresponding to the location information. The second cloud SIM card set includes at least one cloud SIM card, including: The cloud cards in the first cloud card set whose network quality level of the corresponding operator is higher than the first level threshold are merged into the second cloud card set, where the first level threshold is a preset fixed value. Alternatively, the cloud cards of the n operators with the highest network quality rankings in the first cloud card set can be merged into the second cloud card set, where n is a positive integer and is less than or equal to the total number of operators corresponding to the cloud cards in the first cloud card set.

7. The method according to claim 2, characterized in that, When the network quality information includes network standard type and network quality level, the process of using a trained network quality prediction model to process the first data sequence corresponding to the SIM card request to obtain the network quality information of each operator in the operator list corresponding to the location information includes: The trained network quality prediction model is used to process the area grid information corresponding to the location information in the first data sequence and the device information of the target terminal to obtain the network standard type of each operator in the operator list; The trained network quality prediction model is used to process the first data sequence to obtain the network quality level of each operator in the operator list.

8. The method according to claim 2, characterized in that, The method further includes: Acquire historical network access data, which is network-related source data generated when multiple terminals access different operator networks within a historical period; Determine the regional grid information corresponding to the historical location information based on the historical location information in the source data; By associating the historical time information, historical location information, corresponding regional grid information, and device information of each set of source data in the historical online data, a set of training data corresponding to each set of source data is constructed. The information related to network quality in each group of source data in the historical network data is quantified to obtain the network quality label corresponding to each group of source data; The training data corresponding to each set of source data and the network quality label corresponding to each set of source data are combined into a training sample, thus obtaining the training sample set. The network quality prediction model is trained using the training sample set to obtain the trained network quality prediction model.

9. The method according to claim 8, characterized in that, The network quality labels corresponding to each set of source data include network type labels and network quality level labels; the network type labels are extracted from the network type information in each set of source data, and the network quality level labels are obtained by quantifying the network quality-related information in each set of source data in the historical network data.

10. A device for allocating cloud cards, characterized in that, The device includes: A request receiving module is used to receive a card allocation request, which is used to request the allocation of a cloud card required for the target terminal to access the operator's network. The card allocation request includes the location information of the target terminal when it initiates the card allocation request. The allocation module is used to determine the target cloud card from the cloud card set corresponding to the operator list based on the network quality information of each operator in the operator list corresponding to the location information. The network quality information includes at least one of network standard type and network quality level. The allocation module is also used to send the target cloud card to the target terminal.

11. An electronic device, characterized in that, The device includes one or more processors, a memory, and a computer program stored in the memory and executable on the processor, characterized in that, when the one or more processors execute the computer program, the electronic device causes the electronic device to perform the method as described in any one of claims 1 to 9.

12. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 9.

Images