Cell avoidance method, system, electronic device and storage medium

Abstract

Embodiments of the present disclosure provide a cell avoidance method, system, electronic device and storage medium. The cell avoidance method comprises: using a first prediction model to predict an abnormal probability of a cell associated with a first terminal based on first space-time information of the first terminal, wherein the first prediction model is obtained by training a network behavior data collected by a second terminal on a server; and performing a corresponding avoidance operation on an abnormal cell with an abnormal probability greater than a probability threshold.

Description

Technical Field

[0001] This invention relates to the field of communications, and in particular to a cell avoidance method, system, electronic device, and storage medium. Background Technology

[0002] In cellular networks, abnormal cells may exist due to factors such as incorrect parameter configuration in the access network or core network, terminal compatibility issues, and uneven network load. In related technologies, terminals (such as mobile phones) primarily rely on the fault recovery mechanisms defined by the 3GPP protocol to passively respond to abnormal cells.

[0003] The fault recovery mechanisms defined by the 3GPP protocol usually attempt to avoid the cell only after an anomaly occurs, which cannot prevent it in advance, and the interruption of user experience cannot be reduced.

[0004] Therefore, how to avoid abnormal cells in advance, reduce terminal communication anomalies, and improve user experience is an urgent problem to be solved. Summary of the Invention

[0005] This disclosure provides cell evasion methods, systems, electronic devices, and storage media.

[0006] According to a first aspect of the present disclosure, a cell avoidance method is proposed, applied to a first terminal, the method comprising: Using a first prediction model, based on the first spatiotemporal information of the first terminal, the probability of anomalies in the cell associated with the first terminal is predicted. The first prediction model is trained on the server based on network behavior data collected by the second terminal. Perform corresponding avoidance operations on abnormal cells whose anomaly probability is greater than the probability threshold.

[0007] In some embodiments, performing the corresponding avoidance operation on abnormal cells with an anomaly probability greater than a probability threshold includes at least one of the following: In response to the anomaly probability being greater than or equal to a first probability threshold and less than or equal to a second probability threshold, the measured value of the anomalous cell is negatively biased based on the anomaly probability, wherein the second probability threshold is greater than the first probability threshold. In response to the anomaly probability being greater than the second probability threshold, the abnormal cell will not be camped for at least the prohibition period, wherein the prohibition period is determined based on the anomaly probability.

[0008] In some embodiments, the bias value for negatively biasing the measured value includes: the product of a first probability weight and a first preset value, plus the sum of a second preset value, wherein the first probability weight is positively correlated with the anomaly probability.

[0009] In some embodiments, the prohibition duration includes: the product of a second probability weight and a first duration, plus the sum of the second durations, wherein the second probability weight is positively correlated with the anomaly probability.

[0010] In some embodiments, the network behavior data includes: The identification information of the second terminal; The second spatiotemporal information corresponding to the second terminal; The cell identifier corresponding to the network behavior of the second terminal; The network type in which the second terminal performs network behavior; The result of the second terminal performing network actions at the corresponding network layer.

[0011] In some embodiments, the first prediction model includes a lightweight format model of the second prediction model obtained by the server based on the network behavior data.

[0012] According to a second aspect of the present disclosure, a cell avoidance method is proposed, applied to a server, the method comprising: A first prediction model is obtained by training based on network behavior data collected by the second terminal. The first prediction model is used to enable the first terminal to predict the anomaly probability of the cell associated with the first terminal based on the first spatiotemporal information of the first terminal, and to perform corresponding avoidance operations on the abnormal cells whose anomaly probability is greater than the probability threshold.

[0013] In some embodiments, the first terminal performs a corresponding avoidance operation for abnormal cells with an abnormal probability greater than a probability threshold, including at least one of the following: The first terminal responds to the fact that the anomaly probability is greater than or equal to a first probability threshold and less than or equal to a second probability threshold by negatively biasing the measured value of the abnormal cell based on the anomaly probability, wherein the second probability threshold is greater than the first probability threshold; The first terminal responds to the anomaly probability being greater than the second probability threshold by not camping on the abnormal cell for at least the prohibition period, wherein the prohibition period is determined based on the anomaly probability.

[0014] In some embodiments, the bias value for negatively biasing the measured value includes: the product of a first probability weight and a first preset value, plus the sum of a second preset value, wherein the first probability weight is positively correlated with the anomaly probability.

[0015] In some embodiments, the prohibition duration includes: the product of a second probability weight and a first duration, plus the sum of the second durations, wherein the second probability weight is positively correlated with the anomaly probability.

[0016] In some embodiments, the network behavior data includes: The identification information of the second terminal; The second spatiotemporal information corresponding to the second terminal; The cell identifier corresponding to the network behavior of the second terminal; The network type in which the second terminal performs network behavior; The result of the second terminal performing network actions at the corresponding network layer.

[0017] In some embodiments, the first prediction model includes a lightweight format model of the second prediction model obtained by the server based on the network behavior data.

[0018] According to a third aspect of the present disclosure, a cell avoidance device is provided, comprising: the device including: a processing module, the processing module being configured to: The first prediction model is used to predict the probability of anomalies in the cell associated with the first terminal based on the current spatiotemporal information of the first terminal. The first prediction model is trained on the server based on network behavior data collected by the second terminal. Perform corresponding avoidance operations on abnormal cells whose anomaly probability is greater than the probability threshold.

[0019] According to a fourth aspect of the present disclosure, a cell avoidance device is provided, comprising: the device including: a processing module, the processing module being configured to: A first prediction model is obtained by training based on network behavior data collected by the second terminal. The first prediction model is used to enable the first terminal to predict the anomaly probability of the cell associated with the first terminal based on the first spatiotemporal information of the first terminal, and to perform corresponding avoidance operations on the abnormal cells whose anomaly probability is greater than the probability threshold.

[0020] According to a fifth aspect of the present disclosure, an electronic device is provided, the electronic device comprising: One or more processors; The processor is used to invoke instructions to cause the electronic device to execute the cell avoidance method described in the first or second aspect.

[0021] According to a sixth aspect of the present disclosure, a storage medium is provided that stores instructions that, when executed on an electronic device, cause the electronic device to perform the cell avoidance method described in the first or second aspect.

[0022] This disclosure provides a cell avoidance method, system, electronic device, and storage medium. The cell avoidance method includes: employing a first prediction model to predict the probability of anomalies in cells associated with the first terminal based on current spatiotemporal information of the first terminal, wherein the first prediction model is trained on a server using network behavior data collected from a second terminal; and performing corresponding avoidance operations on abnormal cells with an anomaly probability greater than a probability threshold. Thus, by deploying a prediction model on the first terminal and training it using network behavior data collected from multiple terminals, early identification and proactive avoidance of abnormal cells are achieved. This reduces user communication interruptions and experience degradation caused by passive response mechanisms after cell anomalies occur, thereby reducing terminal communication anomalies and improving the overall user experience in cellular networks. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating a cell avoidance method according to an exemplary embodiment. Figure 1 ; Figure 2 This is a flowchart illustrating a cell avoidance method according to an exemplary embodiment. Figure 2 ; Figure 3 This is a schematic diagram of an active avoidance system for abnormal cells, according to an exemplary embodiment. Figure 4 This is a flowchart illustrating a cell avoidance method according to an exemplary embodiment. Figure 3 ; Figure 5 This is a schematic diagram of a prediction model structure according to an exemplary embodiment; Figure 6 This is a schematic diagram of the structure of an electronic device according to an exemplary embodiment. Detailed Implementation

[0024] To make the technical solution and beneficial effects of the present invention more apparent and understandable, a detailed description is provided below by listing specific embodiments. The accompanying drawings are not necessarily drawn to scale, and local features may be enlarged or reduced to more clearly show the details of the local features; unless otherwise defined, the technical and scientific terms used herein have the same meanings as those in the technical field to which this application pertains.

[0025] This disclosure is not exhaustive, but merely illustrative of some embodiments, and is not intended to limit the scope of protection of this disclosure. Unless otherwise specified, each step in a particular embodiment can be implemented as an independent embodiment, and the steps can be arbitrarily combined. For example, a solution after removing some steps in a particular embodiment can also be implemented as an independent embodiment, and the order of the steps in a particular embodiment can be arbitrarily interchanged. Furthermore, the optional implementation methods in a particular embodiment can be arbitrarily combined; moreover, the embodiments can be arbitrarily combined, for example, some or all steps of different embodiments can be arbitrarily combined, and a particular embodiment can be arbitrarily combined with the optional implementation methods of other embodiments.

[0026] In each of the disclosed embodiments, unless otherwise specified or in case of logical conflict, the terminology and / or descriptions of the embodiments are consistent and can be referenced by each other. Technical features in different embodiments can be combined to form new embodiments based on their inherent logical relationships.

[0027] The terminology used in the embodiments of this disclosure is for the purpose of describing particular embodiments only and is not intended to limit the scope of this disclosure.

[0028] In this disclosure, unless otherwise stated, elements expressed in the singular form, such as "a," "an," "the," "the," "the," "the," "the," "the," "this," etc., can mean "one and only one," or "one or more," "at least one," etc. For example, when using articles such as "a," "an," "the," etc. in translation, the noun following the article can be understood as either a singular or a plural expression.

[0029] In the embodiments disclosed herein, "multiple" refers to two or more.

[0030] In some embodiments, the terms “at least one of”, “one or more”, “a plurality of”, “multiple”, etc., may be used interchangeably.

[0031] In some embodiments, the notation "at least one of A and B", "A and / or B", "A in one case, B in another", "A in one case, B in another", etc., may include the following technical solutions depending on the situation: in some embodiments, A (A is executed regardless of B); in some embodiments, B (B is executed regardless of A); in some embodiments, execution is selected from A and B (A and B are selectively executed); in some embodiments, both A and B are executed. The same applies when there are more branches such as A, B, C, etc.

[0032] In some embodiments, the notation "A or B" may include the following technical solutions, depending on the situation: in some embodiments, A (execution of A regardless of B); in some embodiments, B (execution of B regardless of A); in some embodiments, selective execution from A and B (A and B are selectively executed). The same applies when there are more branches such as A, B, and C.

[0033] The prefixes "first," "second," etc., used in the embodiments of this disclosure are merely for distinguishing different descriptive objects and do not impose restrictions on the position, order, priority, value, or content of the descriptive objects. The description of the descriptive objects should be found in the claims or the context of the embodiments, and the use of prefixes should not constitute unnecessary restrictions. For example, if the descriptive object is a "field," the ordinal numbers preceding "field" in "first field" and "second field" do not restrict the position or order of the "fields." "First" and "second" do not restrict whether the "fields" they modify are in the same message, nor do they restrict the order of "first field" and "second field." Similarly, if the descriptive object is a "level," the ordinal numbers preceding "level" in "first level" and "second level" do not restrict the priority between "levels." Furthermore, the value of the descriptive object is not limited by ordinal numbers and can be one or more. For example, in "first device," the value of "device" can be one or more. Furthermore, the objects modified by different prefixes can be the same or different. For example, if the object being described is "device", then "first device" and "second device" can be the same device or different devices, and their types can be the same or different. Similarly, if the object being described is "information", then "first information" and "second information" can be the same information or different information, and their content can be the same or different.

[0034] In some embodiments, “including A,” “containing A,” “for indicating A,” and “carrying A” can be interpreted as directly carrying A or indirectly indicating A.

[0035] In some embodiments, terms such as “…”, “determine…”, “in the case of…”, “when…”, “when…”, “if…”, etc. can be used interchangeably.

[0036] In some embodiments, the terms “greater than,” “greater than or equal to,” “not less than,” “more than,” “more than or equal to,” “not less than,” “higher than,” “higher than or equal to,” “not lower than,” and “above” can be used interchangeably, as can the terms “less than,” “less than or equal to,” “not greater than,” “less than,” “less than or equal to,” “not more than,” “lower than,” “lower than or equal to,” “not higher than,” and “below”.

[0037] In some embodiments, devices, etc., can be interpreted as physical or virtual, and their names are not limited to the names recorded in the embodiments. Terms such as “device”, “equipment”, “circuit”, “network element”, “graph node”, “function”, “unit”, “section”, “system”, “network”, “chip”, “chip system”, “entity”, and “subject” can be used interchangeably.

[0038] Furthermore, each element, each row, or each column in the table of this disclosure can be implemented as an independent embodiment, and any combination of any element, any row, or any column can also be implemented as an independent embodiment.

[0039] When a terminal fails repeatedly during a specific process, the 3GPP protocol defines limited recovery mechanisms. For example: When AS layer processes such as RRC establishment, handover, redirection, and reselection fail consecutively, the terminal may add the problematic cell to the block list (e.g., via the T324 timer) or block the entire TAC area.

[0040] When NAS layer processes (such as Attach / TAU / MRU / Service Request / PDN establishment / PDU session establishment, etc.) fail consecutively, the terminal may attempt PLMN reselection or fall back to other network standards.

[0041] These mechanisms are typically triggered based on a fixed threshold (such as the number of consecutive failures) and work independently for a specific process.

[0042] However, there is a lack of effective handling mechanisms for certain types of anomalies, such as: Anomaly in cell congestion detection: The protocol does not define a standard avoidance mechanism, which relies on the terminal manufacturer for implementation and may result in no response.

[0043] Paging resolution failure: This usually does not trigger cell evasion, but may cause the terminal to remain in a state of service unavailability.

[0044] Caller / called party failure: The failure may only be recorded and may not trigger cell-level avoidance.

[0045] IMS-related anomalies: such as VoLTE / VoNR registration failure, may only trigger CSFB and will not actively avoid problematic cells.

[0046] Current terminal response measures for abnormal cells have the following drawbacks: Passive response: Attempts to escape only after an anomaly occurs, failing to prevent it in advance and inevitably leading to user experience disruption. Lack of coordination: Each protocol layer handles anomalies independently, lacking cross-layer collaboration, potentially resulting in conflicting or insufficient avoidance decisions. Inadequate mechanism: Many anomaly scenarios lack standardized avoidance solutions, relying on vendor implementations with inconsistent effectiveness. Poor adaptability: Fixed thresholds cannot adapt to dynamic network changes, easily leading to misjudgment or underjudgment of abnormal cells. Lack of predictive capability: Unable to utilize historical data to identify anomalies in advance, resulting in repeated failures. The terminal only triggers the avoidance mechanism after failing to interact with an abnormal cell, inevitably degrading the user experience. Furthermore, this passive response cannot address the probabilistic and spatiotemporal correlation characteristics of network anomalies, resulting in limited avoidance effectiveness.

[0047] This disclosure proposes a cell avoidance method, applied to a first terminal, such as... Figure 1 As shown, the method includes: Step 101: Using a first prediction model, based on the first spatiotemporal information of the first terminal, predict the probability of anomalies in the cell associated with the first terminal. The first prediction model is trained on the server based on network behavior data collected by the second terminal. Step 102: Perform corresponding avoidance operations for abnormal cells whose abnormal probability is greater than the probability threshold.

[0048] Here, the cell avoidance method can be executed by a first terminal, such as a smartphone, tablet, or other device with cellular network communication capabilities. The first terminal can collect its own information and execute avoidance decisions.

[0049] The first prediction model can be a computational model used to assess the risk of anomalies in a cell. This first prediction model can be deployed on a first terminal to output a quantified probability value of anomalies based on the input data.

[0050] The first spatiotemporal information may include at least one of the following: the current time information of the first terminal; the current geographical location information of the first terminal. The geographical location information may include at least one of the following: GPS coordinates, base station cell identifier, location area code, routing area code, and tracking area code.

[0051] Anomaly probability can include the likelihood of a network anomaly occurring in the cell. For example, the anomaly probability can range from 0 to 1, with a higher value indicating a greater risk of anomaly.

[0052] The probability threshold can be a preset critical value. When the predicted probability of an anomaly exceeds this threshold, the cell is considered to have an anomaly risk, and corresponding avoidance operations need to be triggered.

[0053] Avoidance maneuvers can be actions taken by a terminal to reduce or mitigate interactions with abnormal cells. The aim of avoidance maneuvers is to reduce the chances of a terminal staying on or connecting to abnormal cells, thereby improving the user experience.

[0054] The server may include one or more server entities that provide computing, storage, and management services. The server may be used to train the first prediction model.

[0055] The second terminal may include a mobile communication terminal device that collects network behavior data. The data from the second terminal is aggregated at the server for model training. The second terminal may be the same device as the first terminal, or it may be a different device. The second terminal may be the same type of device as the first terminal, or it may be a different type of device.

[0056] Network behavior data can be various types of data generated by a second terminal during network communication. This data reflects the interaction between the terminal and the network, such as the success or failure of connection establishment, data transmission rate, signal strength, cell handover records, etc.

[0057] The cell associated with the first terminal may include at least one of the following: the cell where the first terminal is currently located, the target cell that the first terminal may connect to (access, camp, etc.), the cell within a predetermined distance range of the first terminal, the cell within the signal range of the first terminal, and the cell that the first terminal may enter based on the first terminal's movement. The first prediction model is trained on the server side using network behavior data collected from the second terminal. The server can collect network behavior data from one or more second terminals. This data can include connection status, service request results, signal quality, and other information when the terminal communicates with different cells at different times, locations, and locations. For example, the server can collect metrics such as success rate, drop rate, and latency when the second terminal makes voice calls or transmits data in a specific cell. The server can use this data for statistical analysis to identify patterns or features related to cell anomalies and construct the first prediction model accordingly. After training, the trained first prediction model is distributed to the first terminal.

[0058] The first terminal may periodically or upon triggering a specific event (e.g., before attempting to access a cell, during cell reselection or handover decisions). The first spatiotemporal information may include the terminal's geographic location information (e.g., latitude and longitude coordinates obtained via GPS, cell identifier obtained via base station cell identifier), time information (e.g., current date and timestamp), and other environmental information (e.g., the terminal's moving speed, current network type). The first predictive model may be a statistical analysis-based model, for example, predicting the probability of anomalies in associated cells by analyzing the failure rate of a specific spatiotemporal region in historical data.

[0059] After obtaining the predicted anomaly probability, the first terminal compares it with a preset probability threshold. If the predicted anomaly probability exceeds the threshold, the associated cell is considered an abnormal cell. At this point, the first terminal can take a series of avoidance measures. For example, the first terminal can simply record the identifier and anomaly probability of the abnormal cell and issue a prompt to the user; or, the first terminal can temporarily add the abnormal cell to a temporary blacklist and reduce its activity on that cell for a fixed period of time.

[0060] In one possible implementation, the first terminal has at least one associated cell, and there may be at least one predicted abnormal cell.

[0061] Thus, by deploying a predictive model on the first terminal and training it using network behavior data collected from multiple terminals, early identification and proactive avoidance of abnormal cells are achieved. This reduces user communication interruptions and experience degradation caused by passive response mechanisms after cell anomalies occur, thereby minimizing terminal communication anomalies and improving the overall user experience in cellular networks.

[0062] In some embodiments, performing the corresponding avoidance operation on abnormal cells with an anomaly probability greater than a probability threshold includes at least one of the following: In response to the anomaly probability being greater than or equal to a first probability threshold and less than or equal to a second probability threshold, the measured value of the anomalous cell is negatively biased based on the anomaly probability, wherein the second probability threshold is greater than the first probability threshold. In response to the anomaly probability being greater than the second probability threshold, the abnormal cell will not be camped for at least the prohibition period, wherein the prohibition period is determined based on the anomaly probability.

[0063] The probability of an anomaly refers to the likelihood of a network anomaly occurring in the cell associated with the first terminal. Its value is usually between 0 and 1, with a higher value indicating a higher probability of an anomaly.

[0064] The first and second probability thresholds are preset values ​​used to divide the anomaly probability intervals, with the second probability threshold being greater than the first probability threshold. These thresholds can be set according to the actual network environment, operator policies, or user experience requirements. For example, they can be set to fixed values, such as 0.7 and 0.9; or they can be dynamically adjusted based on factors such as historical network data, terminal type, or geographical location to adapt to different application scenarios.

[0065] The measured values ​​may include data obtained by the first terminal from measuring the wireless signal of abnormal cells. For example, the measured values ​​may include at least one of the following: Reference Signal Received Power (RSRP), Reference Signal Received Quality (RSRQ), Received Signal Strength Indication (RSSI), etc. These measured values ​​reflect the network signal quality and strength perceived by the terminal and are important bases for the terminal to select and reselect cells.

[0066] Negative bias refers to reducing the measurement values ​​of abnormal cells during cell selection or reselection by the terminal. By applying a negative bias to the measurement values, the priority of abnormal cells in the cell selection algorithm can be reduced, weakening their attractiveness to the terminal and thus prompting the terminal to prioritize other non-abnormal cells. The implementation of negative bias can include, but is not limited to: directly subtracting a preset fixed value from the original measurement value; or dynamically calculating a bias amount based on the probability of abnormality and performing the subtraction operation; or multiplying by a coefficient less than 1 to reduce the effective strength of the measurement value.

[0067] The prohibition period refers to the length of time under specific conditions during which a first terminal is prohibited from camping on an abnormal cell. During this prohibition period, even if the signal conditions of the abnormal cell meet the camping requirements, the terminal will not choose to camp on that cell. The prohibition period can be implemented in ways including, but not limited to: setting a fixed time length, such as 5 minutes or 10 minutes; or dynamically calculating the prohibition period based on the probability of anomalies using a function or lookup table, for example, a higher probability of anomalies results in a longer prohibition period.

[0068] Through the above technical solution, this application can perform differentiated avoidance operations on abnormal cells based on different levels of anomaly probability. When the anomaly probability is greater than or equal to a first probability threshold and less than or equal to a second probability threshold, the system applies a negative bias to the measured value of the abnormal cell based on the anomaly probability. This effectively reduces the priority of abnormal cells in terminal cell selection, weakening their attractiveness and guiding terminals to actively avoid potentially problematic cells, while not completely prohibiting them from camping, preserving the possibility of connection when necessary, and reducing service interruptions that may result from excessive avoidance.

[0069] When the anomaly probability is high, exceeding the second probability threshold, the system will not reside in the abnormal cell for at least the specified prohibition period. This reduces repeated connection attempts by the terminal on high-risk abnormal cells, significantly lowering the risk of user experience interruption. Since the prohibition period is determined based on the anomaly probability, the avoidance strategy can dynamically adapt to the severity of the anomaly; the higher the anomaly probability, the longer the prohibition period, thus improving the flexibility and effectiveness of the avoidance strategy.

[0070] In this way, by adopting a tiered and refined avoidance mechanism, the first terminal can respond to abnormal cells of different degrees more intelligently and effectively, significantly improving the user experience and reducing communication interruptions caused by network anomalies.

[0071] In some embodiments, the bias value for negatively biasing the measured value includes: the product of a first probability weight and a first preset value, plus the sum of a second preset value, wherein the first probability weight is positively correlated with the anomaly probability.

[0072] The first probability weight can be a weighting factor used to adjust the contribution of a first preset value to the bias value. The higher the anomaly probability, the larger the weight, thus making the bias value more effective at avoiding anomalous cells. The first probability weight can be calculated based on the anomaly probability using a linear function, a nonlinear function (e.g., an exponential or logarithmic function), or a piecewise function, or it can be determined using a pre-trained model or a lookup table.

[0073] The first preset value is a fundamental parameter in the bias value calculation. Multiplied by the first probability weight, it constitutes the anomaly probability-related portion of the bias value. The first preset value can be a fixed value preset by the system or a dynamically configured value based on factors such as network environment, terminal type, or user policy. The second preset value is an additional parameter in the bias value calculation, providing a basic bias amount. Even when the anomaly probability is low, resulting in a small product of the first probability weight and the first preset value, the second preset value ensures that the bias value reaches a minimum negative bias, thereby reducing insufficient avoidance. The second preset value can be a fixed value or a value configured according to actual needs. The first probability weight is positively correlated with the anomaly probability; this positive correlation ensures the adaptability of the bias value. When the predicted anomaly probability is high, the first probability weight increases accordingly, resulting in a larger final negative bias value, thus more strongly suppressing the first terminal from residing in the anomalous cell. Conversely, when the anomaly probability is low, the negative bias value is relatively small, reducing over-avoidance. This relationship can be achieved through various mathematical models, such as a simple linear relationship or a more complex nonlinear increasing function.

[0074] This allows the negative bias operation to more accurately reflect the actual risk level of abnormal cells, improving the accuracy and adaptability of avoidance operations. As a result, when the first terminal predicts a cell anomaly, it can more effectively adjust its behavior, reduce interaction with abnormal cells, and ultimately improve the user's communication experience.

[0075] In some embodiments, the prohibition duration includes: the product of a second probability weight and a first duration, plus the sum of the second durations, wherein the second probability weight is positively correlated with the anomaly probability.

[0076] The second probability weight is a coefficient used to adjust the calculation of the prohibition duration. It maps the predicted probability of anomalies to the influence factor of the prohibition duration, so that a higher probability of anomalies results in a larger weight, leading to a longer prohibition duration. The second probability weight can be a linear function of the anomaly probability; it can also be a non-linear function of the anomaly probability, such as a logarithmic function, exponential function, or piecewise function, to more precisely reflect the sensitivity to different anomaly probability intervals, thereby achieving more accurate control over the prohibition duration.

[0077] The first duration can be a preset base time unit, whose main function is to serve as a scaling factor in the calculation of the prohibition duration, determining the baseline strength of the impact of the anomaly probability on the prohibition duration. The first duration can be a fixed value preset by the system, such as a value in seconds, minutes, or hours; it can also be a parameter dynamically configured according to the network operator's policy, terminal type, or user preferences; or it can be an empirical value obtained by analyzing historical network behavior data or through simulation optimization.

[0078] The second duration can be a preset fixed time value, serving as a basic guarantee for the prohibition duration. Its function is to ensure that even in situations with a low probability of anomalies, a minimum prohibition duration is provided, preventing the prohibition time from being too short to effectively avoid anomalies, while also reducing the risk of the prohibition time being zero, which would lead to evasion failure. The second duration can be a system-preset fixed value, such as 1 minute or 2 minutes; it can also be a parameter set according to the network operator's policy or the terminal manufacturer's recommended value; or it can be a configurable parameter that allows adjustment according to actual needs to achieve a balance between evasion effectiveness and user experience.

[0079] The second probability weight is positively correlated with the anomaly probability, meaning that as the predicted anomaly probability increases, the second probability weight also increases. The prohibition duration can be dynamically adjusted according to the severity of the anomaly (reflected by the anomaly probability), thereby achieving a more refined avoidance strategy.

[0080] In some embodiments, the network behavior data includes: The identification information of the second terminal; The second spatiotemporal information corresponding to the second terminal; The cell identifier corresponding to the network behavior of the second terminal; The network type in which the second terminal performs network behavior; The result of the second terminal performing network actions at the corresponding network layer.

[0081] The identification information of the second terminal refers to the code or string used to uniquely identify the second terminal. For example, the identification information could be the International Mobile Equipment Identity (IMEI), a globally unique 15-digit number string that accurately identifies the mobile device. Alternatively, the identification information could be the International Mobile Subscriber Identity (IMSI), which is typically stored in the SIM card and used to identify the mobile user. Or, the identification information could be the device's MAC address or a device serial number assigned by the manufacturer. By recording the identification information of the second terminal, the behavioral patterns of different terminals can be distinguished, reducing data confusion and enabling the predictive model to learn the impact of individual differences on anomalies.

[0082] The second spatiotemporal information corresponding to the second terminal refers to the time information and / or geographical location information that records the network behavior of the second terminal. For example, the second spatiotemporal information may include Global Positioning System (GPS) coordinates (such as longitude, latitude, and altitude) and the corresponding timestamp. Alternatively, the second spatiotemporal information may be obtained based on base station positioning technology, including information such as cell ID, location area code (LAC) or tracking area code (TAC), combined with the timestamp.

[0083] The cell identifier corresponding to the network behavior of the second terminal is a unique identifier of the cellular cell to which the second terminal is connected when performing network behavior. For example, the cell identifier can be a Global Cell ID. The Global Cell ID can be composed of a Public Land Mobile Network ID (PLMN ID) and a Cell Identity. Alternatively, the cell identifier can be a Local Cell ID, which is unique within a specific base station or area. Or, the cell identifier can be a Physical Cell Identifier (PCI), used to distinguish cells operating on the same frequency. By directly associating behavioral data with specific target cells, the prediction model can focus on cell-level anomaly features, improving the targeting of detection.

[0084] The network standard type used by the second terminal to perform network behavior refers to the communication technology standard or mode adopted when the second terminal performs network behavior. For example, the network standard type may include 4G (such as LTE) or 5G (such as NR). Alternatively, the network standard type can be further subdivided into Frequency Division Duplex (FDD) mode or Time Division Duplex (TDD) mode. Or, for 5G networks, it can be distinguished between Standalone (SA) mode or Non-Standalone (NSA) mode. By considering the environmental differences of different standards (such as 4G or 5G), the prediction model can adapt to abnormal patterns under various network conditions, thereby enhancing the model's generalization ability.

[0085] The execution result of the second terminal's network behavior at the corresponding network layer refers to recording the final state of the second terminal performing specific operations at different network protocol layers. The network layer may include at least one of the following: AS layer (Access Stratum), NAS layer (Non-Access Stratum), and IMS layer (IP Multimedia Subsystem). For example, the execution result of the AS layer may include the success or failure status of Radio Resource Control (RRC) connection establishment, or the success or failure status of handover or reselection operations. The execution result of the NAS layer may include the success or failure status of attach or Tracking Area Update (TAU) procedures, or the success or failure status of service request or PDU session establishment. For example, the execution result of the IMS layer may include the success or failure status of IMS registration, or the success or failure status of VoLTE / VoNR call establishment. The execution result can be represented as success / failure or a specific error code to indicate the final state of the operation. By providing cross-layer information, the predictive model can comprehensively learn the abnormal behavior of each network layer, achieving a comprehensive prediction of the probability of anomalies.

[0086] For example, network behavior data is presented in the form of Table (1).

[0087]

[0088] In this way, the server can train the first prediction model based on more comprehensive and accurate network behavior data, which enables the first terminal to more accurately predict the anomaly probability of its associated cell based on the model. This allows the terminal to perform more precise and timely avoidance operations on abnormal cells with an anomaly probability greater than the probability threshold, significantly improving the user experience and effectively reducing the occurrence of communication interruptions.

[0089] In some embodiments, the first prediction model includes a lightweight format model of the second prediction model obtained by the server based on the network behavior data.

[0090] The first prediction model is deployed on the first terminal and is used to predict the probability of anomalies in the cells associated with the first terminal based on the current spatiotemporal information of the first terminal. Considering the computing power and storage limitations of the terminal device, this model needs to have the characteristics of efficient operation. The second prediction model is the original and complete prediction model trained by the server based on the network behavior data collected by the second terminal. This model usually has high complexity and can make full use of the server's powerful computing resources and massive network behavior data for deep learning and optimization to ensure the accuracy and robustness of the prediction. The lightweight format model refers to the optimization and simplification of the second prediction model trained by the server through model compression, pruning, quantization, knowledge distillation and other techniques, so as to significantly reduce the model size, computational complexity and memory consumption while maintaining high prediction performance. For example, model pruning can be used to remove unimportant connections or neurons in the model, thereby reducing the model size; or model quantization can be used to convert model parameters from floating-point numbers to low-bit integers to reduce storage space and accelerate computation. Another approach is to use knowledge distillation, using a large teacher model to guide the training of a small student model, so that the student model can achieve near-teacher model performance on a smaller scale.

[0091] In this way, the server leverages its powerful computing capabilities and abundant network behavior data to train a high-precision, highly robust second prediction model, ensuring the reliability of anomaly probability prediction. By converting this second prediction model into a lightweight format and deploying it as the first prediction model on the first terminal, the computational resources and memory consumption of the model on the terminal are significantly reduced. This enables real-time and efficient anomaly probability prediction, reducing prediction latency and terminal performance degradation caused by model complexity.

[0092] In one possible implementation, the first prediction model includes an MLP model. The MLP model, as a multilayer perceptron, is capable of processing high-dimensional feature data and learning complex nonlinear patterns. The input layer of the MLP receives feature inputs such as the current spatiotemporal information of the first terminal. For example, it can encode or preprocess raw features such as timestamps, geographic coordinates, and cell identifiers before input, or directly embed these features before input. The multiple hidden layers progressively extract deep feature representations through a hierarchical structure. For example, they can consist of fully connected layers with different numbers of neurons, or include structures such as skip connections to enhance information flow. The hidden layers use nonlinear activation functions, such as ReLU, Leaky ReLU, or ELU, to introduce nonlinear transformations, thereby capturing complex dependencies between features and reducing the limitations of linear models. The output layer generates anomaly probabilities using the sigmoid function, mapping the model's output to a probability range of 0 to 1, facilitating subsequent decision-making. The MLP model learns the mapping relationship between features and anomalies by optimizing a loss function; for example, a binary cross-entropy loss function can be used to quantify the prediction error. To minimize the loss function, the model iteratively updates the weights using gradient descent algorithms. For example, stochastic gradient descent (SGD), mini-batch gradient descent, or adaptive optimizers such as Adam can be used to adjust the model parameters. Furthermore, regularization techniques such as L1 regularization, L2 regularization, or Dropout are employed during training to constrain model complexity, prevent overfitting, and enhance the model's generalization ability on unknown data.

[0093] This disclosure proposes a cell avoidance method, applied to a server, such as... Figure 2 As shown, the method includes: Step 201: Train a first prediction model based on the network behavior data collected by the second terminal. The first prediction model is used to enable the first terminal to predict the abnormal probability of the cell associated with the first terminal based on the first spatiotemporal information of the first terminal, and to perform corresponding avoidance operations on abnormal cells with abnormal probabilities greater than the probability threshold.

[0094] Here, the cell avoidance method can be executed by a first terminal, such as a smartphone, tablet, or other device with cellular network communication capabilities. The first terminal can collect its own information and execute avoidance decisions.

[0095] The first prediction model can be a computational model used to assess the risk of anomalies in a cell. This first prediction model can be deployed on a first terminal to output a quantified probability value of anomalies based on the input data.

[0096] The first spatiotemporal information may include at least one of the following: the current time information of the first terminal; the current geographical location information of the first terminal. The geographical location information may include at least one of the following: GPS coordinates, base station cell identifier, location area code, routing area code, and tracking area code.

[0097] Anomaly probability can include the likelihood of a network anomaly occurring in the cell. For example, the anomaly probability can range from 0 to 1, with a higher value indicating a greater risk of anomaly.

[0098] The probability threshold can be a preset critical value. When the predicted probability of an anomaly exceeds this threshold, the cell is considered to have an anomaly risk, and corresponding avoidance operations need to be triggered.

[0099] Avoidance maneuvers can be actions taken by a terminal to reduce or mitigate interactions with abnormal cells. The aim of avoidance maneuvers is to reduce the chances of a terminal staying on or connecting to abnormal cells, thereby improving the user experience.

[0100] The server may include one or more server entities that provide computing, storage, and management services. The server may be used to train the first prediction model.

[0101] The second terminal may include a mobile communication terminal device that collects network behavior data. The data from the second terminal is aggregated at the server for model training. The second terminal may be the same device as the first terminal, or it may be a different device. The second terminal may be the same type of device as the first terminal, or it may be a different type of device.

[0102] Network behavior data can be various types of data generated by a second terminal during network communication. This data reflects the interaction between the terminal and the network, such as the success or failure of connection establishment, data transmission rate, signal strength, cell handover records, etc.

[0103] The cell associated with the first terminal may include at least one of the following: the cell where the first terminal is currently located, the target cell that the first terminal may connect to (access, camp, etc.), the cell within a predetermined distance range of the first terminal, the cell of the first terminal with signal range, and the cell that the first terminal may enter based on the first terminal's movement.

[0104] The first prediction model is trained on the server side using network behavior data collected from the second terminal. The server can collect network behavior data from one or more second terminals. This data can include connection status, service request results, signal quality, and other information when the terminal communicates with different cells at different times, locations, and locations. For example, the server can collect metrics such as success rate, drop rate, and latency when the second terminal makes voice calls or transmits data in a specific cell. The server can use this data for statistical analysis to identify patterns or features related to cell anomalies and construct the first prediction model accordingly. After training, the trained first prediction model is distributed to the first terminal.

[0105] The first terminal may periodically or upon triggering a specific event (e.g., before attempting to access a cell, during cell reselection or handover decisions). The first spatiotemporal information may include the terminal's geographic location information (e.g., latitude and longitude coordinates obtained via GPS, cell identifier obtained via base station cell identifier), time information (e.g., current date and timestamp), and other environmental information (e.g., the terminal's moving speed, current network type). The first predictive model may be a statistical analysis-based model, for example, predicting the probability of anomalies in associated cells by analyzing the failure rate of a specific spatiotemporal region in historical data.

[0106] After obtaining the predicted anomaly probability, the first terminal compares it with a preset probability threshold. If the predicted anomaly probability exceeds the threshold, the associated cell is considered an abnormal cell. At this point, the first terminal can take a series of avoidance measures. For example, the first terminal can simply record the identifier and anomaly probability of the abnormal cell and issue a prompt to the user; or, the first terminal can temporarily add the abnormal cell to a temporary blacklist and reduce its activity on that cell for a fixed period of time.

[0107] Thus, by deploying a predictive model on the first terminal and training it using network behavior data collected from multiple terminals, early identification and proactive avoidance of abnormal cells are achieved. This reduces user communication interruptions and experience degradation caused by passive response mechanisms after cell anomalies occur, thereby minimizing terminal communication anomalies and improving the overall user experience in cellular networks.

[0108] In some embodiments, the first terminal performs a corresponding avoidance operation for abnormal cells with an abnormal probability greater than a probability threshold, including at least one of the following: The first terminal responds to the fact that the anomaly probability is greater than or equal to a first probability threshold and less than or equal to a second probability threshold by negatively biasing the measured value of the abnormal cell based on the anomaly probability, wherein the second probability threshold is greater than the first probability threshold; The first terminal responds to the anomaly probability being greater than the second probability threshold by not camping on the abnormal cell for at least the prohibition period, wherein the prohibition period is determined based on the anomaly probability.

[0109] The probability of an anomaly refers to the likelihood of a network anomaly occurring in the cell associated with the first terminal. Its value is usually between 0 and 1, with a higher value indicating a higher probability of an anomaly.

[0110] The first and second probability thresholds are preset values ​​used to divide the anomaly probability intervals, with the second probability threshold being greater than the first probability threshold. These thresholds can be set according to the actual network environment, operator policies, or user experience requirements. For example, they can be set to fixed values, such as 0.7 and 0.9; or they can be dynamically adjusted based on factors such as historical network data, terminal type, or geographical location to adapt to different application scenarios.

[0111] The measured values ​​may include data obtained by the first terminal from measuring the wireless signal of abnormal cells. For example, the measured values ​​may include at least one of the following: Reference Signal Received Power (RSRP), Reference Signal Received Quality (RSRQ), Received Signal Strength Indication (RSSI), etc. These measured values ​​reflect the network signal quality and strength perceived by the terminal and are important bases for the terminal to select and reselect cells.

[0112] Negative bias refers to reducing the measurement values ​​of abnormal cells during cell selection or reselection by the terminal. By applying a negative bias to the measurement values, the priority of abnormal cells in the cell selection algorithm can be reduced, weakening their attractiveness to the terminal and thus prompting the terminal to prioritize other non-abnormal cells. The implementation of negative bias can include, but is not limited to: directly subtracting a preset fixed value from the original measurement value; or dynamically calculating a bias amount based on the probability of abnormality and performing the subtraction operation; or multiplying by a coefficient less than 1 to reduce the effective strength of the measurement value.

[0113] The prohibition period refers to the length of time under specific conditions during which a first terminal is prohibited from camping on an abnormal cell. During this prohibition period, even if the signal conditions of the abnormal cell meet the camping requirements, the terminal will not choose to camp on that cell. The prohibition period can be implemented in ways including, but not limited to: setting a fixed time length, such as 5 minutes or 10 minutes; or dynamically calculating the prohibition period based on the probability of anomalies using a function or lookup table, for example, a higher probability of anomalies results in a longer prohibition period.

[0114] Through the above technical solution, this application can perform differentiated avoidance operations on abnormal cells based on different levels of anomaly probability. When the anomaly probability is greater than or equal to a first probability threshold and less than or equal to a second probability threshold, the system applies a negative bias to the measured value of the abnormal cell based on the anomaly probability. This effectively reduces the priority of abnormal cells in terminal cell selection, weakening their attractiveness and guiding terminals to actively avoid potentially problematic cells, while not completely prohibiting them from camping, preserving the possibility of connection when necessary, and reducing service interruptions that may result from excessive avoidance.

[0115] When the anomaly probability is high, exceeding the second probability threshold, the system will not reside in the abnormal cell for at least the specified prohibition period. This reduces repeated connection attempts by the terminal on high-risk abnormal cells, significantly lowering the risk of user experience interruption. Since the prohibition period is determined based on the anomaly probability, the avoidance strategy can dynamically adapt to the severity of the anomaly; the higher the anomaly probability, the longer the prohibition period, thus improving the flexibility and effectiveness of the avoidance strategy.

[0116] In this way, by adopting a tiered and refined avoidance mechanism, the first terminal can respond to abnormal cells of different degrees more intelligently and effectively, significantly improving the user experience and reducing communication interruptions caused by network anomalies.

[0117] In some embodiments, the bias value for negatively biasing the measured value includes: the product of a first probability weight and a first preset value, plus the sum of a second preset value, wherein the first probability weight is positively correlated with the anomaly probability.

[0118] The first probability weight can be a weighting factor used to adjust the contribution of a first preset value to the bias value. The higher the anomaly probability, the larger the weight, thus making the bias value more effective at avoiding anomalous cells. The first probability weight can be calculated based on the anomaly probability using a linear function, a nonlinear function (e.g., an exponential or logarithmic function), or a piecewise function, or it can be determined using a pre-trained model or a lookup table.

[0119] The first preset value is a fundamental parameter in the bias value calculation. Multiplied by the first probability weight, it constitutes the anomaly probability-related portion of the bias value. The first preset value can be a fixed value preset by the system or a dynamically configured value based on factors such as network environment, terminal type, or user policy. The second preset value is an additional parameter in the bias value calculation, providing a basic bias amount. Even when the anomaly probability is low, resulting in a small product of the first probability weight and the first preset value, the second preset value ensures that the bias value reaches a minimum negative bias, thereby reducing insufficient avoidance. The second preset value can be a fixed value or a value configured according to actual needs. The first probability weight is positively correlated with the anomaly probability; this positive correlation ensures the adaptability of the bias value. When the predicted anomaly probability is high, the first probability weight increases accordingly, resulting in a larger final negative bias value, thus more strongly suppressing the first terminal from residing in the anomalous cell. Conversely, when the anomaly probability is low, the negative bias value is relatively small, reducing over-avoidance. This relationship can be achieved through various mathematical models, such as a simple linear relationship or a more complex nonlinear increasing function.

[0120] This allows the negative bias operation to more accurately reflect the actual risk level of abnormal cells, improving the accuracy and adaptability of avoidance operations. As a result, when the first terminal predicts a cell anomaly, it can more effectively adjust its behavior, reduce interaction with abnormal cells, and ultimately improve the user's communication experience.

[0121] In some embodiments, the prohibition duration includes: the product of a second probability weight and a first duration, plus the sum of the second durations, wherein the second probability weight is positively correlated with the anomaly probability.

[0122] The second probability weight is a coefficient used to adjust the calculation of the prohibition duration. It maps the predicted probability of anomalies to the influence factor of the prohibition duration, so that a higher probability of anomalies results in a larger weight, leading to a longer prohibition duration. The second probability weight can be a linear function of the anomaly probability; it can also be a non-linear function of the anomaly probability, such as a logarithmic function, exponential function, or piecewise function, to more precisely reflect the sensitivity to different anomaly probability intervals, thereby achieving more accurate control over the prohibition duration.

[0123] The first duration can be a preset base time unit, whose main function is to serve as a scaling factor in the calculation of the prohibition duration, determining the baseline strength of the impact of the anomaly probability on the prohibition duration. The first duration can be a fixed value preset by the system, such as a value in seconds, minutes, or hours; it can also be a parameter dynamically configured according to the network operator's policy, terminal type, or user preferences; or it can be an empirical value obtained by analyzing historical network behavior data or through simulation optimization.

[0124] The second duration can be a preset fixed time value, serving as a basic guarantee for the prohibition duration. Its function is to ensure that even in situations with a low probability of anomalies, a minimum prohibition duration is provided, preventing the prohibition time from being too short to effectively avoid anomalies, while also reducing the risk of the prohibition time being zero, which would lead to evasion failure. The second duration can be a system-preset fixed value, such as 1 minute or 2 minutes; it can also be a parameter set according to the network operator's policy or the terminal manufacturer's recommended value; or it can be a configurable parameter that allows adjustment according to actual needs to achieve a balance between evasion effectiveness and user experience.

[0125] The second probability weight is positively correlated with the anomaly probability, meaning that as the predicted anomaly probability increases, the second probability weight also increases. The prohibition duration can be dynamically adjusted according to the severity of the anomaly (reflected by the anomaly probability), thereby achieving a more refined avoidance strategy.

[0126] In some embodiments, the network behavior data includes: The identification information of the second terminal; The second spatiotemporal information corresponding to the second terminal; The cell identifier corresponding to the network behavior of the second terminal; The network type in which the second terminal performs network behavior; The result of the second terminal performing network actions at the corresponding network layer.

[0127] The identification information of the second terminal refers to the code or string used to uniquely identify the second terminal. For example, the identification information could be the International Mobile Equipment Identity (IMEI), a globally unique 15-digit number string that accurately identifies the mobile device. Alternatively, the identification information could be the International Mobile Subscriber Identity (IMSI), which is typically stored in the SIM card and used to identify the mobile user. Or, the identification information could be the device's MAC address or a device serial number assigned by the manufacturer. By recording the identification information of the second terminal, the behavioral patterns of different terminals can be distinguished, reducing data confusion and enabling the predictive model to learn the impact of individual differences on anomalies.

[0128] The second spatiotemporal information corresponding to the second terminal refers to the time information and / or geographical location information that records the network behavior of the second terminal. For example, the second spatiotemporal information may include Global Positioning System (GPS) coordinates (such as longitude, latitude, and altitude) and the corresponding timestamp. Alternatively, the second spatiotemporal information may be obtained based on base station positioning technology, including information such as cell ID, location area code (LAC) or tracking area code (TAC), combined with the timestamp.

[0129] The cell identifier corresponding to the network behavior of the second terminal is a unique identifier of the cellular cell to which the second terminal is connected when performing network behavior. For example, the cell identifier can be a Global Cell ID. The Global Cell ID can be composed of a Public Land Mobile Network ID (PLMN ID) and a Cell Identity. Alternatively, the cell identifier can be a Local Cell ID, which is unique within a specific base station or area. Or, the cell identifier can be a Physical Cell Identifier (PCI), used to distinguish cells operating on the same frequency. By directly associating behavioral data with specific target cells, the prediction model can focus on cell-level anomaly features, improving the targeting of detection.

[0130] The network standard type used by the second terminal to perform network behavior refers to the communication technology standard or mode adopted when the second terminal performs network behavior. For example, the network standard type may include 4G (such as LTE) or 5G (such as NR). Alternatively, the network standard type can be further subdivided into Frequency Division Duplex (FDD) mode or Time Division Duplex (TDD) mode. Or, for 5G networks, it can be distinguished between Standalone (SA) mode or Non-Standalone (NSA) mode. By considering the environmental differences of different standards (such as 4G or 5G), the prediction model can adapt to abnormal patterns under various network conditions, thereby enhancing the model's generalization ability.

[0131] The execution result of the second terminal's network behavior at the corresponding network layer refers to recording the final state of the second terminal performing specific operations at different network protocol layers. The network layer may include at least one of the following: AS layer (Access Stratum), NAS layer (Non-Access Stratum), and IMS layer (IP Multimedia Subsystem). For example, the execution result of the AS layer may include the success or failure status of Radio Resource Control (RRC) connection establishment, or the success or failure status of handover or reselection operations. The execution result of the NAS layer may include the success or failure status of attach or Tracking Area Update (TAU) procedures, or the success or failure status of service request or PDU session establishment. For example, the execution result of the IMS layer may include the success or failure status of IMS registration, or the success or failure status of VoLTE / VoNR call establishment. The execution result can be represented as success / failure or a specific error code to indicate the final state of the operation. By providing cross-layer information, the predictive model can comprehensively learn the abnormal behavior of each network layer, achieving a comprehensive prediction of the probability of anomalies.

[0132] In this way, the server can train the first prediction model based on more comprehensive and accurate network behavior data, which enables the first terminal to more accurately predict the anomaly probability of its associated cell based on the model. This allows the terminal to perform more precise and timely avoidance operations on abnormal cells with an anomaly probability greater than the probability threshold, significantly improving the user experience and effectively reducing the occurrence of communication interruptions.

[0133] In some embodiments, the first prediction model includes a lightweight format model of the second prediction model obtained by the server based on the network behavior data.

[0134] The first prediction model is deployed on the first terminal and is used to predict the probability of anomalies in the cells associated with the first terminal based on the current spatiotemporal information of the first terminal. Considering the computing power and storage limitations of the terminal device, this model needs to have the characteristics of efficient operation. The second prediction model is the original and complete prediction model trained by the server based on the network behavior data collected by the second terminal. This model usually has high complexity and can make full use of the server's powerful computing resources and massive network behavior data for deep learning and optimization to ensure the accuracy and robustness of the prediction. The lightweight format model refers to the optimization and simplification of the second prediction model trained by the server through model compression, pruning, quantization, knowledge distillation and other techniques, so as to significantly reduce the model size, computational complexity and memory consumption while maintaining high prediction performance. For example, model pruning can be used to remove unimportant connections or neurons in the model, thereby reducing the model size; or model quantization can be used to convert model parameters from floating-point numbers to low-bit integers to reduce storage space and accelerate computation. Another approach is to use knowledge distillation, using a large teacher model to guide the training of a small student model, so that the student model can achieve near-teacher model performance on a smaller scale.

[0135] In this way, the server leverages its powerful computing capabilities and abundant network behavior data to train a high-precision, highly robust second prediction model, ensuring the reliability of anomaly probability prediction. By converting this second prediction model into a lightweight format and deploying it as the first prediction model on the first terminal, the computational resources and memory consumption of the model on the terminal are significantly reduced. This enables real-time and efficient anomaly probability prediction, reducing prediction latency and terminal performance degradation caused by model complexity.

[0136] In one possible implementation, the first prediction model includes an MLP model. The MLP model, as a multilayer perceptron, is capable of processing high-dimensional feature data and learning complex nonlinear patterns. The input layer of the MLP receives feature inputs such as the current spatiotemporal information of the first terminal. For example, it can encode or preprocess raw features such as timestamps, geographic coordinates, and cell identifiers before input, or directly embed these features before input. The multiple hidden layers progressively extract deep feature representations through a hierarchical structure. For example, they can consist of fully connected layers with different numbers of neurons, or include structures such as skip connections to enhance information flow. The hidden layers use nonlinear activation functions, such as ReLU, Leaky ReLU, or ELU, to introduce nonlinear transformations, thereby capturing complex dependencies between features and reducing the limitations of linear models. The output layer generates anomaly probabilities using the sigmoid function, mapping the model's output to a probability range of 0 to 1, facilitating subsequent decision-making. The MLP model learns the mapping relationship between features and anomalies by optimizing a loss function; for example, a binary cross-entropy loss function can be used to quantify the prediction error. To minimize the loss function, the model iteratively updates the weights using gradient descent algorithms. For example, stochastic gradient descent (SGD), mini-batch gradient descent, or adaptive optimizers such as Adam can be used to adjust the model parameters. Furthermore, regularization techniques such as L1 regularization, L2 regularization, or Dropout are employed during training to constrain model complexity, prevent overfitting, and enhance the model's generalization ability on unknown data.

[0137] A specific example is provided in conjunction with the above embodiments for illustration.

[0138] In this example, AI machine learning technology is applied to the identification and avoidance of abnormal cells in cellular networks at the edge, constructing a complete edge-cloud collaborative intelligent avoidance system. Its features include: 1. Innovative application of AI machine learning models in the identification of abnormal cells at the edge.

[0139] This example introduces machine learning models into the field of edge-side cellular network anomaly detection, replacing traditional threshold judgment with a data-driven approach.

[0140] The model can learn the spatiotemporal distribution patterns of abnormal cells, the specific behaviors of terminals, and the dynamic changes in the network environment, thus realizing a paradigm shift from "passive response" to "active prediction".

[0141] Through a multi-layer neural network structure, the model can capture complex nonlinear feature relationships and identify abnormal patterns that are difficult for humans to detect.

[0142] 2. Distributed learning and inference architecture with edge-cloud collaboration.

[0143] Design a collaborative architecture that combines centralized cloud training with distributed edge inference to balance the requirements of model accuracy and real-time performance.

[0144] The cloud uses massive amounts of terminal data to train and optimize models, while the edge uses lightweight models to achieve low-latency inference, ensuring a continuous user experience.

[0145] 3. Multi-dimensional feature engineering and spatiotemporal context modeling Construct a multi-dimensional feature system that includes time period features, spatial location features, terminal historical features, and network environment features.

[0146] Periodic coding techniques are used to handle the continuity of time features, reducing information loss in traditional discretization methods.

[0147] 4. Tiered dynamic avoidance mechanism A tiered avoidance strategy based on anomaly probability values ​​enables a smooth transition from mild punishment to complete prohibition.

[0148] The avoidance intensity is dynamically adapted to the anomaly probability to reduce the decrease in network availability caused by excessive avoidance.

[0149] Continuous learning and adaptive update system The model has online learning capabilities and can automatically adjust its parameters according to changes in the network environment.

[0150] By leveraging a closed-loop data system between the edge and cloud, the model can be continuously optimized, ensuring its long-term effectiveness.

[0151] For example, this example provides an active avoidance system for abnormal cells based on an AI model, the system structure of which is as follows: Figure 3 As shown, predictive avoidance is achieved through cloud (server-side) training and terminal-side inference. Data collection, model training, and model updates are handled by the cloud. Terminal-side execution location is obtained, model inference is performed, and interaction with the modem is completed. The core of the solution includes data collection, model training, terminal-side inference, and avoidance execution.

[0152] 1. Data Collection: The terminal (second terminal) collects cellular network behavior data throughout the entire process, including terminal identification, spatiotemporal information, cell information, network type, and process results at each layer (such as success / failure records at the AS / NAS / IMS layer).

[0153] Network behavior data is uploaded to the cloud in an efficient serialization format to ensure coverage of multi-dimensional features (such as time, location, terminal type, and network status).

[0154] For example, network behavior data is presented in the form of Table (1).

[0155] 2. Model Training: Training data: After aggregating data in the cloud, feature engineering is performed, including time period encoding, terminal historical statistics, spatial gridding, etc., to generate standardized feature vectors.

[0156] Model (prediction model) selection: A multilayer perceptron (MLP) model is adopted, such as... Figure 4 As shown, the model consists of an input layer, multiple hidden layers, and an output layer. The hidden layers use non-linear activation functions (such as ReLU) to learn feature interactions, and the output layer generates anomaly probabilities using the sigmoid function. The model is trained independently for different anomaly types to focus on specific patterns.

[0157] Training principle: The model learns the mapping relationship between features and anomalies by optimizing the loss function (such as cross-entropy), and iteratively updates the weights using the gradient descent algorithm. Regularization techniques (such as Dropout) are used during training to prevent overfitting and ensure generalization ability.

[0158] Output: The prediction model outputs the probability value of each cell for different anomaly types, which is used for edge decision-making.

[0159] 3. End-to-end reasoning: Models trained in the cloud can be converted into a lightweight format and deployed to the terminal, and integrated through the system interface.

[0160] Inference is triggered based on changes in terminal location and / or time to predict the probability of anomalies in surrounding cells.

[0161] The probability value is compared with a preset threshold to generate a list of abnormal cells.

[0162] like Figure 5 As shown, the specific steps for circumvention on the terminal side include: Step 501: Obtain location and time.

[0163] Step 502: Load the model (first prediction model).

[0164] Step 503: List of communities with abnormal decommissioning status.

[0165] Step 504: Bypass via Modem.

[0166] Step 505: Update the results.

[0167] 4. Avoid execution logic: The design incorporates different levels of proactive avoidance based on anomaly probability values: Mild avoidance (suitable for anomalous cells with a moderate probability (e.g., 0.7~0.9)): During cell selection / reselection and measurement report reporting, a penalty deduction is applied to the measurement values ​​of abnormal cells (such as RSRP / RSRQ) (i.e., the measurement values ​​of abnormal cells are negatively biased).

[0168] Impairment Amount = Basic Impairment + Probability Weight × Additional Impairment Effect: Significantly reduces the probability of dwelling, but does not completely prohibit it.

[0169] Heavy avoidance (suitable for anomalous cells with a high probability of occurrence, such as greater than 0.9): Adding a residential area to the prohibited list will result in a prohibited time period equal to the base time plus a probability weight multiplied by an extension time. This cell is not included in the cell handover measurement report. Effect: Completely reduces dwell time.

[0170] The method used in this example has the following advantages.

[0171] Proactive prevention: The model predicts the probability of anomalies and avoids them before they occur, thus improving the continuity of user experience.

[0172] High accuracy: Machine learning models learn complex patterns, and the anomaly detection accuracy is significantly higher than that of threshold methods.

[0173] Adaptive optimization: The model is continuously trained and updated to adapt to network changes, and is effective in the long term.

[0174] Resource efficiency: Lightweight model design results in low resource consumption for edge inference, without affecting terminal performance.

[0175] Cross-scenario collaboration: A unified model handles multiple types of anomalies, and avoidance strategies are consistent and collaborative.

[0176] The MLP model achieves a good balance in this scenario. It has a simple structure, stable training, fast inference speed, is suitable for deployment on mobile terminals, and can effectively learn the feature interactions in this solution.

[0177] In addition to MLP, the prediction model can also use at least one of the following models: Logistic Regression: Simple to calculate and highly interpretable.

[0178] Decision trees / random forests: can handle non-linear relationships, and feature importance is clear.

[0179] Gradient Boosting Tree (XGBoost / LightGBM): Performs well on many tabular data tasks.

[0180] Convolutional Neural Networks (CNNs): can handle spatial features (such as gridding geographic locations).

[0181] Recurrent Neural Networks (RNNs) / Long Short-Term Memory Networks (LSTMs): Suitable for processing time series data, they can capture patterns of anomalies evolving over time.

[0182] Graph Neural Networks (GNNs): Model cells and locations as graph structures, making them suitable for capturing network topology relationships.

[0183] Attention mechanism / Transformer: can process sequential data and capture long-term dependencies.

[0184] CNN+MLP combination: CNN is used to process spatial features, and MLP is used to process other features.

[0185] Combination of temporal model and MLP: LSTM is used to process temporal features, and MLP is used to process static features.

[0186] This disclosure also proposes a cell evasion device, installed in a first terminal. The cell evasion device includes a processing module, which is used for: The first prediction model is used to predict the probability of anomalies in the cell associated with the first terminal based on the current spatiotemporal information of the first terminal. The first prediction model is trained on the server based on network behavior data collected by the second terminal. Perform corresponding avoidance operations on abnormal cells whose anomaly probability is greater than the probability threshold.

[0187] In some embodiments, the processing module is specifically used for at least one of the following: In response to the anomaly probability being greater than or equal to a first probability threshold and less than or equal to a second probability threshold, the measured value of the anomalous cell is negatively biased based on the anomaly probability, wherein the second probability threshold is greater than the first probability threshold. In response to the anomaly probability being greater than the second probability threshold, the abnormal cell will not be camped for at least the prohibition period, wherein the prohibition period is determined based on the anomaly probability.

[0188] In some embodiments, the bias value for negatively biasing the measured value includes: the product of a first probability weight and a first preset value, plus the sum of a second preset value, wherein the first probability weight is positively correlated with the anomaly probability.

[0189] In some embodiments, the prohibition duration includes: the product of a second probability weight and a first duration, plus the sum of the second durations, wherein the second probability weight is positively correlated with the anomaly probability.

[0190] In some embodiments, the network behavior data includes: The identification information of the second terminal; The second spatiotemporal information corresponding to the second terminal; The cell identifier corresponding to the network behavior of the second terminal; The network type in which the second terminal performs network behavior; The result of the second terminal performing network actions at the corresponding network layer.

[0191] In some embodiments, the first prediction model includes a lightweight format model of the second prediction model obtained by the server based on the network behavior data.

[0192] This disclosure also proposes a cell evasion device, configured on a server, the cell evasion device comprising: a processing module, the processing module being used for: A first prediction model is obtained by training based on network behavior data collected by the second terminal. The first prediction model is used to enable the first terminal to predict the anomaly probability of the cell associated with the first terminal based on the first spatiotemporal information of the first terminal, and to perform corresponding avoidance operations on the abnormal cells whose anomaly probability is greater than the probability threshold.

[0193] In some embodiments, the first terminal performs a corresponding avoidance operation for abnormal cells with an abnormal probability greater than a probability threshold, including at least one of the following: The first terminal responds to the fact that the anomaly probability is greater than or equal to a first probability threshold and less than or equal to a second probability threshold by negatively biasing the measured value of the abnormal cell based on the anomaly probability, wherein the second probability threshold is greater than the first probability threshold; The first terminal responds to the anomaly probability being greater than the second probability threshold by not camping on the abnormal cell for at least the prohibition period, wherein the prohibition period is determined based on the anomaly probability.

[0194] In some embodiments, the bias value for negatively biasing the measured value includes: the product of a first probability weight and a first preset value, plus the sum of a second preset value, wherein the first probability weight is positively correlated with the anomaly probability.

[0195] In some embodiments, the prohibition duration includes: the product of a second probability weight and a first duration, plus the sum of the second durations, wherein the second probability weight is positively correlated with the anomaly probability.

[0196] In some embodiments, the network behavior data includes: The identification information of the second terminal; The second spatiotemporal information corresponding to the second terminal; The cell identifier corresponding to the network behavior of the second terminal; The network type in which the second terminal performs network behavior; The result of the second terminal performing network actions at the corresponding network layer.

[0197] In some embodiments, the first prediction model includes a lightweight format model of the second prediction model obtained by the server based on the network behavior data.

[0198] It should be understood that the division of units or modules in the above device is only a logical functional division. In actual implementation, they can be fully or partially integrated into a single physical entity, or they can be physically separated. Furthermore, the units or modules in the device can be implemented by a processor calling software: for example, the device includes a processor connected to a memory containing instructions. The processor calls the instructions stored in the memory to implement any of the above methods or to implement the functions of the units or modules in the above device. The processor can be, for example, a general-purpose processor, such as a Central Processing Unit (CPU) or a microprocessor, and the memory can be internal or external to the device. Alternatively, the units or modules in the device can be implemented in the form of hardware circuits. The functionality of some or all of the units or modules can be achieved through the design of these hardware circuits, which can be understood as one or more processors. For example, in one implementation, the hardware circuit is an application-specific integrated circuit (ASIC), and the functionality of some or all of the units or modules is achieved through the design of the logical relationships between the components within the circuit. In another implementation, the hardware circuit can be implemented using a programmable logic device (PLD), such as a field-programmable gate array (FPGA), which can include a large number of logic gates. The connection relationships between the logic gates are configured through configuration files, thereby achieving the functionality of some or all of the units or modules. All units or modules of the above device can be implemented entirely through processor-called software, entirely through hardware circuits, or partially through processor-called software with the remaining parts implemented through hardware circuits.

[0199] In this disclosure, the processor is a circuit with signal processing capabilities. In one implementation, the processor can be a circuit with instruction read and execute capabilities, such as a Central Processing Unit (CPU), a microprocessor, a graphics processing unit (GPU) (which can be understood as a type of microprocessor), or a digital signal processor (DSP). In another implementation, the processor can implement certain functions through the logical relationships of hardware circuits. The logical relationships of the aforementioned hardware circuits are fixed or reconfigurable. For example, the processor is a hardware circuit implemented using an application-specific integrated circuit (ASIC) or a programmable logic device (PLD), such as an FPGA. In a reconfigurable hardware circuit, the process of the processor loading a configuration document and configuring the hardware circuit can be understood as the process of the processor loading instructions to implement the functions of some or all of the above units or modules. In addition, it can also be a hardware circuit designed for artificial intelligence, which can be understood as an ASIC, such as a Neural Network Processing Unit (NPU), a Tensor Processing Unit (TPU), a Deep Learning Processing Unit (DPU), etc.

[0200] Figure 6 This is a schematic diagram of the structure of the electronic device 9100 provided in this embodiment. The electronic device 9100 can be a computer terminal, a server, a chip, chip system, or processor that supports the implementation of any of the above methods, or a chip, chip system, or processor that supports the implementation of any of the above information transmission methods in the terminal. The electronic device 9100 can be used to implement the document processing methods described in the above method embodiments; for details, please refer to the descriptions in the above method embodiments.

[0201] like Figure 6 As shown, the electronic device 9100 includes one or more processors 9101. The processor 9101 can be a general-purpose processor or a special-purpose processor, etc. The processor 9101 is used to invoke instructions to cause the electronic device 9100 to execute any of the above-mentioned document processing methods.

[0202] In some embodiments, the electronic device 9100 further includes one or more memories 9102 for storing instructions. Optionally, all or part of the memories 9102 may also be located outside the electronic device 9100.

[0203] In some embodiments, the electronic device 9100 further includes one or more transceivers 9103. When the electronic device 9100 includes one or more transceivers 9103, the steps of sending, receiving and / or acquiring in the above method are performed by the transceivers 9103, and the other steps are performed by the processor 9101.

[0204] In some embodiments, the acquisition steps in the above method can also be executed by the processor 9101, for example, acquiring information from the memory 9102.

[0205] Optionally, the electronic device 9100 further includes one or more interface circuits 9104 connected to the memory 9102. The interface circuits 9104 can be used to receive signals from the memory 9102 or other devices, and can be used to send signals to the memory 9102 or other devices. For example, the interface circuits 9104 can read instructions stored in the memory 9102 and send the instructions to the processor 9101.

[0206] The electronic device 9100 described in the above embodiments may be a network device or a terminal, but the scope of the electronic device 9100 described in this disclosure is not limited thereto, and the structure of the electronic device 9100 may vary. Figure 6 There are limitations. Electronic devices can be standalone devices or part of a larger device.

[0207] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program commands. The aforementioned program can be stored in a storage medium, including various media capable of storing program code such as mobile storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks or optical disks.

[0208] Alternatively, if the integrated units described above are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several commands to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.

[0209] It should be understood that the above embodiments are exemplary and are not intended to encompass all possible implementations included in the claims. Various modifications and changes can be made to the above embodiments without departing from the scope of this disclosure. Similarly, the various technical features of the above embodiments can be arbitrarily combined to form other embodiments of the present invention that may not be explicitly described. Therefore, the above embodiments only illustrate several implementations of the present invention and do not limit the scope of protection of this patent.

Claims

1. A method for avoiding residential areas, characterized in that, Applied to a first terminal, the method includes: Using a first prediction model, based on the first spatiotemporal information of the first terminal, the probability of anomalies in the cell associated with the first terminal is predicted. The first prediction model is trained on the server based on network behavior data collected by the second terminal. Perform corresponding avoidance operations on abnormal cells whose anomaly probability is greater than the probability threshold.

2. The community avoidance method according to claim 1, characterized in that, The step of performing corresponding avoidance operations on abnormal cells with an anomaly probability greater than a probability threshold includes at least one of the following: In response to the anomaly probability being greater than or equal to a first probability threshold and less than or equal to a second probability threshold, the measured value of the anomalous cell is negatively biased based on the anomaly probability, wherein the second probability threshold is greater than the first probability threshold. In response to the anomaly probability being greater than the second probability threshold, the abnormal cell will not be camped for at least the prohibition period, wherein the prohibition period is determined based on the anomaly probability.

3. The community avoidance method according to claim 2, characterized in that, The bias value for negatively biasing the measured value includes: the product of a first probability weight and a first preset value, plus the sum of a second preset value, wherein the first probability weight is positively correlated with the anomaly probability.

4. The community avoidance method according to claim 2, characterized in that, The prohibition duration includes: the product of the second probability weight and the first duration, plus the sum of the second duration, wherein the second probability weight is positively correlated with the anomaly probability.

5. The community avoidance method according to claim 1, characterized in that, The network behavior data includes: The identification information of the second terminal; The second spatiotemporal information corresponding to the second terminal; The cell identifier corresponding to the network behavior of the second terminal; The network type in which the second terminal performs network behavior; The result of the second terminal performing network actions at the corresponding network layer.

6. The cell avoidance method according to claim 1 or 2, characterized in that, The first prediction model includes a lightweight format model of the second prediction model obtained by the server based on the network behavior data.

7. A method for avoiding conflicts in residential communities, characterized in that, Applied to a server, the method includes: A first prediction model is obtained by training based on network behavior data collected by the second terminal. The first prediction model is used to enable the first terminal to predict the anomaly probability of the cell associated with the first terminal based on the first spatiotemporal information of the first terminal, and to perform corresponding avoidance operations on the abnormal cells whose anomaly probability is greater than the probability threshold.

8. The community avoidance method according to claim 7, characterized in that, The first terminal performs corresponding avoidance operations on abnormal cells with an anomaly probability greater than a probability threshold, including at least one of the following: The first terminal responds to the fact that the anomaly probability is greater than or equal to a first probability threshold and less than or equal to a second probability threshold by negatively biasing the measured value of the abnormal cell based on the anomaly probability, wherein the second probability threshold is greater than the first probability threshold; The first terminal responds to the anomaly probability being greater than the second probability threshold by not camping on the abnormal cell for at least the prohibition period, wherein the prohibition period is determined based on the anomaly probability.

9. The community avoidance method according to claim 8, characterized in that, The bias value for negatively biasing the measured value includes: the product of a first probability weight and a first preset value, plus the sum of a second preset value, wherein the first probability weight is positively correlated with the anomaly probability.

10. The cell avoidance method according to claim 8, characterized in that, The prohibition duration includes: the product of the second probability weight and the first duration, plus the sum of the second duration, wherein the second probability weight is positively correlated with the anomaly probability.

11. A community avoidance device, characterized in that, include: The device includes a processing module, the processing module being used for: The first prediction model is used to predict the probability of anomalies in the cell associated with the first terminal based on the current spatiotemporal information of the first terminal. The first prediction model is trained on the server based on network behavior data collected by the second terminal. Perform corresponding avoidance operations on abnormal cells whose anomaly probability is greater than the probability threshold.

12. A community avoidance device, characterized in that, include: The device includes a processing module, the processing module being used for: A first prediction model is obtained by training based on network behavior data collected by the second terminal. The first prediction model is used to enable the first terminal to predict the anomaly probability of the cell associated with the first terminal based on the first spatiotemporal information of the first terminal, and to perform corresponding avoidance operations on the abnormal cells whose anomaly probability is greater than the probability threshold.

13. An electronic device, characterized in that, The electronic device includes: One or more processors; The processor is configured to invoke instructions to cause the electronic device to execute the cell avoidance method according to any one of claims 1 to 10.

14. A storage medium, characterized in that, The storage medium stores instructions that, when executed on an electronic device, cause the electronic device to perform the cell avoidance method according to any one of claims 1 to 10.

Images