Tolerance threshold prediction method, apparatus, device, medium, and program product
By acquiring user characteristic information and a complaint prediction model, and combining it with probability threshold search, the maximum tolerance threshold of users is calculated in reverse. This solves the problem of low prediction accuracy in existing technologies, achieves more accurate prediction of user tolerance thresholds, and reduces user complaint rates and operator compensation.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHINA UNITED NETWORK COMM GRP CO LTD
- Filing Date
- 2026-03-11
- Publication Date
- 2026-06-12
Smart Images
Figure CN122196742A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of artificial intelligence technology, and in particular to a tolerance threshold prediction method, apparatus, device, medium, and program product. Background Technology
[0002] Currently, users have a certain psychological tolerance threshold for the service quality of various services provided by telecommunications operators. If this threshold is exceeded, it may lead to user complaints. For example, in the use of data and billing services by telecommunications operators, the extra charges incurred due to users exceeding their data allowance are one of the main reasons for user complaints.
[0003] In related technologies, a direct regression prediction scheme is mainly used to predict the minimum psychological tolerance value of users for the service quality of various services, thereby preventing user complaints and achieving precise customer retention. Specifically, this prediction method uses the user's historical complaint records and the corresponding actual amount exceeding the package limit as training labels. Through a regression model, the mapping between user characteristics and this amount is learned, thus directly predicting the user's maximum tolerance amount for exceeding the package limit.
[0004] However, in scenarios with a very small number of positive samples, the above regression model tends to predict values close to negative samples, which leads to low accuracy of the predicted values. Summary of the Invention
[0005] This application provides a tolerance threshold prediction method, apparatus, device, medium, and program product, which improves the accuracy of predicting the maximum tolerance threshold of a user for a certain business indicator.
[0006] In a first aspect, embodiments of this application provide a tolerance threshold prediction method, the method comprising: acquiring first user feature information of a first user, the first user feature information including user profile feature information and behavioral feature information of the first user's historical resource over-limit behavior for a target business indicator; determining a first tolerance threshold for the first user's excess amount for the target business indicator based on a preset tolerance threshold range; obtaining a first complaint probability based on the first user feature information and the first tolerance threshold; if the absolute difference between the first complaint probability and the preset complaint probability is less than the first threshold, then determining the first tolerance threshold as the maximum tolerance threshold for the first user's excess amount for the target business indicator; if the absolute difference between the first complaint probability and the preset complaint probability is greater than or equal to the first threshold, then narrowing the preset tolerance threshold range, and determining the maximum tolerance threshold based on the narrowed tolerance threshold range.
[0007] The technical solution provided in this application brings at least the following beneficial effects: by obtaining the first user characteristic information of the first user, which includes user profile characteristic information and behavioral characteristic information of the first user's historical resource over-limit behavior for the target business indicator, then, based on a preset tolerance threshold range, determining the first tolerance threshold for the first user's excess amount for the target business indicator, and obtaining the first complaint probability based on the first user characteristic information and the first tolerance threshold, subsequently, if the absolute difference between the first complaint probability and the preset complaint probability is less than the first threshold, then determining the first tolerance threshold as the maximum tolerance threshold for the first user's excess amount for the target business indicator; conversely, if the absolute difference between the first complaint probability and the preset complaint probability is greater than or equal to the first threshold, then narrowing the preset tolerance threshold range, and determining the above-mentioned maximum tolerance threshold based on the narrowed tolerance threshold range. In this solution, by combining the user complaint probability output by the complaint prediction model with the probability threshold search method, and by continuously narrowing the preset tolerance threshold range, the maximum tolerance threshold for a certain business indicator corresponding to the critical point when the complaint probability of each user reaches the threshold can be accurately calculated in reverse. This improves the accuracy of predicting the maximum tolerance threshold of a user for a certain business indicator, thereby providing a direct and quantifiable decision-making basis for the operator's refined operation.
[0008] One possible implementation is that, based on a preset tolerance threshold range, determining the first tolerance threshold for the first user's excess amount for the target business indicator includes: using the midpoint between the upper and lower limits of the preset tolerance threshold range as the first tolerance threshold.
[0009] Another possible implementation, the aforementioned narrowing of the preset tolerance threshold range, includes: if the first complaint probability is less than the preset complaint probability, then adjusting the lower limit of the preset tolerance threshold range to the first tolerance threshold; if the first complaint probability is greater than or equal to the preset complaint probability, then adjusting the upper limit of the preset tolerance threshold range to the first tolerance threshold.
[0010] Another possible implementation is that the above tolerance threshold prediction method obtains the first complaint probability based on the first user feature information and the first tolerance threshold, including: generating second user feature information based on the first user feature information and the first tolerance threshold; inputting the second user feature information into the complaint prediction model and outputting the first complaint probability.
[0011] Another possible implementation, wherein determining the maximum tolerance threshold based on the narrowed tolerance threshold range, includes: determining a second tolerance threshold for the first user's excess amount for the target business indicator based on the narrowed tolerance threshold range; obtaining a second complaint probability based on the first user's characteristic information and the second tolerance threshold; determining the second tolerance threshold as the maximum tolerance threshold if the absolute difference between the second complaint probability and the preset complaint probability is less than the first threshold; and further narrowing the tolerance threshold range and determining the maximum tolerance threshold based on the narrowed tolerance threshold range if the absolute difference between the second complaint probability and the preset complaint probability is greater than or equal to the first threshold.
[0012] Another possible implementation, the above tolerance threshold prediction method includes: acquiring user feature information of at least one user, and marking the user feature information of users who have historically complained about resource overuse of the target business indicator as positive samples, and the user feature information of other users as negative samples, wherein each of the above user feature information includes user profile feature information and behavioral feature information of the user's historical resource overuse behavior of the target business indicator; training a neural network model based on the user feature information of the at least one user to obtain the above complaint prediction model; wherein, during the model training process, a higher loss weight is assigned to positive samples than to negative samples.
[0013] Another possible implementation method, the tolerance threshold prediction, also includes: based on the above maximum tolerance threshold, publishing business recommendation information related to the business corresponding to the above target business indicators to the terminal of the first user.
[0014] Another possible implementation method, the tolerance threshold prediction method, also includes: based on the above maximum tolerance threshold, publishing business recommendation information related to the business corresponding to the above target business indicator to the terminal device of the first user.
[0015] Secondly, embodiments of this application provide a tolerance threshold prediction device, comprising: an acquisition module, a determination module, and a processing module. The acquisition module is configured to acquire first user characteristic information of a first user, the first user characteristic information including user profile characteristic information and behavioral characteristic information of the first user's historical resource over-limit behavior regarding a target business indicator; the determination module is configured to determine a first tolerance threshold for the first user's excess amount regarding the target business indicator based on a preset tolerance threshold range; the processing module is configured to obtain a first complaint probability based on the first user characteristic information acquired by the acquisition module and the first tolerance threshold determined by the determination module; the determination module is further configured to determine the first tolerance threshold as the maximum tolerance threshold for the first user's excess amount regarding the target business indicator if the absolute difference between the first complaint probability obtained by the processing module and the preset complaint probability is less than the first threshold; the processing module is further configured to narrow the preset tolerance threshold range if the absolute difference between the first complaint probability obtained by the processing module and the preset complaint probability is greater than or equal to the first threshold; the determination module is further configured to determine the maximum tolerance threshold based on the narrowed tolerance threshold range.
[0016] In one possible implementation, the determining module in the aforementioned tolerance threshold prediction device is specifically used to: take the midpoint between the upper limit and the lower limit of the aforementioned preset tolerance threshold range as the aforementioned first tolerance threshold.
[0017] Another possible implementation is that the processing module in the above-mentioned tolerance threshold prediction device is specifically used to: if the first complaint probability is less than the preset complaint probability, adjust the lower limit of the preset tolerance threshold range to the first tolerance threshold; if the first complaint probability is greater than or equal to the preset complaint probability, adjust the upper limit of the preset tolerance threshold range to the first tolerance threshold.
[0018] In another possible implementation, the processing module in the tolerance threshold prediction device is further used to: generate second user feature information based on the first user feature information and the first tolerance threshold; input the second user feature information into the complaint prediction model and output the first complaint probability.
[0019] Another possible implementation is that the processing module in the aforementioned tolerance threshold prediction device is further used for:
[0020] Based on the narrowed tolerance threshold range, a second tolerance threshold for the first user's excess amount for the target business indicator is determined; based on the first user's characteristic information and the second tolerance threshold, a second complaint probability is obtained; if the absolute difference between the second complaint probability and the preset complaint probability is less than the first threshold, the second tolerance threshold is determined as the maximum tolerance threshold; if the absolute difference between the second complaint probability and the preset complaint probability is greater than or equal to the first threshold, the tolerance threshold range is further narrowed, and the maximum tolerance threshold is determined based on the narrowed tolerance threshold range.
[0021] Another possible implementation, the tolerance threshold prediction method, includes: the acquisition module is further configured to acquire user feature information of at least one user; the processing module is further configured to label the user feature information of users who have historically complained about resource overuse of the target business indicator as positive samples, and the user feature information of other users as negative samples, wherein each user feature information includes user profile feature information and behavioral feature information of the user's historical resource overuse behavior regarding the target business indicator; the processing module is further configured to train a neural network model based on the user feature information of the at least one user to obtain the complaint prediction model; wherein, during the model training process, a higher loss weight is assigned to positive samples than to negative samples.
[0022] In another possible implementation, the tolerance threshold prediction device further includes a publishing module, which is used to publish business recommendation information related to the business corresponding to the target business indicator to the terminal of the first user based on the maximum tolerance threshold.
[0023] Thirdly, this application provides an electronic device comprising: a processor and a memory; the memory stores a program or instructions executable on the processor, wherein the program or instructions, when executed by the processor, implement the method of the first aspect described above.
[0024] Fourthly, this application provides a readable storage medium on which a program or instructions are stored, which, when executed by a computer, implement the method of the first aspect described above.
[0025] Fifthly, this application provides a computer program product stored in a storage medium, which, when executed by a computer, implements the method described in the first aspect.
[0026] In a sixth aspect, embodiments of this application provide a chip including a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement the method described in the first aspect.
[0027] The beneficial effects of the second to sixth aspects mentioned above are described in the corresponding description of the first aspect and will not be repeated here. Attached Figure Description
[0028] Figure 1 A schematic diagram of the network architecture for an application of a tolerance threshold prediction method provided in this application embodiment;
[0029] Figure 2 A flowchart illustrating a tolerance threshold prediction method provided in an embodiment of this application;
[0030] Figure 3 A flowchart illustrating another tolerance threshold prediction method provided in this application embodiment;
[0031] Figure 4 A flowchart illustrating another tolerance threshold prediction method provided in this application embodiment;
[0032] Figure 5 A flowchart illustrating another tolerance threshold prediction method provided in this application embodiment;
[0033] Figure 6 A flowchart illustrating another tolerance threshold prediction method provided in this application embodiment;
[0034] Figure 7 A flowchart illustrating the implementation process of a tolerance threshold prediction method provided in this application embodiment;
[0035] Figure 8 A schematic diagram illustrating an iterative process using a search algorithm, provided as an embodiment of this application;
[0036] Figure 9 This is a schematic diagram of the structure of a tolerance threshold prediction device provided in an embodiment of this application;
[0037] Figure 10 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation
[0038] The tolerance threshold prediction method, apparatus, equipment, medium, and program products provided in this application will now be described in detail with reference to the accompanying drawings.
[0039] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.
[0040] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.
[0041] The terms "at least one," "at least one of," etc., used in the specification and claims of this application refer to any one, any two, or a combination of two or more of the included items. For example, at least one of a, b, and c can mean: "a," "b," "c," "a and b," "a and c," "b and c," and "a, b, and c," where a, b, and c can be single or multiple. Similarly, "at least two" refers to two or more items, and its meaning is similar to that of "at least one."
[0042] In the description of this application, unless otherwise stated, "a plurality of" means two or more.
[0043] Currently, among users of telecommunications operators' data and billing services, additional charges incurred due to exceeding data allowances are a major cause of user complaints. Operators, facing public pressure, are often forced to compensate users to quell these complaints. Therefore, accurately predicting the maximum amount users can tolerate for exceeding their data allowances is crucial for developing personalized plans, conducting targeted marketing, and providing early warnings of service suspension to prevent complaints.
[0044] In related technologies, prediction is mainly carried out through the following two methods:
[0045] 1) Direct Regression Prediction Scheme: This method treats the prediction of a user's maximum tolerable amount as a standard regression problem. Its core principle is to assume a definite functional relationship between the user's maximum tolerable amount and their features. This relationship is learned by minimizing the mean squared error between the predicted and actual amounts, or other loss functions. Specifically, the technical steps include: a) collecting user features (such as historical over-limit amounts) from the business system; b) using the actual over-limit amounts of users who have historically complained about over-limit fees as training labels for the regression model; c) using a regression algorithm (such as linear regression or other machine learning models) to fit the features and labels, training a mapping function M=F(X) from features to amounts; d) for new users, directly inputting their feature vector X into the trained regression model, outputting a predicted maximum tolerable amount M.
[0046] 2) Simple threshold classification scheme: This scheme defines the problem as a binary classification problem, aiming to predict whether a user will complain at a certain cost level. Its core principle is based on statistical classification theory, distinguishing two types of users through features. However, in business applications, this method can only determine the risk level of a user's complaint in the current state, and cannot directly give a specific threshold amount. Specifically, the technical steps include: a) collecting feature data similar to the regression scheme; b) defining binary labels: labeling users who have complained as positive samples (1), and users who have not complained as negative samples (0); c) training a binary classification model (such as logistic regression, support vector machine, or LightGBM classifier), which outputs the probability that a user belongs to the complaint class; d) setting a fixed probability threshold (usually 0.5). When the model predicts a probability P≥0.5, it is determined that the user will complain; when P<0.5, it is determined that the user will not complain.
[0047] However, none of the above solutions can effectively address the challenge of accurately and interpretably predicting the extreme value of a user's maximum tolerance for over-leveraging under extremely imbalanced data. In other words, the drawback of the above solutions is:
[0048] 1. Sensitive to imbalance in the number of positive and negative samples: In scenarios with very few positive samples, the direct regression method tends to predict values close to those of negative samples, and cannot accurately capture the critical point of the amount exceeding the limit that leads to complaints.
[0049] 2. The prediction results lack direct business meaning: The probability values output by the classification model cannot be directly converted into a "maximum tolerable amount" that can be understood at the business level. They can only express the probability of a user complaint based on the current excess amount. Furthermore, it cannot accurately express or obtain the excess amount at which a user is about to file a complaint.
[0050] 3. Inability to proactively locate extreme points: Existing methods passively predict the probability of a user's behavior in the current state, and cannot proactively calculate and simulate the critical cost value for a user to go from "not complaining" to "complaining".
[0051] To address the aforementioned issues, this application provides a tolerance threshold prediction method, apparatus, device, medium, and program product that can be applied to core business scenarios such as personalized package development, precision marketing, traffic usage intervention, service outage warning, pre-complaint prevention, customer retention, and tariff optimization. This can reduce the rate of complaints about excessive charges from the source, reduce unreasonable compensation and profit loss for operators due to public opinion pressure, and improve user perception and tariff transparency.
[0052] In one possible scenario, such as Scenario 1:
[0053] Scenario 1: Scenario where mobile data usage exceeds the plan limit, resulting in additional charges.
[0054] In cases where users exceed their mobile data allowance and incur additional charges, and given the extremely limited historical sample of user complaints about exceeding the allowance, the system can accurately predict the user's psychological tolerance limit for these charges and provide advance data usage reminders, tariff warnings, or plan upgrade recommendations to avoid complaints arising from exceeding the user's tolerance threshold.
[0055] In one possible scenario, such as scenario 2:
[0056] Scenario 2: Incurring additional charges when basic communication services such as voice calls or SMS exceed the plan's limit.
[0057] For scenarios where users exceed their package limits for basic communication services such as voice calls or SMS, resulting in additional charges, the system predicts the upper limit of user tolerance for over-limit charges for user groups with sparse complaint samples, thus avoiding complaints caused by over-limit charges exceeding the user's tolerance threshold.
[0058] Furthermore, the applications of this application are not limited to credit control scenarios. They can also be applied to key areas such as churn warning, value-added service marketing, and network quality sensitivity assessment. For example, in churn warning scenarios, it can predict the minimum tolerance level of users for network quality or service experience, thereby achieving more precise customer retention. It can also be applied to any scenario involving the prediction of extreme user behavior and the judgment of risk thresholds. For instance, in the internet service sector, it can predict the tolerance boundaries of users for price adjustments and service degradation.
[0059] To address the aforementioned technical problems, embodiments of this application provide a method, apparatus, device, medium, and program product for predicting tolerance thresholds. The method involves acquiring first user characteristic information of a first user, including user profile characteristic information and behavioral characteristic information of the first user's historical resource over-limit behavior regarding a target business indicator. Then, based on a preset tolerance threshold range, a first tolerance threshold for the first user's excess amount regarding the target business indicator is determined. A first complaint probability is obtained based on the first user characteristic information and the first tolerance threshold. Subsequently, if the absolute difference between the first complaint probability and the preset complaint probability is less than the first threshold, the first tolerance threshold is determined as the maximum tolerance threshold for the first user's excess amount regarding the target business indicator. Conversely, if the absolute difference between the first complaint probability and the preset complaint probability is greater than or equal to the first threshold, the preset tolerance threshold range is narrowed, and the maximum tolerance threshold is determined based on the narrowed tolerance threshold range. In this solution, by combining the user complaint probability output by the complaint prediction model with the probability threshold search method, and by continuously narrowing the preset tolerance threshold range, the maximum tolerance threshold for a certain business indicator corresponding to the critical point when the complaint probability of each user reaches the threshold can be accurately calculated in reverse. This improves the accuracy of predicting the maximum tolerance threshold of a user for a certain business indicator, thereby providing a direct and quantifiable decision-making basis for the operator's refined operation.
[0060] The tolerance threshold prediction method, apparatus, device, medium, and program products provided in this application will be described in detail below with reference to the accompanying drawings.
[0061] Figure 1 The diagram illustrates a network architecture for applying a tolerance threshold prediction method according to an embodiment of this application. For example... Figure 1 As shown, the network architecture includes a tolerance threshold prediction device 101 and a terminal device 102. The tolerance threshold prediction device 101 and the terminal device 102 are interconnected.
[0062] In some embodiments, the tolerance threshold prediction device 101 described above may be a server, a computer, or a processor or processing unit within a server or computer. The server may be a single server or a server cluster consisting of multiple servers. It should be noted that the embodiments of this application do not limit the specific device form of the tolerance threshold prediction device 101. Figure 1 The tolerance threshold prediction device 101 is illustrated as a single server.
[0063] In some embodiments, the terminal device may be a mobile phone, tablet computer, laptop computer, handheld computer, in-vehicle electronic device, mobile internet device (MID), augmented reality (AR) / virtual reality (VR) device, robot, wearable device, personal computer (PC), ultra-mobile personal computer (UMPC), netbook, or personal digital assistant (PDA), etc., and the embodiments of this application do not specifically limit it. Figure 1 The example shown is a mobile phone, with terminal device 102 as an example.
[0064] In some embodiments, the tolerance threshold prediction device 101 acquires first user characteristic information of a first user, which includes user profile characteristic information and behavioral characteristic information of the first user's historical resource over-limit behavior for the target business indicator. Then, based on a preset tolerance threshold range, it determines a first tolerance threshold for the first user's excess amount for the target business indicator and obtains a first complaint probability based on the first user characteristic information and the first tolerance threshold. Subsequently, if the absolute difference between the first complaint probability and the preset complaint probability is less than the first threshold, the first tolerance threshold is determined as the maximum tolerance threshold for the first user's excess amount for the target business indicator; conversely, if the absolute difference between the first complaint probability and the preset complaint probability is greater than or equal to the first threshold, the preset tolerance threshold range is narrowed, and the maximum tolerance threshold is determined based on the narrowed tolerance threshold range. Finally, based on the maximum tolerance threshold, business recommendation information related to the business corresponding to the target business indicator can be published to the terminal device 102 corresponding to the first user.
[0065] It should be noted that the network architecture described in the embodiments of this application is for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and does not constitute a limitation on the technical solutions provided in the embodiments of this application. As network architectures evolve, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.
[0066] See Figure 2 This is a flowchart illustrating a tolerance threshold prediction method provided in an embodiment of this application. Figure 2 As shown, the tolerance threshold prediction method provided in this application embodiment can be implemented by the above-mentioned tolerance threshold prediction device, specifically including the following steps 201 to 205.
[0067] Step 201: The tolerance threshold prediction device acquires the first user feature information of the first user.
[0068] In this embodiment of the application, the aforementioned first user feature information includes user profile feature information and behavioral feature information of the first user’s historical resource over-limit behavior in relation to the target business indicators.
[0069] In this embodiment of the application, the aforementioned user profile feature information is used to characterize at least one of the user's attribute features, behavioral features, preference features, and status features. For example, age, gender, network access duration, and service plan level.
[0070] In this embodiment of the application, the aforementioned user profile feature information can be considered as the user's static features, and the aforementioned behavioral feature information can be considered as dynamic traffic feature information.
[0071] In this application embodiment, the aforementioned target service indicators include, but are not limited to: mobile data traffic, basic communication service package amount, network quality, network signal strength, etc.
[0072] In this application embodiment, the above-mentioned resource over-limit behavior includes, but is not limited to: using mobile data traffic exceeding the package limit, using basic communication services such as voice calls or SMS exceeding the package limit and incurring additional charges, etc.
[0073] For example, the aforementioned behavioral characteristic information may include the user's recent records of excessive sexual activity, such as the average amount of excessive sexual activity in the past three months, the maximum amount of excessive sexual activity in the past three months, and the frequency of excessive sexual activity.
[0074] Step 202: The tolerance threshold prediction device determines the first tolerance threshold for the first user's excess amount for the aforementioned target business indicators based on a preset tolerance threshold range.
[0075] In this embodiment, the lower limit of the preset tolerance threshold range can be 0, and the upper limit can be determined based on the target business indicators. In other words, the upper limit can be the upper limit of the excess amount set based on the target business indicators.
[0076] In this embodiment, the aforementioned preset tolerance threshold range can also be referred to as the tolerance threshold search range. That is, the tolerance threshold prediction device uses a search algorithm to continuously iterate over the preset tolerance threshold range to find the maximum tolerance threshold for the first user's excess amount for the target business indicator.
[0077] In the embodiments of this application, the tolerance threshold prediction device may use a binary search method, a golden section method, a gradient descent method, or a more complex optimization algorithm for iterative processing.
[0078] In some embodiments, combined with Figure 2 ,like Figure 3As shown, step 202 above can be implemented through step 202a as follows.
[0079] Step 202a: The tolerance threshold prediction device uses the midpoint between the upper and lower limits of the preset tolerance threshold range as the first tolerance threshold.
[0080] In this way, by using the midpoint between the upper and lower limits of the preset tolerance threshold range as the assumed tolerance threshold, the computational complexity is reduced and the prediction efficiency is improved.
[0081] Step 203: The tolerance threshold prediction device obtains the first complaint probability based on the first user characteristic information and the first tolerance threshold.
[0082] In some embodiments, combined with Figure 2 ,like Figure 4 As shown, step 203 above can be implemented through steps 203a to 203b.
[0083] Step 203a: The tolerance threshold prediction device generates second user feature information based on the first user feature information and the first tolerance threshold.
[0084] Step 203b: The tolerance threshold prediction device inputs the aforementioned second user characteristic information into the complaint prediction model and outputs the first complaint probability.
[0085] Step 204: If the absolute difference between the first complaint probability and the preset complaint probability is less than the first threshold, the tolerance threshold prediction device determines the first tolerance threshold as the maximum tolerance threshold for the first user's excess amount for the target business indicator.
[0086] Step 205: If the absolute difference between the first complaint probability and the preset complaint probability is greater than or equal to the first threshold, the tolerance threshold prediction device narrows the preset tolerance threshold range and determines the maximum tolerance threshold based on the narrowed tolerance threshold range.
[0087] In some embodiments, combined with Figure 2 ,like Figure 5 As shown, the "reducing the range of the preset tolerance threshold" in step 205 can be achieved through the following steps 205a1 and 205a2.
[0088] Step 205a1: If the first complaint probability is less than the preset complaint probability, the tolerance threshold prediction device adjusts the lower limit of the preset tolerance threshold range to the first tolerance threshold.
[0089] Step 205a2: If the first complaint probability is greater than or equal to the preset complaint probability, the tolerance threshold prediction device adjusts the upper limit of the preset tolerance threshold range to the first tolerance threshold.
[0090] In some embodiments, combined with Figure 2 ,like Figure 5 As shown, the "determining the maximum tolerance threshold based on the reduced tolerance threshold range" in step 205 can be specifically implemented through steps 205b1 to 205b4.
[0091] Step 205b1: The tolerance threshold prediction device determines the second tolerance threshold for the first user's excess amount for the target business indicator based on the narrowed tolerance threshold range.
[0092] Step 205b2: The tolerance threshold prediction device obtains the second complaint probability based on the first user characteristic information and the second tolerance threshold.
[0093] In some embodiments, step 205b2 can be implemented through the following process: the tolerance threshold prediction device generates third user feature information based on the first user feature information and the second tolerance threshold; the tolerance threshold prediction device inputs the third user feature information into the complaint prediction model and outputs a second complaint probability.
[0094] In some embodiments, after determining a new hypothetical tolerance threshold each time, the tolerance threshold prediction device combines the new hypothetical tolerance threshold with the first user feature information to construct new user feature information, and predicts the complaint probability corresponding to the new user feature information. By continuously iterating and optimizing the hypothetical tolerance threshold, a tolerance limit that is infinitely close to the "warning probability" that the first user is about to file a complaint can be determined.
[0095] Step 205b3: If the absolute difference between the second complaint probability and the preset complaint probability is less than the first threshold, the tolerance threshold prediction device determines the second tolerance threshold as the maximum tolerance threshold.
[0096] Step 205b4: If the absolute difference between the second complaint probability and the preset complaint probability is greater than or equal to the first threshold, the tolerance threshold prediction device continues to narrow the tolerance threshold range and determines the maximum tolerance threshold based on the narrowed tolerance threshold range.
[0097] In this embodiment of the application, if the absolute difference between the second complaint probability and the preset complaint probability is greater than or equal to the first threshold, the tolerance threshold range is further narrowed, and the narrowed tolerance threshold range is further searched and iterated until a tolerance threshold corresponding to a complaint probability whose absolute difference with the preset complaint probability is less than the first threshold is found.
[0098] In this way, by continuously iterating the search region, the adjusted search region can be continuously focused on the vicinity of the optimal solution, thereby improving the accuracy of the prediction.
[0099] In some embodiments, combined with Figure 2 ,like Figure 6 As shown, the tolerance threshold prediction method provided in this application embodiment may further include the following step 301:
[0100] Step 301: Based on the above maximum tolerance threshold, the tolerance threshold prediction device publishes business recommendation information related to the target business indicator to the terminal device of the first user.
[0101] It should be noted that, since this application can predict the corresponding maximum tolerance threshold for different users based on different user profiles, different business recommendation information can be published to the terminal devices of different users based on the corresponding maximum tolerance threshold.
[0102] In this way, different services can be recommended to different users, effectively preventing high costs incurred by users due to resource overruns, thereby significantly reducing user complaint rates and helping operators reduce unnecessary compensation payments due to public pressure. This provides operators with direct and quantifiable decision-making basis for refined operations.
[0103] It should be noted that step 301 can be performed after step 204 or step 205.
[0104] In some embodiments, the tolerance threshold prediction method provided in this application may further include the following steps 401 and 402:
[0105] Step 401: The tolerance threshold prediction device acquires user characteristic information of at least one user, and marks the user characteristic information of users who have complained about the excessive use of resources for the above-mentioned target business indicators in the past as positive samples, and the user characteristic information of other users as negative samples.
[0106] Each of the aforementioned user feature information includes user profile feature information and behavioral feature information of the user's historical resource over-limit behavior in relation to the target business indicator.
[0107] Typically, the number of negative samples is much greater than the number of positive samples.
[0108] Step 402: The tolerance threshold prediction device trains the neural network model based on the user feature information of at least one user to obtain a complaint prediction model.
[0109] In this process, positive samples are assigned a higher loss weight than negative samples during model training.
[0110] In this embodiment, users who have historically complained about excessive resource usage related to the aforementioned target business metrics can be marked as 1 (positive samples) and otherwise as 0 (negative samples). Typically, since positive samples are extremely rare, techniques to address type imbalance need to be employed during model training, such as assigning significantly higher weights to positive samples than to negative samples.
[0111] For example, taking the above complaint prediction model as a distributed LightGBM classification model, the weights of positive samples can be set to be much higher than those of negative samples in LightGBM (e.g., setting is_unbalance=True or scale_pos_weight = (number of negative samples / number of positive samples)) to alleviate the class imbalance problem.
[0112] In this way, predictive models are trained based on user characteristic information, and data-driven methods are used to uncover deep patterns from historical behavior, building reusable and accurate user profiles and demand prediction capabilities. This provides a stable and reliable technical foundation for proactive and personalized service recommendations, and also makes preventative intervention strategies (such as preventing traffic overload) scalable, thereby systematically reducing operational risks.
[0113] It should be noted that steps 401 and 402 above can be performed before step 203b above.
[0114] This application, based on training a high-precision complaint classification model, uses a search algorithm to predict complaint probabilities to back-calculate the excess fee amount corresponding to when the complaint probability of each user reaches a preset business threshold, i.e., the maximum excess fee tolerance amount. This solution is insensitive to data imbalance, the results have clear business interpretability, and it can proactively locate the extreme points for individual users.
[0115] The tolerance threshold prediction method provided in this application combines the user complaint probability output by the complaint prediction model with the probability threshold search method. By continuously narrowing the preset tolerance threshold range, it accurately back-calculates the maximum tolerance threshold for a certain business indicator corresponding to the critical point when the complaint probability of each user reaches the threshold. This improves the accuracy of predicting the maximum tolerance threshold of a user for a certain business indicator, thereby providing a direct and quantifiable decision-making basis for the operator's refined operation.
[0116] The tolerance threshold prediction method of this application will be described below through specific embodiments.
[0117] like Figure 7As shown, the implementation process of the tolerance threshold prediction method provided in this application embodiment includes the following four stages: data preparation stage, model training stage, extreme value prediction probability search stage, and extreme value business application stage, wherein:
[0118] 1) Data preparation stage:
[0119] The tolerance threshold prediction device extracts historical user data from the business system for the past N months (e.g., twelve months). Then, it constructs static and dynamic traffic features, defining the label Y as a binary variable: users who have filed complaints about exceeding service limits during the observation period are marked as 1 (positive sample), otherwise as 0 (negative sample). The static features include, but are not limited to, user attributes such as age, gender, duration of service, and service plan level. The dynamic traffic features include, but are not limited to, recent records of exceeding service limits, such as the average over-limit fee in the past three months, the maximum over-limit amount in the past three months, and the frequency of over-limit fees.
[0120] (ii) Model training phase;
[0121] A distributed LightGBM classification model is trained using a distributed machine learning framework (such as SynapseML). The static features and dynamic traffic features built in the previous stage are used as input to the model, outputting the probability P (0 ≤ P ≤ 1) of a user complaining about the current overcharge level. Simultaneously, positive samples are assigned significantly higher weights than negative samples during training (e.g., is_unbalance=True or scale_pos_weight = (number of negative samples / number of positive samples)). Evaluation metrics are then selected, typically PR-AUC (area under the precision-recall curve) and F1-Score as the primary metrics because they are more sensitive to class imbalance and more reliable than accuracy. Based on this, distributed training techniques are used to iteratively train the model, leveraging LightGBM's distributed learning capabilities to efficiently handle large-scale data and perform distributed training.
[0122] III) Extreme Value Prediction Probability Search Stage
[0123] This stage is the core of this application. In this stage, for the trained model and each user, the goal is to find an excess amount M such that when the user's "excess amount" feature is replaced with M, the model's predicted complaint probability P approaches a threshold P_t set by business experts (e.g., P_t = 0.8, indicating that an 80% complaint probability is considered the tolerance limit, and an error tolerance value ε = 0.01 is set). This threshold can be considered the "warning probability" that the user is about to file a complaint.
[0124] Specifically, such as Figure 8As shown, this stage includes two processes: an initialization process and an iterative search process, wherein:
[0125] a) Initialize the search range
[0126] For the user to be predicted, obtain all characteristic values except for "excess amount", denoted as . Set an initial search range for excess amounts. (i.e., the aforementioned tolerance threshold range), for example, =0, =Based on the upper limit value of the over-set limit given by the business.
[0127] b) Iterative search process, which can be performed iteratively using a search algorithm (e.g., binary search). The specific iterative process may include the following steps:
[0128] Step 1: Find the midpoint of the current search range .
[0129] Step 2: Set the user's "over-spending amount" feature value to M_mid obtained in a, while keeping other features unchanged, to form a new feature vector. .
[0130] Step 3: Input the new feature vector Χ' into the LightGBM model trained during the above model training process to obtain the predicted complaint probability. .
[0131] Step 4: Calculate the absolute difference between P_mid and the target probability P_t, |P_mid-P_t|.
[0132] Step 5: Determine: If |P_mid-P_t| < ε, then the extreme point has been found, and the current M_mid is the user's maximum tolerance amount; otherwise, continue execution.
[0133] Step 6: Adjust the search scope.
[0134] If P_mid < P_t, it means that the current amount M_mid has not yet reached the user's tolerance limit. The next search should be conducted in a higher amount range, that is, let M_low = M_mid.
[0135] If P_mid > P_t, it means that the current amount M_mid has exceeded the user's tolerance limit, and the next search should be conducted in a lower amount range, that is, let M_high = M_mid.
[0136] Output: When the convergence condition is met, the iteration stops, and the final M_mid is output as the user's maximum overcharge tolerance amount.
[0137] IV) Extreme Value Business Application Stage
[0138] At this stage, the maximum over-limit amount for each user can be applied to the actual business system to achieve refined operations and prevent complaints. Specific business application examples include, but are not limited to: targeted marketing of service packages, risk warnings and friendly reminders for over-limit fees, and automatic shutdown control based on tolerance limits (complaint prevention).
[0139] For targeted marketing of service packages, one approach is to compare a user's maximum tolerance amount, M_mid, with the user's current package price and the average over-limit amount over the past six months. If M_mid is significantly and consistently higher than the price of the current package's upgrade tier, then a more suitable upgrade package can be precisely recommended to the user via app push notifications, SMS, or an account manager. In this way, implicit tolerance can be transformed into explicit marketing signals, increasing the success rate of package upgrades and fundamentally reducing the risk of over-limit payments and potential complaints due to incompatible packages.
[0140] For over-usage fee risk warnings and friendly reminders, the application method can be to trigger the warning mechanism when a user's monthly over-usage fee approaches a certain percentage (e.g., 80%) of their historical M_mid value. At this time, the system can automatically send a friendly reminder SMS or APP message to the user, informing them of their current usage and cost status, preventing a bill "shock". In this way, the system shifts from passively handling complaints to proactively managing expectations, greatly improving the user experience and resolving potential complaints before they occur, reflecting the refinement and humanization of the service.
[0141] For automatic shutdown control based on tolerance limits (complaint prevention), the application method can be to set rules in the business support system. For example, when a user's real-time overcharge for the month reaches their predicted M_mid value, the system can automatically trigger a shutdown operation, or switch to a speed-limited usage mode (such as reducing the speed to 128kbps). In this way, it can completely avoid users generating "sky-high bills" that far exceed their psychological tolerance, and technically eliminate serious complaints caused by such bills.
[0142] In summary, this application successfully transforms the probability output of a classification model into an extreme value with clear business significance through the four-stage process described above, providing operators with direct and quantifiable decision-making basis for refined operations. It ultimately transforms raw user data into intelligent strategies that can directly drive business actions, forming a complete closed loop from data perception to decision execution, realizing the ultimate value of technology empowering business.
[0143] In other words, this application proposes a framework transformation mechanism for extreme value prediction, which can transform regression or simple classification problems into a two-stage framework of "classification model + probability threshold search". The core is to use the classification model to predict the probability, and then use the predicted probability search algorithm to infer the critical point. At the same time, for scenarios with a very small number of positive samples in massive data, an extreme value localization scheme is designed that combines efficient search algorithms such as binary search to pre-train the classification model, so that the model has a certain robustness in extreme value prediction under imbalanced data classes.
[0144] It should be noted that the specific steps of the above four stages in this embodiment can be found in the descriptions in the above embodiments, and will not be repeated here.
[0145] It should be noted that the above-described method embodiments, or the various possible implementations of the method embodiments, can be executed individually, or, provided there is no conflict, they can be combined with each other. The specific implementation can be determined according to actual usage requirements, and this application embodiment does not impose any restrictions on this.
[0146] As can be seen, the above mainly describes the solutions provided by the embodiments of this application from a methodological perspective. To achieve the above functions, the embodiments of this application provide corresponding hardware structures and / or software modules for executing each function. Those skilled in the art should readily recognize that, in conjunction with the modules and algorithm steps of the various examples described in the embodiments disclosed herein, the embodiments of this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.
[0147] This application embodiment can divide the tolerance threshold prediction device into functional modules according to the above method example. For example, each function can be divided into a separate functional module, or two or more functions can be integrated into one processing module. The integrated module can be implemented in hardware or as a software functional module. Optionally, the module division in this application embodiment is illustrative and only represents one logical functional division; other division methods may be used in actual implementation.
[0148] In some embodiments, this application also provides a tolerance threshold prediction apparatus. The tolerance threshold prediction apparatus may include one or more functional modules for implementing the tolerance threshold prediction method of the above method embodiments.
[0149] For example, Figure 9 This is a schematic diagram of a tolerance threshold prediction device provided in an embodiment of this application. Figure 9 As shown, the tolerance threshold prediction device 900 includes: an acquisition module 901, a determination module 902, and a processing module 903, wherein:
[0150] The acquisition module 901 is used to acquire the first user feature information of the first user, which includes user profile feature information and behavioral feature information of the first user’s historical resource over-limit behavior in response to the target business indicators.
[0151] The determination module 902 is used to determine the first tolerance threshold for the first user's excess amount for the target business indicator based on a preset tolerance threshold range.
[0152] The processing module 903 is used to obtain a first complaint probability based on the first user feature information obtained by the acquisition module 902 and the first tolerance threshold determined by the determination module.
[0153] The determining module 902 is further configured to determine the first tolerance threshold as the maximum tolerance threshold for the first user's excess amount for the target business indicator if the absolute difference between the first complaint probability obtained by the processing module 903 and the preset complaint probability is less than the first threshold.
[0154] The processing module 903 is further configured to narrow the range of the preset tolerance threshold if the absolute difference between the first complaint probability and the preset complaint probability is greater than or equal to the first threshold.
[0155] The determination module 902 is also used to determine the maximum tolerance threshold mentioned above based on the reduced tolerance threshold range.
[0156] The tolerance threshold prediction device provided in this application combines the user complaint probability output by the complaint prediction model with the probability threshold search method. By continuously narrowing the preset tolerance threshold range, it accurately back-calculates the maximum tolerance threshold for a certain business indicator corresponding to the critical point when the complaint probability of each user reaches the threshold. This improves the accuracy of predicting the maximum tolerance threshold of a user for a certain business indicator, thereby providing a direct and quantifiable decision-making basis for the operator's refined operation.
[0157] In some embodiments, the determining module 902 is specifically used to: take the midpoint between the upper limit and the lower limit of the preset tolerance threshold range as the first tolerance threshold.
[0158] In some embodiments, the processing module 903 is specifically used for:
[0159] If the first complaint probability is less than the preset complaint probability, then the lower limit of the preset tolerance threshold range will be adjusted to the first tolerance threshold.
[0160] If the first complaint probability is greater than or equal to the preset complaint probability, then the upper limit of the preset tolerance threshold range will be adjusted to the first tolerance threshold.
[0161] In some embodiments, the processing module 903 is specifically used for:
[0162] Based on the first user feature information and the first tolerance threshold mentioned above, second user feature information is generated;
[0163] Input the aforementioned second user characteristic information into the complaint prediction model and output the first complaint probability.
[0164] In some embodiments, the processing module 903 is specifically used for:
[0165] Based on the narrowed tolerance threshold range, a second tolerance threshold is determined for the first user's excess amount for the target business metric.
[0166] Based on the first user characteristic information and the second tolerance threshold, the second complaint probability is obtained;
[0167] If the absolute difference between the second complaint probability and the preset complaint probability is less than the first threshold, then the second tolerance threshold is determined to be the aforementioned maximum tolerance threshold.
[0168] If the absolute difference between the second complaint probability and the preset complaint probability is greater than or equal to the first threshold, the range of the above tolerance threshold is further narrowed, and the maximum tolerance threshold is determined based on the narrowed range of tolerance threshold.
[0169] In some embodiments, the acquisition module 901 is further configured to acquire user feature information of at least one user; the processing module 903 is further configured to mark the user feature information of users who have historically complained about excessive resource usage of target business indicators as positive samples, and the user feature information of other users as negative samples, wherein each user feature information includes user profile feature information and behavioral feature information of the user's historical excessive resource usage of target business indicators; and to train a neural network model based on the user feature information of at least one user to obtain the complaint prediction model; wherein, during the model training process, a higher loss weight is assigned to positive samples than to negative samples.
[0170] In some embodiments, the tolerance threshold prediction device 900 further includes: a publishing module;
[0171] The aforementioned publishing module is used to publish business recommendation information related to the target business indicators to the terminal of the first user, based on the aforementioned maximum tolerance threshold.
[0172] It should be noted that the tolerance threshold prediction device 900 can implement all the processes implemented in the above method embodiments and achieve the same beneficial effects. To avoid repetition, it will not be described again here.
[0173] In the case where the functions of the integrated modules described above are implemented in hardware, this application provides a possible structural schematic diagram of the electronic device involved in the above embodiments. For example... Figure 10 As shown, the electronic device 90 includes: a processor 92, a communication interface 93, and a bus 94. Optionally, the electronic device 90 may also include a memory 91.
[0174] Processor 92 may implement or execute various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 92 may be a central processing unit, a general-purpose processor, a digital signal processor, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It may implement or execute various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 92 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.
[0175] Communication interface 93 is used to connect with other devices via a communication network. This communication network can be Ethernet, wireless access network, wireless local area network (WLAN), etc.
[0176] The memory 91 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto.
[0177] As one possible implementation, the memory 91 can exist independently of the processor 92. The memory 91 can be connected to the processor 92 via a bus 94 and is used to store instructions or program code. When the processor 92 calls and executes the instructions or program code stored in the memory 91, it can implement the tolerance threshold prediction method provided in the embodiments of this application.
[0178] In another possible implementation, memory 91 can also be integrated with processor 92.
[0179] Bus 94 can be an Extended Industry Standard Architecture (EISA) bus, etc. Bus 94 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 10 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.
[0180] Through the above description of the implementation methods, those skilled in the art can clearly understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the service calling device can be divided into different functional modules to complete all or part of the functions described above.
[0181] This application embodiment also provides a chip, which includes a processor and a communication interface. The communication interface is coupled to the processor. The processor is used to run programs or instructions to implement the various processes of the above-described tolerance threshold prediction method embodiment and can achieve the same technical effect. To avoid repetition, it will not be described again here.
[0182] It should be understood that the chip mentioned in the embodiments of this application may also be referred to as a system-on-a-chip, system chip, chip system, or system-on-a-chip, etc.
[0183] This application also provides a readable storage medium storing a program or instructions that, when executed by a computer, implement the tolerance threshold prediction method provided in the above embodiments. It is understood that all or part of the processes in the above method embodiments can be executed by computer instructions instructing related hardware; the readable storage medium can be any of the foregoing embodiments or memory; the readable storage medium can also be an external storage device of the service invocation device, such as a pluggable hard drive, SmartMedia Card (SMC), Secure Digital (SD) card, flash card, etc., equipped on the service invocation device. Further, the readable storage medium can include both internal storage units of the service invocation device and external storage devices. The readable storage medium is used to store the computer program and other programs and data required by the service invocation device. The readable storage medium can also be used to temporarily store data that has been output or will be output.
[0184] This application also provides a computer program product, which is stored in a storage medium and implements the tolerance threshold prediction method provided in the above embodiments when the computer program product is executed by a computer.
[0185] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.
[0186] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a computer software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of this application.
[0187] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.
Claims
1. A method for predicting tolerance thresholds, characterized in that, The method includes: Obtain the first user feature information of the first user, which includes user profile feature information and behavioral feature information of the first user’s historical resource over-limit behavior in response to the target business indicators; Based on a preset tolerance threshold range, a first tolerance threshold for the first user's excess amount for the target business indicator is determined; Based on the first user characteristic information and the first tolerance threshold, the first complaint probability is obtained; If the absolute difference between the first complaint probability and the preset complaint probability is less than the first threshold, then the first tolerance threshold is determined to be the maximum tolerance threshold for the first user for the excess amount of the target business indicator. If the absolute difference between the first complaint probability and the preset complaint probability is greater than or equal to the first threshold, then the range of the preset tolerance threshold is narrowed, and the maximum tolerance threshold is determined based on the narrowed tolerance threshold range.
2. The tolerance threshold prediction method according to claim 1, characterized in that, The step of determining the first tolerance threshold for the first user's excess amount for the target business indicator based on a preset tolerance threshold range includes: The midpoint between the upper and lower limits of the preset tolerance threshold range is taken as the first tolerance threshold.
3. The tolerance threshold prediction method according to claim 1 or 2, characterized in that, The reduction of the preset tolerance threshold range includes: If the first complaint probability is less than the preset complaint probability, then the lower limit of the preset tolerance threshold range is adjusted to the first tolerance threshold. If the first complaint probability is greater than or equal to the preset complaint probability, then the upper limit of the preset tolerance threshold range is adjusted to the first tolerance threshold.
4. The tolerance threshold prediction method according to claim 1, characterized in that, The step of obtaining the first complaint probability based on the first user characteristic information and the first tolerance threshold includes: Based on the first user feature information and the first tolerance threshold, second user feature information is generated; The second user feature information is input into the complaint prediction model, and the first complaint probability is output.
5. The tolerance threshold prediction method according to claim 1, characterized in that, Determining the maximum tolerance threshold based on the narrowed tolerance threshold range includes: Based on the narrowed tolerance threshold range, a second tolerance threshold is determined for the first user regarding the excess amount for the target business indicator; Based on the first user characteristic information and the second tolerance threshold, a second complaint probability is obtained; If the absolute difference between the second complaint probability and the preset complaint probability is less than the first threshold, then the second tolerance threshold is determined to be the maximum tolerance threshold. If the absolute difference between the second complaint probability and the preset complaint probability is greater than or equal to the first threshold, the tolerance threshold range is further narrowed, and the maximum tolerance threshold is determined based on the narrowed tolerance threshold range.
6. The tolerance threshold prediction method according to claim 1, characterized in that, The method further includes: Obtain user feature information of at least one user, and mark the user feature information of users who have complained about the resource overuse of the target business indicator in the past as positive samples, and mark the user feature information of other users as negative samples. Each user feature information includes user profile feature information and behavioral feature information of the user's past resource overuse behavior in relation to the target business indicator. The neural network model is trained based on the user feature information of at least one user to obtain the complaint prediction model; wherein, during the model training process, the positive samples are assigned a higher loss weight than the negative samples.
7. The tolerance threshold prediction method according to claim 1, characterized in that, The method further includes: Based on the maximum tolerance threshold, business recommendation information related to the target business indicator is published to the first user's terminal device.
8. A tolerance threshold prediction device, characterized in that, The tolerance threshold prediction device includes: an acquisition module, a determination module, and a processing module; The acquisition module is used to acquire the first user feature information of the first user, which includes user profile feature information and behavioral feature information of the first user’s historical resource over-limit behavior in response to the target business indicators. The determining module is used to determine, based on a preset tolerance threshold range, the first tolerance threshold for the first user's excess amount for the target business indicator; The processing module is used to obtain a first complaint probability based on the first user feature information obtained by the acquisition module and the first tolerance threshold determined by the determination module. The determining module is further configured to determine the first tolerance threshold as the maximum tolerance threshold for the first user's excess amount of the target business indicator if the absolute difference between the first complaint probability and the preset complaint probability obtained by the processing module is less than the first threshold. The processing module is further configured to narrow the range of the preset tolerance threshold if the absolute difference between the first complaint probability and the preset complaint probability obtained by the processing module is greater than or equal to the first threshold. The determining module is further configured to determine the maximum tolerance threshold based on the narrowed tolerance threshold range.
9. An electronic device, characterized in that, It includes a processor and a memory, the memory storing a program or instructions that can run on the processor, the program or instructions being executed by the processor to implement the tolerance threshold prediction method as described in any one of claims 1-8.
10. A readable storage medium, characterized in that, The readable storage medium stores a program or instructions that, when executed by a computer, implement the tolerance threshold prediction method as described in any one of claims 1-7.
11. A computer program product, characterized in that, The computer program product is stored in a storage medium, and when executed by a computer, the computer program product implements the tolerance threshold prediction method as described in any one of claims 1-7.