Broadband service auditing method, apparatus, device, medium, and product

Abstract

This disclosure relates to the field of communication technology, and in particular to a method, apparatus, device, medium, and product for broadband service auditing. In this disclosure, feature extraction is performed on broadband service data to obtain target service features, which characterize the data features of broadband service activation and maintenance. Based on the target service features and service audit rules, a first audit result is determined. Based on on-site images of broadband services and a pre-established service database, a second audit result is determined, whereby the service database consists of images of on-site activation and maintenance of historical broadband services. The first and second audit results are then fused to obtain a target audit result, which characterizes the service status of the broadband service. By combining the audit results from feature judgment and on-site image judgment, false service reporting can be more accurately located and prevented, thereby ensuring the authenticity and reliability of service data. Therefore, the accuracy of broadband service auditing can be improved.

Description

Technical Field

[0001] This disclosure belongs to the field of communication technology, specifically relating to a broadband service auditing method, apparatus, equipment, medium, and product. Background Technology

[0002] Residential broadband and dedicated line services are broadband services launched to meet the needs of different user groups. With the growth of residential broadband and dedicated line services, related maintenance fees and installation / relocation costs have also increased year by year, leading to a sharp rise in management risks. Among these risks, there may be abnormal broadband service situations, such as services that have been declared but not implemented, or services that have been discontinued but not declared. Auditing broadband services can identify business problems and risks, propose improvement suggestions, standardize cost management, and promote cost reduction and efficiency improvement for the company. Therefore, auditing and verifying residential broadband and dedicated line services is particularly important.

[0003] In related technologies, online data can be monitored, supplemented by sampling methods such as telephone follow-ups and on-site inspections to identify abnormal services and thus achieve broadband service auditing. However, this auditing method may misidentify normal services, affecting end-users of legitimate services. Furthermore, telephone follow-ups and on-site inspections are time-consuming, labor-intensive, and have low coverage. Therefore, they reduce the accuracy of broadband service auditing. Summary of the Invention

[0004] This disclosure addresses some deficiencies mentioned in the background art by providing a broadband service auditing method, apparatus, device, medium, and product, which can improve the accuracy of broadband service auditing.

[0005] In a first aspect, embodiments of this disclosure provide a broadband service auditing method, comprising: Feature extraction is performed on broadband service data to obtain target service features, which characterize the data features of broadband service activation and maintenance; Based on the target business characteristics and business audit rules, the first audit result is determined; Based on on-site images of broadband services and a pre-established service database, a second audit result is determined, wherein the service database consists of images of the activation and maintenance of services at the service site during the execution of historical broadband services. The first audit result and the second audit result are merged to obtain the target audit result, which represents the service status of the broadband service.

[0006] Optionally, the step of extracting features from broadband service data to obtain target service features includes: Feature extraction is performed on the broadband service data to obtain multiple candidate service features; The importance coefficients of each candidate business feature are obtained by analyzing each candidate business feature based on the pre-trained feature selection model. The importance coefficients of each candidate business feature are sorted to obtain the target business feature, which includes at least one of the candidate business features.

[0007] Optionally, if the feature selection model is a decision tree model, the pre-trained feature selection model analyzes each of the candidate business features to obtain the importance coefficient of each candidate business feature, including: Based on the pre-trained feature selection model, each of the candidate business features is segmented to obtain the total number of segmentations and the sum of segmentation gains for each of the candidate business features; The importance coefficient is obtained based on the quotient of the total number of segmentations and the sum of the segmentation gains.

[0008] Optionally, the target service characteristics include at least traffic; the target service characteristics also include at least one of activation time, usage duration, user status, and rate encoding.

[0009] Optionally, determining the first audit result based on the target business characteristics and business audit rules includes: Invoke the preset business audit rules; When the target business characteristics meet the conditions of the business audit rules, the first audit result is determined to represent an audit anomaly; The business audit rules include at least the following: When the target service characteristics are less than the first preset threshold for N consecutive months, the broadband service is determined to be an abnormal service. The abnormal service indicates that the authenticity of the user account is abnormal, and N is a natural number greater than 1.

[0010] Optionally, the method further includes: The feature selection model was tested using multiple candidate first thresholds and broadband service sample data to obtain the prediction accuracy and prediction recall corresponding to each candidate first threshold. From the predicted precision and the predicted recall, determine the maximum predicted precision and the maximum predicted recall. The candidate first threshold corresponding to the maximum prediction precision and the maximum prediction recall is used as the first preset threshold.

[0011] Optionally, determining the second audit result based on broadband service field images and a pre-established service database includes: Based on the broadband service channel category in the broadband service data, the broadband service field images are classified to obtain the field image classification results. Based on the classification results of each of the aforementioned on-site images, the pre-established service database is used to perform similarity recognition on each of the aforementioned broadband service on-site images to obtain similarity recognition results; The second audit result is determined based on the similarity recognition result and the second preset threshold.

[0012] Optionally, determining the second audit result based on the similarity recognition result and the second preset threshold includes: When the similarity recognition result is greater than the second preset threshold, the second audit result is determined to indicate an audit anomaly; When the similarity recognition result is less than or equal to the second preset threshold, the second audit result is determined to indicate that the audit is normal.

[0013] Optionally, the step of fusing the first audit result and the second audit result to obtain the target audit result includes: When both the first audit result and the second audit result indicate an audit anomaly, the target audit result is determined to be an audit anomaly. When the first audit result indicates an audit anomaly and the second audit result indicates an audit normality, determine whether the target business characteristic in the N+1th month is less than a first preset threshold. If yes, the target audit result is determined to be an audit anomaly; otherwise, the target audit result is determined to be an audit normal.

[0014] Optionally, the method further includes: Obtain the device type and / or service address from the on-site images of multiple broadband services; Cluster analysis is performed based on the device type and / or the service address to obtain cluster results; When the clustering results indicate that the device types are the same and / or the service addresses are the same, the authenticity of the broadband service site image is determined to be false, and the false authenticity indicates that the target audit result is an audit anomaly.

[0015] Optionally, the method further includes: Obtain initial data for broadband services; The initial data of the broadband service is processed to obtain the broadband service data; the data processing includes at least one of the following: data cleaning, removal of abnormal data, and data filling.

[0016] In a second aspect, embodiments of this disclosure provide a broadband service auditing apparatus, comprising: The extraction module is used to extract features from broadband service data to obtain target service features, which represent the data features of broadband service activation and maintenance. The first determining module is used to determine the first audit result based on the target business characteristics and business audit rules; The second determining module is used to determine the second audit result based on broadband service site images and a pre-established service database, wherein the service database consists of images of service site activation and maintenance during the execution of historical broadband services. The fusion module is used to fuse the first audit result and the second audit result to obtain the target audit result, which represents the service status of the broadband service.

[0017] In a third aspect, embodiments of this disclosure provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described broadband service auditing method.

[0018] In a fourth aspect, embodiments of this disclosure provide a computer-readable storage medium having a computer program stored thereon, the program being executed by a processor to implement the above-described broadband service auditing method.

[0019] In a fifth aspect, embodiments of this disclosure provide a computer program product including computer-readable code, or a non-volatile computer-readable storage medium carrying computer-readable code, wherein when the computer-readable code is run in a processor of an electronic device, the processor in the electronic device executes the above-described broadband service auditing method.

[0020] In this disclosure, feature extraction is performed on broadband service data to obtain target service features, which characterize the data features of broadband service activation and maintenance. Based on the target service features and service audit rules, a first audit result is determined. Based on on-site images of broadband services and a pre-established service database, a second audit result is determined. The service database consists of images of on-site activation and maintenance of historical broadband services. The first and second audit results are then fused to obtain the target audit result, which characterizes the service status of the broadband service. By feature filtering on the large amount of data generated during service activation and maintenance, priority target service features are determined as the evaluation criteria for judging whether a service is abnormal. Combining the audit results from feature judgment and on-site image judgment helps to confirm services that have been declared but not implemented or services that have been discontinued but not declared. This combination of on-site image auditing allows for more accurate location and prevention of false business reporting, thereby ensuring the authenticity and reliability of service data. Therefore, it can improve the accuracy of broadband service auditing.

[0021] It should be understood that both the foregoing general description and the following detailed description are exemplary and intended to provide further illustration of the claimed technology. Attached Figure Description

[0022] Figure 1 This is a flowchart of a broadband service auditing method provided in this disclosure.

[0023] Figure 2 Another flowchart for a broadband service auditing method provided in this disclosure.

[0024] Figure 3 A flowchart for determining the first audit result provided in this disclosure.

[0025] Figure 4 This is a schematic diagram of the structure of a broadband service auditing device provided in this disclosure.

[0026] Figure 5 This is a hardware block diagram of an electronic device provided in this disclosure.

[0027] Figure 6 This is a schematic diagram of a computer program product provided in this disclosure. Detailed Implementation

[0028] To enable those skilled in the art to better understand the technical solution of this application, the application scenario of this application will be described first below.

[0029] Home broadband is an internet access service for individual / family users, meeting consumer-level needs such as daily internet browsing, video streaming, and gaming. Leased line services are dedicated network services for enterprises / organizations, providing dedicated bandwidth, high stability, and customized solutions. Home broadband and leased line services are broadband services launched to meet the needs of different user groups. With the growth of home broadband and leased line services, related maintenance fees and installation / relocation costs have also increased year by year, leading to a sharp rise in management risks. Among these risks, there may be abnormal broadband service situations, such as services that have been declared but not implemented, or services that have been discontinued but not declared. Auditing broadband services can identify business problems and risks, propose improvement suggestions, standardize cost management, and promote cost reduction and efficiency improvement for the company. Therefore, auditing and verifying home broadband and leased line services is particularly important.

[0030] In related technologies, online data monitoring, especially traffic data monitoring, is commonly used in the auditing and verification of home broadband and leased line services, supplemented by sampling methods such as telephone follow-ups and on-site spot checks. In traffic data monitoring, existing technologies often incorporate existing monitoring methods, such as silent user data, into the identification of service anomalies to achieve broadband service auditing. However, online data, especially traffic monitoring, may misidentify normal services, lacking accuracy guarantees. Misidentified normal services being interrupted can negatively impact end users. Telephone follow-ups and on-site spot checks are time-consuming, labor-intensive, and have low coverage. Introducing silent user data, on the other hand, results in lower accuracy because many silent users are not considered to be experiencing service anomalies. Therefore, this reduces the accuracy of broadband service auditing.

[0031] To address the aforementioned technical problems, this disclosure provides an inventive concept: by performing feature filtering on the large amount of data generated during service activation and maintenance, priority target service characteristics are identified as evaluation criteria to determine whether a service is abnormal. Combining feature-based judgment with on-site image analysis provides dual audit results, enabling the identification of services that have been declared but not implemented, or services that have been discontinued but not declared. This combination of on-site image auditing allows for more accurate location and prevention of falsely reported services, thereby ensuring the authenticity and reliability of service data. Therefore, it can improve the accuracy of broadband service auditing.

[0032] The present disclosure will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present disclosure and not intended to limit it. Furthermore, it should be noted that, for ease of description, only the parts relevant to the present disclosure are shown in the drawings, not the entire structure.

[0033] Figure 1 This is a flowchart illustrating a broadband service auditing method provided in this disclosure. Figure 1 As shown, the method includes: S101: Extract features from broadband service data to obtain target service features.

[0034] Specifically, feature extraction is performed on the processed broadband service data to identify and prioritize target service features that are important for broadband service auditing. Reducing the number of service features through this filtering process effectively reduces the audit workload and improves the efficiency of broadband service auditing. These target service features represent the data characteristics related to broadband service activation and maintenance.

[0035] S102: Determine the first audit result based on the target business characteristics and business audit rules.

[0036] Specifically, by performing business anomaly correlation analysis on multiple features of broadband service data through feature extraction, the target service features are determined. Combined with preset business audit rules, it is determined whether the target service features are in an abnormal state, thereby determining whether the broadband service is in an abnormal state, so as to obtain the first audit result based on the data feature dimension.

[0037] S103: Determine the second audit result based on on-site images of broadband services and a pre-established service database.

[0038] Specifically, by comparing on-site broadband service images with images in the service database, the authenticity of the user's service is verified, and any anomalies in the on-site broadband service images are determined, thereby identifying any anomalies in the broadband service and obtaining a second audit result based on the on-site image dimension. By judging the on-site images, it is possible to more intuitively determine whether there are any services that have been declared but not implemented, or services that have been discontinued but not declared. The service database contains images of on-site service activation and maintenance during the execution of historical broadband services.

[0039] S104: The first audit result and the second audit result are merged to obtain the target audit result.

[0040] Specifically, the first audit result is the audit result of data monitoring, and the second audit result is the audit result of on-site images. These two results are combined to determine abnormal data states and abnormal image similarity states. The introduction of on-site broadband service image recognition and comparison overcomes the drawback of remote auditing, which can only retrieve online data and lacks accuracy assurance. In remote auditing, by combining online data and on-site images, abnormal situations such as user accounts being fraudulently activated / maintained broadband / dedicated lines are verified, thereby ensuring the authenticity and reliability of service data. The target audit result represents the service status of the broadband service.

[0041] In this disclosure, feature extraction is performed on broadband service data to obtain target service features, which characterize the data features of broadband service activation and maintenance. Based on the target service features and service audit rules, a first audit result is determined. Based on on-site images of broadband services and a pre-established service database, a second audit result is determined. The service database consists of images of on-site activation and maintenance of historical broadband services. The first and second audit results are then fused to obtain the target audit result, which characterizes the service status of the broadband service. By feature filtering on the large amount of data generated during service activation and maintenance, priority target service features are determined as the evaluation criteria for judging whether a service is abnormal. Combining the audit results from feature judgment and on-site image judgment helps to confirm services that have been declared but not implemented or services that have been discontinued but not declared. This combination of on-site image auditing allows for more accurate location and prevention of false business reporting, thereby ensuring the authenticity and reliability of service data. Therefore, it can improve the accuracy of broadband service auditing.

[0042] In one possible implementation, an exemplary method for extracting features from broadband service data to obtain target service features includes: Feature extraction is performed on broadband service data to obtain multiple candidate service features; each candidate service feature is analyzed based on a pre-trained feature selection model to obtain the importance coefficient of each candidate service feature; the importance coefficients of each candidate service feature are ranked to obtain the target service feature.

[0043] Specifically, the target service characteristics include at least one candidate service characteristic. For example, the target service characteristics include at least traffic; the target service characteristics also include at least one of activation time, usage duration, user status, and rate coding.

[0044] Preliminary feature screening of broadband service data yields multiple candidate service features, such as account number, activation time, service type, application channel, service item identifier code, status, month, city, phone number, data usage, usage duration, user status, and rate coding. A feature selection model is then constructed using the LightGBM feature selection algorithm to extract features such as activation time, data usage, usage duration, user status, and rate coding from the broadband service data, ensuring that the features are both service-relevant and data-representative.

[0045] In this embodiment, when calling LightGBM, the model is a decision tree model. Its hyperparameters significantly impact model performance. Therefore, considering both model performance and generalization ability, the key parameters are configured as follows: Learning rate = 0.05. The learning rate controls the contribution of each tree to the final prediction result. This configuration allows for more precise weight adjustment during training, reducing the risk of overfitting. Maximum tree depth = 7. Deeper trees may capture noise in the training set. This configuration allows the model to capture sufficient information while maintaining a certain level of generalization ability. Maximum number of leaf nodes per tree = 80. Typically, num_leaves is related to the maximum tree depth (max_depth). For example, num_leaves ≤ 2^(max_depth). This configuration provides the model with high expressive power and low overfitting risk, while ensuring the model is not overly complex.

[0046] LightGBM uses node impurities in each decision tree to calculate the importance coefficients of candidate business features. The final importance coefficient of the candidate business feature is the average of the feature importance of all decision trees. The LightGBM algorithm's built-in feature selection algorithm calculates the importance coefficients of features, sorts them, and thus determines the target business feature from multiple candidate business features.

[0047] For example, based on the selected broadband service sample data, the LightGMB algorithm and model are first trained, and the importance coefficients of each candidate service feature are obtained by utilizing the feature_importances_ attribute of the algorithm, as shown in Table 1.

[0048] Table 1 Importance Coefficients of Candidate Business Features

[0049] Based on the importance coefficients obtained by the algorithm, the feature corresponding to the largest coefficient, traffic, is selected as the optimal target business feature.

[0050] Furthermore, the LightGBM algorithm used in this embodiment for feature selection of business data is a gradient boosting algorithm based on decision trees, which has the advantage of being lightweight. In actual implementation, it can also be replaced with existing feature selection methods such as Pearson correlation coefficient, distance correlation coefficient, feature ranking based on learning models, and regularization models, depending on the team's algorithm selection preferences.

[0051] For example, if the feature selection model is a decision tree model, the importance coefficients of each candidate business feature are obtained by analyzing each candidate business feature based on the pre-trained feature selection model, including: The pre-trained feature selection model is used to segment each candidate business feature, and the total number of segmentations and the sum of segmentation gains for each candidate business feature are obtained. The importance coefficient is obtained based on the quotient of the total number of segmentations and the sum of segmentation gains.

[0052] Specifically, in decision tree models, the importance metrics include the total number of splits and the sum of split gains. These importance metrics determine the final importance coefficient. An example formula for determining the importance coefficient is as follows:

[0053]

[0054]

[0055] in, The total number of times each candidate business feature is segmented in the iteration tree. Let K be the sum of the segmentation gains resulting from the use of candidate business features in all decision trees for segmentation, and K be the K decision trees generated in K iterations.

[0056] For example, candidate business feature x1 appears on two trees, and the two trees split a total of 6 times. x1 appears 3 times, so the importance coefficient of x1 is 3 / 6, or 1 / 2. Candidate business feature x4 appears on one tree, and the tree splits 3 times. x4 appears once, so the importance coefficient of x4 is 1 / 3.

[0057] In one possible implementation, an exemplary method for determining the first audit result based on target business characteristics and business audit rules includes: Invoke the preset business audit rules; when the target business characteristics meet the conditions of the business audit rules, determine the first audit result to characterize the audit anomaly.

[0058] For example, business audit rules should include at least: When the target service characteristics are less than the first preset threshold for N consecutive months, the broadband service is determined to be an abnormal service.

[0059] Specifically, taking the target business characteristics as traffic characteristics as an example, the selected traffic characteristics are analyzed. By comparing the traffic characteristics of normal and abnormal business operations during audit verification, abnormal business operations often show traffic remaining at 0 for extended periods, while normal business operations show traffic that is generally not zero. In this embodiment, the business audit rule's judgment standard is that if the traffic is 0 for N consecutive months (1 < N ≤ 12), then the business is judged as abnormal, and the authenticity of the user account is questionable. Here, abnormal business indicates that the authenticity of the user account is abnormal, and N is a natural number greater than 1.

[0060] In one possible implementation, the method further includes: The feature selection model is tested using multiple candidate first thresholds and broadband service sample data to obtain the prediction precision and prediction recall corresponding to each candidate first threshold. From each prediction precision and prediction recall, the maximum prediction precision and maximum prediction recall are determined. The candidate first threshold corresponding to the maximum prediction precision and maximum prediction recall is used as the first preset threshold.

[0061] Specifically, the precision and recall of predictions are calculated using a threshold tuning method. The threshold is fine-tuned over a number of consecutive months with zero traffic to maximize precision without reducing recall, and a first preset threshold is determined. For each candidate value of n, this threshold is applied to the test set to predict the authenticity of user accounts, and the corresponding confusion matrix, precision, and recall are calculated.

[0062] Using broadband service sample data and manually verifying its authenticity under actual circumstances, a test set with positive and negative examples was established. The results of user account authenticity identification were compared and analyzed with the actual situation, as shown in Table 2, and then the prediction precision and recall were calculated.

[0063] Table 2 Comparison and Analysis of Authenticity Identification Results with Actual Situation

[0064] The number of consecutive months with zero traffic was determined based on a threshold tuning method. The threshold, N, ranged from 1 to N and then to 12. A comparison of the 12 thresholds (n values) is shown in Table 3 below.

[0065] Table 3. Precision and Recall corresponding to the first candidate threshold values.

[0066] Based on the calculated metrics, an N value was chosen as the final standard that balances high recall with maximum precision. Considering that N=2 might lead to high recall but low precision because a short time window could misclassify some normal fluctuations as anomalies, and that while N=4 improves precision, recall decreases, meaning more genuine anomalies are missed, N=3 takes into account both short-term normal fluctuations (such as holidays and maintenance periods) and is long enough to identify prolonged periods of inactivity. Therefore, N=3 provides the highest precision while maintaining a certain recall, demonstrating a good balance between the two. Analysis in this embodiment shows that N=3 is the optimal choice; that is, if traffic is zero for three consecutive months, the user account is considered a fraudulently activated / maintained broadband / dedicated line and subject to focused investigation.

[0067] To more comprehensively evaluate model performance, the F1-score was calculated for model performance with validation thresholds N of 2, 3, or 4. The F1-score is the harmonic mean of precision and recall, providing a balance between the two. The calculated F1-scores were approximately 0.339 for a first candidate threshold of 2, approximately 0.493 for a first candidate threshold of 3, and approximately 0.450 for a first candidate threshold of 4. The highest F1-score was achieved with a first candidate threshold of 3, indicating that the model performed best in balancing precision and recall at this threshold.

[0068] In this embodiment, the N value is taken as 1-12 for the month. In scenarios with more intensive detection needs, such as weekly detection, the range of N values ​​is larger. Grid search can be used to traverse all candidate N values ​​and record the model performance index corresponding to each N value. Alternatively, strategies such as random search or Bayesian optimization can be used to find the optimal n value in order to complete threshold tuning more efficiently.

[0069] In one possible implementation, an exemplary method for determining the second audit result based on broadband service field images and a pre-established service database includes: Based on the broadband service channel category in the broadband service data, the broadband service field images are classified to obtain field image classification results; for each field image classification result, a pre-established service database is used to perform similarity recognition on each broadband service field image to obtain similarity recognition results; based on the similarity recognition results and a second preset threshold, a second audit result is determined.

[0070] Specifically, this embodiment employs Paddle vision detection technology, using on-site photographs as targets for training and recognition. The labelimg tool is used to label the installation and maintenance images, enabling the identification of specific modules within the images, such as optical splitters, broadband lines, and optical modems. Mobile services require on-site personnel to take and upload photos of the broadband / leaseline service activation / maintenance process. Therefore, the service database contains image information corresponding to these service activities. This embodiment can obtain the required image data by accessing existing system resources without altering the workflow of on-site personnel.

[0071] Next, similarity comparison is performed using mature algorithms from the OpenCV vision library. Due to the massive number of images, they are first categorized and grouped according to broadband service channel types. Then, within each group, both difference hashing and RGB histograms are used for similarity identification. For materials suspected of rotation or deformation, the SIFT algorithm is used to extract invariant feature points for enhanced comparison to obtain the similarity identification result. For example, when the similarity identification result is greater than a second preset threshold, the second audit result is determined to indicate an audit anomaly; when the similarity identification result is less than or equal to the second preset threshold, the second audit result is determined to indicate a normal audit.

[0072] Auditing on-site images of broadband services can not only directly serve the troubleshooting and on-site confirmation of home broadband / dedicated line services, but also, through optimized algorithm logic and model configuration, can be flexibly extended to other operation and maintenance management processes, such as image comparison before and after construction, and equipment status monitoring. With the evolution of 5G and cloud computing technologies, this embodiment can also be integrated into a wider range of intelligent operation and maintenance systems. For example, at the site of home broadband equipment installation or dedicated line maintenance, by accurately identifying and comparing on-site images with standard construction images, the compliance and quality of the work can be quickly verified. How to accurately identify abnormal situations in broadband installation and maintenance services from a large number of users during auditing and verification?

[0073] In addition, it can also identify anomalies in images uploaded by staff through technical processing, such as retrieving image attributes to check if they have been edited using software like Photoshop, or directly scanning images to identify Photoshop traces. Specifically, it identifies the device type / model based on the interface layout characteristics of splitters, optical modems, etc., in the image. This can prevent false business operations from deliberately altering the shooting angle and environment of the on-site images, enabling targeted identification in the early stages of false business operations.

[0074] In one possible implementation, an exemplary method for fusing the first audit result and the second audit result to obtain the target audit result includes: When both the first and second audit results indicate an audit anomaly, the target audit result is determined to be an audit anomaly. When both the first and second audit results indicate an audit normality, it is determined whether the target business characteristics in the N+1 month are less than the first preset threshold. If yes, the target audit result is determined to be an audit anomaly; otherwise, the target audit result is determined to be an audit normality.

[0075] Specifically, the image recognition results and feature audit results are combined to determine the target audit result. As mentioned earlier, in the online data monitoring mentioned above, the method for judging false business activity in traffic monitoring is that if the traffic is 0 for N consecutive months (N=3), it is judged as an abnormal business activity, and the authenticity of the user account is questionable. However, according to the precision and recall tables corresponding to various threshold values, the difference between N=3 and N=4 is not significant enough to warrant one being identified as an abnormal business activity while the other should not.

[0076] Therefore, to improve the accuracy and efficiency of identifying business anomalies, after two consecutive months of suspected fraudulent business activity (i.e., after three months of zero traffic), a comprehensive judgment is made by comparing and identifying images from the broadband service site. If the second audit result indicates "anomaly," the user account is immediately determined to be fraudulent; if the second audit result indicates "normal," monitoring continues for another month before a final judgment is made. If the traffic remains zero during the continued monitoring month, the target audit result is determined to be an audit anomaly; otherwise, the target audit result is determined to be an audit normal. Furthermore, other types of data from the broadband service data can also be used to make the judgment.

[0077] In one possible implementation, the method further includes: Obtain the device type and / or service address from broadband service field images of multiple broadband services; perform cluster analysis based on device type and / or service address to obtain clustering results; when the clustering results indicate that the device type and / or service address are the same, determine that the authenticity of the broadband service field image is false.

[0078] Specifically, in this embodiment, not only is the authenticity of a single broadband service site image analyzed, but the address information and / or device model of broadband service site images from multiple broadband services can also be compared to accurately identify those who use the same address or device to commit batch fraud. Comparing the address information and device model from multiple broadband services can be achieved through conventional clustering analysis or hash value algorithms, or by combining machine learning models for comprehensive judgment. Clustering analysis methods include at least one of the following: K-means clustering, hierarchical clustering, DBSCAN (Density-Based Spatial Clustering of Applications with Noise), and hash value algorithms such as the Perceptual Hash Algorithm. By combining the clustering results of the above device types / models with the judgment criterion of "batch fraud with the same address or device," it is determined that when multiple broadband services have the same device type and / or the same service address, there may be a case of batch fraud, i.e., the authenticity of the broadband service site image is false. Here, a false authenticity indicates that the target audit result is an audit anomaly.

[0079] By introducing a method to identify device type / address information based on features such as interface layout in abnormal image recognition, the accuracy of on-site image recognition is improved, and it is also helpful to accurately locate cases of batch false reporting.

[0080] In one possible implementation, the method further includes: Obtain initial broadband service data; process the initial broadband service data to obtain broadband service data.

[0081] Specifically, the initial broadband service data related to service activation and maintenance for home broadband and dedicated line services is retrieved from the business system. This initial data includes: account number, activation time, service type, application channel, service item identifier code, status, month, city, phone number, data usage, usage duration, user status, and rate code. The data undergoes preliminary cleaning to remove useless input. This data processing includes at least one of the following: data cleaning, removal of abnormal data, and data population.

[0082] Furthermore, data processing specifically includes: handling missing data values, which involves deleting or filling in missing values ​​for a specific attribute in the dataset; data deletion, where records containing missing values ​​are deleted directly if the percentage of missing values ​​is very small (e.g., less than 3%); data filling, where numeric data is filled using the mean, median, or mode; categorical data is filled using the most common category or an unknown label such as "unknown"; for data marked as successfully activated home broadband / dedicated line services but without a matching network usage, the default network usage is set to 0, thereby expanding the scope of data for audit and verification; and removing outlier data, which involves deleting outlier values, which are unreasonable values ​​in the dataset.

[0083] Outliers are identified using a box plot method: the first quartile (Q1) and the third quartile (Q3) are calculated, and then IQR is defined as Q3 - Q1. Any data point below Q1 - 1.5 × IQR or above Q3 + 1.5 × IQR is considered an outlier.

[0084] Figure 2 Another flowchart for a broadband service auditing method provided in this disclosure. Figure 2 As shown, the method includes: S201: Conduct online monitoring of broadband service data to determine the first audit result.

[0085] Specifically, online monitoring of broadband service data involves ranking the importance coefficients of each candidate service feature extracted from the broadband service data, with the feature having the highest importance coefficient being the priority feature. This priority feature is then used as the evaluation criterion to determine whether a service is abnormal, thus obtaining the first audit result.

[0086] Figure 3 A flowchart for determining the first audit result provided in this disclosure. Figure 3 As shown, the method includes: S301: Obtain initial data for broadband services.

[0087] Specifically, the system retrieves data on the activation of home broadband and dedicated line services, installation and relocation costs, maintenance fees for home broadband and dedicated line services, and grid operation data.

[0088] S302: Preprocess the initial broadband service data to obtain broadband service data.

[0089] Specifically, the data undergoes preliminary cleaning to remove data that has not been successfully activated, expired data, and data with empty numbers in the network data. The data processing includes at least one of the following: data cleaning, removal of abnormal data, and data filling.

[0090] S303: Select the optimal target service feature from broadband service data.

[0091] Specifically, the model is invoked to select the optimal feature, and the LightGBM algorithm is used to calculate the feature importance coefficient. The feature with the largest coefficient is the optimal target business feature.

[0092] S304: First preset threshold optimization.

[0093] Specifically, the number of consecutive months with zero traffic is determined by threshold optimization methods to obtain the first preset threshold.

[0094] S305: Determine if a service has been fraudulently activated.

[0095] Specifically, if a user's traffic is zero for three consecutive months, the user's account is deemed to have abnormal business activity and fraudulent business activation, meaning the first audit result is abnormal.

[0096] S202: Identify on-site images of broadband services to determine the second audit result.

[0097] Specifically, images of broadband and leased line equipment are identified, a business database is built, images of the business site are retrieved from the database, and image similarity recognition is performed using the OpenCV vision library to obtain the second audit result.

[0098] S203: Determine the status of broadband services by combining the results of the first and second audits.

[0099] Specifically, combining the first and second audit results, when both are abnormal (i.e., abnormal traffic and abnormal image similarity), the broadband service status is determined to be that there is false reporting.

[0100] Figure 4 This is a schematic diagram of the structure of a broadband service auditing device provided in this disclosure. Figure 4 As shown, the device 400 includes: an extraction module 410, a first determination module 420, a second determination module 430, and a fusion module 440.

[0101] Extraction module 410 is used to extract features from broadband service data to obtain target service features, wherein the target service features characterize the data features of broadband service activation and maintenance; The first determining module 420 is used to determine the first audit result based on the target business characteristics and business audit rules; The second determining module 430 is used to determine the second audit result based on broadband service site images and a pre-established service database, wherein the service database consists of images of service site activation and maintenance during the execution of historical broadband services. The fusion module 440 is used to fuse the first audit result and the second audit result to obtain a target audit result, which represents the service status of the broadband service.

[0102] Optionally, the extraction module is used for: Feature extraction is performed on the broadband service data to obtain multiple candidate service features; The importance coefficients of each candidate business feature are obtained by analyzing each candidate business feature based on the pre-trained feature selection model. The importance coefficients of each candidate business feature are sorted to obtain the target business feature, which includes at least one of the candidate business features.

[0103] Optionally, if the feature selection model is a decision tree model, the extraction module: Based on the pre-trained feature selection model, each of the candidate business features is segmented to obtain the total number of segmentations and the sum of segmentation gains for each of the candidate business features; The importance coefficient is obtained based on the quotient of the total number of segmentations and the sum of the segmentation gains.

[0104] Optionally, the target service characteristics include at least traffic; the target service characteristics also include at least one of activation time, usage duration, user status, and rate encoding.

[0105] Optionally, the first determining module is used to: Invoke the preset business audit rules; When the target business characteristics meet the conditions of the business audit rules, the first audit result is determined to represent an audit anomaly; The business audit rules include at least the following: When the target service characteristics are less than the first preset threshold for N consecutive months, the broadband service is determined to be an abnormal service. The abnormal service indicates that the authenticity of the user account is abnormal, and N is a natural number greater than 1.

[0106] Optionally, the device further includes: The feature selection model was tested using multiple candidate first thresholds and broadband service sample data to obtain the prediction accuracy and prediction recall corresponding to each candidate first threshold. From the predicted precision and the predicted recall, determine the maximum predicted precision and the maximum predicted recall. The candidate first threshold corresponding to the maximum prediction precision and the maximum prediction recall is used as the first preset threshold.

[0107] Optionally, the second determining module: Based on the broadband service channel category in the broadband service data, the broadband service field images are classified to obtain the field image classification results. Based on the classification results of each of the aforementioned on-site images, the pre-established service database is used to perform similarity recognition on each of the aforementioned broadband service on-site images to obtain similarity recognition results; The second audit result is determined based on the similarity recognition result and the second preset threshold.

[0108] Optionally, the second determining module: When the similarity recognition result is greater than the second preset threshold, the second audit result is determined to indicate an audit anomaly; When the similarity recognition result is less than or equal to the second preset threshold, the second audit result is determined to indicate that the audit is normal.

[0109] Optionally, the fusion module is used for: When both the first audit result and the second audit result indicate an audit anomaly, the target audit result is determined to be an audit anomaly. When the first audit result indicates an audit anomaly and the second audit result indicates an audit normality, determine whether the target business characteristic in the N+1th month is less than a first preset threshold. If yes, the target audit result is determined to be an audit anomaly; otherwise, the target audit result is determined to be an audit normal.

[0110] Optionally, the device further includes: Obtain the device type and / or service address from the on-site images of multiple broadband services; Cluster analysis is performed based on the device type and / or the service address to obtain cluster results; When the clustering results indicate that the device types are the same and / or the service addresses are the same, the authenticity of the broadband service site image is determined to be false, and the false authenticity indicates that the target audit result is an audit anomaly.

[0111] Optionally, the device further includes: Obtain initial data for broadband services; The initial data of the broadband service is processed to obtain the broadband service data; the data processing includes at least one of the following: data cleaning, removal of abnormal data, and data filling.

[0112] This application also provides an electronic device for performing the above-described broadband service auditing method. Please refer to... Figure 5 It illustrates a schematic diagram of an electronic device provided by some embodiments of this application. For example... Figure 5 As shown, the electronic device 50 includes: a processor 500, a memory 501, a bus 502, and a communication interface 503. The processor 500, the communication interface 503, and the memory 501 are connected via the bus 502. The memory 501 stores a computer program that can run on the processor 500. When the processor 500 runs the computer program, it executes the broadband service auditing method provided in any of the foregoing embodiments of this application.

[0113] The memory 501 may include high-speed random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Communication between this device network element and at least one other network element is achieved through at least one communication interface 503 (which can be wired or wireless), such as the Internet, wide area network, local area network, metropolitan area network, etc.

[0114] Bus 502 can be an ISA bus, PCI bus, or EISA bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. The memory 501 is used to store programs. After receiving an execution instruction, the processor 500 executes the program. The broadband service auditing method disclosed in any of the foregoing embodiments of this application can be applied to the processor 500, or implemented by the processor 500.

[0115] The processor 500 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of the processor 500 or by instructions in software form. The processor 500 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules may reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in memory 501. The processor 500 reads the information in memory 501 and, in conjunction with its hardware, completes the steps of the above method.

[0116] The electronic device provided in this application embodiment and the broadband service auditing method provided in this application embodiment are based on the same inventive concept and have the same beneficial effects as the methods they adopt, operate or implement.

[0117] This application also provides a computer-readable storage medium corresponding to the broadband service auditing method provided in the foregoing embodiments. The computer-readable storage medium shown can be an optical disc, on which a computer program is stored. When the computer program is run by a processor, it executes the broadband service auditing method provided in any of the foregoing embodiments.

[0118] It should be noted that examples of the computer-readable storage medium may also include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other optical and magnetic storage media, which will not be elaborated here.

[0119] The computer-readable storage medium provided in the above embodiments of this application and the broadband service auditing method provided in the embodiments of this application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the applications stored therein.

[0120] This application also provides a computer program product 600, such as... Figure 6 As shown. This computer program product carries a computer program 601. The instructions included in the program code can be used to execute the steps of the broadband service auditing method described in the above method embodiments. For details, please refer to the above method embodiments, which will not be repeated here.

[0121] The aforementioned computer program product can be implemented through hardware, software, or a combination thereof. In one optional embodiment, the computer program product is specifically embodied in a computer storage medium; in another optional embodiment, the computer program product is specifically embodied in a software product, such as a software development kit (SDK), etc.

[0122] The basic principles of this disclosure have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, and effects mentioned in this disclosure are merely examples and not limitations, and should not be considered as essential features of each embodiment of this disclosure. Furthermore, the specific details disclosed above are for illustrative and facilitative purposes only, and are not limitations. These details do not limit the scope of this disclosure to the necessity of employing the aforementioned specific details for implementation.

[0123] The block diagrams of devices, apparatuses, devices, and systems disclosed herein are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, devices, and systems can be connected, arranged, and configured in any manner. Words such as “comprising,” “including,” “having,” etc., are open-ended terms meaning “including but not limited to,” and are used interchangeably with them. The terms “or” and “and” as used herein refer to the terms “and / or,” and are used interchangeably with them unless the context clearly indicates otherwise. The term “such as” as used herein refers to the phrase “such as but not limited to,” and is used interchangeably with it.

[0124] Additionally, as used herein, the "or" used in a list of items beginning with "at least one" indicates a separate list, such that a list of, for example, "at least one of A, B, or C" means A or B or C, or AB or AC or BC, or ABC (i.e., A and B and C). Furthermore, the word "exemplary" does not imply that the described example is preferred or better than other examples.

[0125] It should also be noted that in the systems and methods of this disclosure, the components or steps can be decomposed and / or recombined. These decompositions and / or recombinations should be considered as equivalent solutions to this disclosure.

[0126] Various changes, substitutions, and modifications can be made to the technology described herein without departing from the teachings defined by the appended claims. Furthermore, the scope of the claims of this disclosure is not limited to the specific aspects of the processes, machines, manufactures, events, means, methods, and actions described above. Currently existing or later-developed processes, machines, manufactures, events, means, methods, or actions that perform substantially the same function or achieve substantially the same result as the corresponding aspects described herein can be utilized. Therefore, the appended claims include such processes, machines, manufactures, events, means, methods, or actions within their scope.

[0127] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use this disclosure. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of this disclosure. Therefore, this disclosure is not intended to be limited to the aspects shown herein, but rather to be carried out within the widest scope consistent with the principles and novel features disclosed herein.

[0128] The above description has been given for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of this disclosure to the forms disclosed herein. Although numerous exemplary aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations therein.

Claims

1. A broadband service auditing method, characterized in that, include: Feature extraction is performed on broadband service data to obtain target service features, which characterize the data features of broadband service activation and maintenance; Based on the target business characteristics and business audit rules, the first audit result is determined; Based on on-site images of broadband services and a pre-established service database, a second audit result is determined, wherein the service database consists of images of the activation and maintenance of services at the service site during the execution of historical broadband services. The first audit result and the second audit result are merged to obtain the target audit result, which represents the service status of the broadband service.

2. The method according to claim 1, characterized in that, The process of extracting features from broadband service data to obtain target service features includes: Feature extraction is performed on the broadband service data to obtain multiple candidate service features; The importance coefficients of each candidate business feature are obtained by analyzing each candidate business feature based on the pre-trained feature selection model. The importance coefficients of each candidate business feature are sorted to obtain the target business feature, which includes at least one of the candidate business features.

3. The method according to claim 2, characterized in that, If the feature selection model is a decision tree model, the pre-trained feature selection model analyzes each of the candidate business features to obtain the importance coefficient of each candidate business feature, including: Based on the pre-trained feature selection model, each of the candidate business features is segmented to obtain the total number of segmentations and the sum of segmentation gains for each of the candidate business features; The importance coefficient is obtained based on the quotient of the total number of segmentations and the sum of the segmentation gains.

4. The method according to claim 2, characterized in that, The target service characteristics include at least traffic; the target service characteristics also include at least one of activation time, usage duration, user status, and rate encoding.

5. The method according to claim 1, characterized in that, The determination of the first audit result based on the target business characteristics and business audit rules includes: Invoke the preset business audit rules; When the target business characteristics meet the conditions of the business audit rules, the first audit result is determined to represent an audit anomaly; The business audit rules include at least the following: When the target service characteristics are less than the first preset threshold for N consecutive months, the broadband service is determined to be an abnormal service. The abnormal service indicates that the authenticity of the user account is abnormal, and N is a natural number greater than 1.

6. The method according to claim 5, characterized in that, The method further includes: The feature selection model was tested using multiple candidate first thresholds and broadband service sample data to obtain the prediction accuracy and prediction recall corresponding to each candidate first threshold. From the predicted precision and the predicted recall, determine the maximum predicted precision and the maximum predicted recall. The candidate first threshold corresponding to the maximum prediction precision and the maximum prediction recall is used as the first preset threshold.

7. The method according to claim 1, characterized in that, The determination of the second audit result based on broadband service field images and a pre-established service database includes: Based on the broadband service channel category in the broadband service data, the broadband service field images are classified to obtain the field image classification results. Based on the classification results of each of the aforementioned on-site images, the pre-established service database is used to perform similarity recognition on each of the aforementioned broadband service on-site images to obtain similarity recognition results; The second audit result is determined based on the similarity recognition result and the second preset threshold.

8. The method according to claim 7, characterized in that, The determination of the second audit result based on the similarity recognition result and the second preset threshold includes: When the similarity recognition result is greater than the second preset threshold, the second audit result is determined to indicate an audit anomaly; When the similarity recognition result is less than or equal to the second preset threshold, the second audit result is determined to indicate that the audit is normal.

9. The method according to claim 1, characterized in that, The process of fusing the first audit result and the second audit result to obtain the target audit result includes: When both the first audit result and the second audit result indicate an audit anomaly, the target audit result is determined to be an audit anomaly. When the first audit result indicates an audit anomaly and the second audit result indicates an audit normality, determine whether the target business characteristic in the N+1th month is less than a first preset threshold. If yes, the target audit result is determined to be an audit anomaly; otherwise, the target audit result is determined to be an audit normal.

10. The method according to claim 1, characterized in that, The method further includes: Obtain the device type and / or service address from the on-site images of multiple broadband services; Cluster analysis is performed based on the device type and / or the service address to obtain cluster results; When the clustering results indicate that the device types are the same and / or the service addresses are the same, the authenticity of the broadband service site image is determined to be false, and the false authenticity indicates that the target audit result is an audit anomaly.

11. The method according to claim 1, characterized in that, The method further includes: Obtain initial data for broadband services; The initial data of the broadband service is processed to obtain the broadband service data; the data processing includes at least one of the following: data cleaning, removal of abnormal data, and data filling.

12. A broadband service auditing device, characterized in that, include: The extraction module is used to extract features from broadband service data to obtain target service features, which represent the data features of broadband service activation and maintenance. The first determining module is used to determine the first audit result based on the target business characteristics and business audit rules; The second determining module is used to determine the second audit result based on broadband service site images and a pre-established service database, wherein the service database consists of images of service site activation and maintenance during the execution of historical broadband services. The fusion module is used to fuse the first audit result and the second audit result to obtain the target audit result, which represents the service status of the broadband service.

13. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The processor executes the computer program to implement the method as described in any one of claims 1-11.

14. A computer-readable storage medium having a computer program stored thereon, characterized in that, The program is executed by a processor to implement the method as described in any one of claims 1-11.

15. A computer program product, characterized in that, Includes computer-readable code, or a non-volatile computer-readable storage medium carrying computer-readable code, wherein when the computer-readable code is run in a processor of an electronic device, the processor in the electronic device performs the method as described in any one of claims 1-11.

Images