A communication link abnormal switching processing method and system

Abstract

The application relates to the technical field of bank collection communication processing, and discloses a communication link abnormal switching processing method and system. A monitoring module of the system acquires real-time operation data of multiple communication links, including connection state data and quality index data; an analysis module acquires abnormal evaluation parameters of each link based on the real-time operation data, so as to represent the abnormality degree of the link connection state; a positioning module determines an abnormal communication area needing to perform a switching operation according to the abnormal evaluation parameters; a decision module determines a main communication link, namely a first target link, needing to be preferentially processed based on the abnormal communication area; and a compensation module performs transmission compensation adjustment on the first target link, so that the communication quality index of the abnormal communication area reaches a preset standard threshold. The system can timely process communication link abnormalities, guarantee the stability of the communication link in the bank arrears collection business, and ensure the smooth development of the collection business.

Description

Technical Field

[0001] This invention relates to the field of bank debt collection communication processing technology, specifically a method and system for handling abnormal switching of communication links. Background Technology

[0002] In the process of debt collection by banks, communication links are a crucial foundation for ensuring the smooth progress of the business. Collection operations often require maintaining real-time contact with debtors through multiple communication links, including voice calls, SMS notifications, and online message pushes. Different communication methods rely on different communication links, and these links together constitute the communication network for the debt collection operation.

[0003] As the scale of debt collection operations continues to expand and the user base becomes increasingly geographically dispersed, the operating environment of communication links becomes more complex. In actual operation, communication links may be affected by various factors, leading to anomalies such as network signal fluctuations, base station equipment failures, and line transmission interference. If these anomalies are not detected and addressed promptly, they will directly impact the efficiency of debt collection operations, and may even cause communication interruptions with users, preventing timely delivery of collection notices and consequently affecting the progress of the bank's debt recovery.

[0004] In managing communication links during bank debt collection, existing technologies mostly only allow for simple monitoring of link connection status, meaning they can only determine whether the link is connected. They lack comprehensive and real-time means to acquire link quality metrics such as call completion rate, SMS delivery rate, and message latency. This means that existing monitoring methods cannot detect minor link anomalies in a timely manner, even when communication quality has significantly deteriorated before a complete disconnection, thus missing the optimal time for intervention.

[0005] Even if some technologies can obtain a small amount of quality indicator data, they are still significantly insufficient in assessing the degree of link anomalies. Existing assessment methods often rely on a single indicator for judgment, lacking comprehensive anomaly assessment parameters. This makes it difficult to accurately characterize the degree of anomalies in the communication link's connection status, preventing staff from clearly understanding the severity of the link anomalies and thus hindering the development of effective response strategies.

[0006] Existing technologies have significant shortcomings in identifying abnormal communication areas. Because they cannot accurately grasp the operational status and degree of abnormality of each part of the link, they can often only make an overall judgment on the abnormal link, rather than precisely locating the specific area within the link that requires a switching operation. This means that when handling anomalies, it may be necessary to switch the entire link, which not only increases the complexity and cost of the operation but may also affect the communication stability of other normal areas.

[0007] When selecting communication links for priority handling, existing technologies lack a scientifically sound decision-making basis. Selection is often random or follows a fixed order, failing to determine the primary communication links to prioritize based on their importance in debt collection operations and the urgency of abnormal situations. This can lead to critical link anomalies not being addressed promptly, further exacerbating the impact of communication problems on debt collection operations. Furthermore, when handling abnormal links, existing technologies mostly resolve the issue by switching links, lacking mechanisms for transmission compensation and adjustment. Even after link switching, the new link may still fail to meet the communication quality requirements of debt collection operations due to various factors, still negatively impacting the debt collection process. Summary of the Invention

[0008] The purpose of this invention is to provide a method and system for handling abnormal switching of communication links, so as to solve the problems mentioned in the background art.

[0009] To achieve the above objectives, the present invention provides a communication link anomaly handover handling system, the system comprising:

[0010] The monitoring module is used to acquire real-time operational data of multiple communication links in the bank's debt collection business; the real-time operational data includes connection status data and quality indicator data of each communication link.

[0011] The analysis module is used to obtain anomaly evaluation parameters for each communication link based on the real-time operating data; the anomaly evaluation parameters are used to characterize the degree of anomaly in the connection status of the corresponding communication link.

[0012] The positioning module is used to determine the abnormal communication area based on various anomaly assessment parameters; the abnormal communication area is the area in the communication link that requires a handover operation.

[0013] The decision module is used to determine a first target link based on the abnormal communication area; the first target link is the primary communication link that needs to be prioritized among all communication links.

[0014] The compensation module is used to perform transmission compensation adjustment on the first target link so that the communication quality index of the abnormal communication area reaches a preset standard threshold.

[0015] Preferably, the analysis module includes:

[0016] The parameter acquisition unit is used to select a second target link from each communication link; the second target link is any one of the communication links.

[0017] A timing processing unit is used to obtain timing quality index data of the second target link based on the real-time running data;

[0018] The difference calculation unit is used to obtain multiple first-level differences based on the time-series quality index data; the first-level difference is the difference between two adjacent time-series quality index data.

[0019] An evaluation generation unit is used to obtain the first evaluation factor based on each first-level difference;

[0020] The parameter output unit is used to obtain the anomaly evaluation parameters of the second target link based on the first evaluation factor.

[0021] Preferably, the evaluation generation unit is specifically used for:

[0022] A first average value is obtained based on each first-level difference; the first average value is the average of the absolute values ​​of each first-level difference.

[0023] The first average value is denoted as the first evaluation factor.

[0024] Preferably, the parameter output unit includes:

[0025] The secondary computation subunit is used to obtain multiple second-level differences based on each first-level difference; the second-level difference is the difference between two temporally adjacent first-level differences;

[0026] The parameter generation subunit is used to obtain the second evaluation factor based on each second-level difference;

[0027] The weighted processing subunit is used to weight the first evaluation factor and the second evaluation factor through a preset weight coefficient to obtain the abnormal evaluation parameters of the second target link.

[0028] Preferably, the positioning module includes:

[0029] The sorting unit is used to sort all anomaly evaluation parameters from largest to smallest.

[0030] The filtering unit is used to select the communication links corresponding to the top-ranked preset number of abnormal evaluation parameters in the sorting results as the third target links.

[0031] The prediction unit is used to input the real-time operating data of each third target link into the prediction model;

[0032] The region determination unit is used to obtain the abnormal prediction values ​​of each location in the communication network based on the prediction model, and to determine the abnormal communication region based on each abnormal prediction value.

[0033] Preferably, the prediction unit includes:

[0034] The link selection subunit is used to select a fourth target link and a fifth target link based on each third target link; the fourth target link and the fifth target link are any two adjacent communication links among the third target links;

[0035] The point generation subunit is used to obtain a preset number of predicted points between the fourth target link and the fifth target link;

[0036] The prediction execution subunit is used to obtain the abnormal prediction values ​​of each prediction point based on the prediction model.

[0037] Preferably, the compensation module includes:

[0038] A quality detection unit is used to obtain the maximum quality deviation value of the abnormal communication area;

[0039] The judgment unit is used to not initiate compensation adjustment when the maximum quality deviation value is less than or equal to the preset standard threshold, and to initiate compensation adjustment when the maximum quality deviation value is greater than the preset standard threshold.

[0040] The adjustment calculation unit is used to obtain the quality deviation difference based on the maximum quality deviation value and a preset standard threshold.

[0041] The compensation execution unit is used to perform transmission compensation adjustment on the first target link based on the quality deviation difference.

[0042] Preferably, the system further includes:

[0043] The optimization tiering module is used to determine the tiering optimization strategy based on bank customer type information; the bank customer type information includes high-risk customer type information, medium-risk customer type information, and low-risk customer type information;

[0044] The strategy execution module is used to perform multi-level optimization control on the transmission compensation adjustment process of the compensation module based on the hierarchical optimization strategy.

[0045] Preferably, the optimization and grading module includes:

[0046] The duration monitoring unit is used to obtain the duration of communication link abnormalities;

[0047] The strategy selection unit is used to adopt a first-level optimization strategy when the duration of the communication link abnormality is less than the preset duration, and to adopt a second-level optimization strategy when the duration of the communication link abnormality is greater than or equal to the preset duration.

[0048] The parameter adjustment unit is used to adjust the control parameters of the compensation module based on different levels of optimization strategies.

[0049] The strategy execution module includes:

[0050] The correlation analysis unit is used to obtain correlation parameters and non-correlation parameters during the transmission compensation adjustment process; the correlation parameters include bandwidth allocation parameters and quality compensation parameters.

[0051] The weight allocation unit is used to determine the adjustment weights corresponding to the associated and unassociated parameters based on the bank customer type information.

[0052] The execution control unit is used to perform multi-level optimization control based on the adjustment weights corresponding to the associated and non-associated parameters, respectively.

[0053] Preferably, the present invention also includes a communication link abnormal handover processing method, which includes all the modules and method flow of the above-mentioned communication link abnormal handover processing system.

[0054] Compared with the prior art, the beneficial effects of the present invention are:

[0055] From the monitoring module's perspective, it can comprehensively acquire real-time operational data from multiple communication links, including not only connection status data for each link but also quality indicator data. This comprehensive data acquisition method breaks through the limitations of existing technologies that can only monitor link connection status, allowing staff to grasp the overall operational status of communication links in real time. Whether a link is disconnected or changes in quality indicators such as call connection rate and SMS delivery rate can be detected promptly. This enables the timely detection of various link anomalies, even minor anomalies that have not yet caused a link disconnection but have already led to a decline in quality, avoiding missing the best opportunity for anomaly handling due to incomplete monitoring, and laying a solid data foundation for subsequent anomaly analysis and processing.

[0056] The analysis module acquires anomaly assessment parameters for each communication link based on real-time operational data. These parameters accurately characterize the degree of anomaly in the connection status of the corresponding communication link. Compared to existing technologies that rely on a single indicator for fuzzy assessment, these precise assessment parameters allow staff to clearly and accurately understand the severity of anomalies in each communication link, such as whether it is a slight quality fluctuation or a severe anomaly approaching interruption. This provides an accurate basis for subsequent anomaly handling and avoids situations where inappropriate handling measures are taken due to unclear assessment of the degree of anomaly.

[0057] The positioning module determines the abnormal communication area based on various anomaly assessment parameters, precisely pinpointing the specific area in the communication link that requires a switching operation. This function overcomes the shortcomings of existing technologies that cannot accurately locate abnormal areas and can only process the entire link. Staff can perform switching operations on specific abnormal areas without adjusting the entire communication link, reducing operational complexity and unnecessary costs. It also avoids interference with normal communication areas in the link, ensuring the stability of communication in other areas and guaranteeing that collection services can continue partially normally while handling abnormal links.

[0058] The decision-making module identifies the primary target link based on the abnormal communication area, thus clarifying the main communication link that needs priority processing among all communication links. In debt collection, different communication links play different roles and have different levels of importance. Some links may be mainly used for real-time communication with key debtors, and their abnormalities have a significant impact on the business; others may be used to send routine notifications, with relatively less impact. This module can scientifically and rationally determine the primary communication link to be prioritized based on the location of the abnormal area and the link's importance in the business, ensuring that abnormalities on critical links are resolved first, minimizing the impact of abnormal links on the core aspects of the debt collection business, and avoiding the situation in existing technologies where random selection or fixed-order selection of links for processing leads to critical issues not being resolved in a timely manner.

[0059] The compensation module performs transmission compensation adjustments on the primary target link to bring the communication quality indicators of the abnormal communication area up to a preset standard threshold. This module fills the gap in existing technologies that only solve problems by switching links, lacking the ability to compensate and adjust the link. After determining the primary link for priority handling, transmission compensation adjustments can specifically improve the communication quality of the abnormal area. Even if it is not possible to immediately switch to a better link, compensation adjustments can ensure that the communication quality of the link meets the requirements of collection services. Simultaneously, this compensation adjustment also enhances the link's anti-interference capability and reduces the probability of subsequent abnormal situations.

[0060] The various modules work collaboratively to form a complete communication link anomaly handling system. From data monitoring, anomaly assessment, and precise location, to priority decision-making and compensation adjustments, each link is closely connected, ensuring that communication link anomalies can be handled quickly and accurately, guaranteeing the stable operation of communication links in the bank's debt collection business. Whether it's timely intervention when minor link anomalies occur or efficient handling of serious anomalies, this system plays a vital role in reducing the interruption or efficiency reduction of collection operations caused by communication link problems, ensuring that collection notices are delivered to users in a timely and accurate manner, and helping banks to carry out debt collection work more smoothly. Attached Figure Description

[0061] Figure 1 This is a timing diagram of the communication link anomaly switching processing system described in this invention;

[0062] Figure 2 A flowchart for the process of obtaining anomaly assessment parameters for the analysis module;

[0063] Figure 3 A flowchart for the process of generating the first evaluation factor for the evaluation unit. Detailed Implementation

[0064] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0065] Please see Figure 1 The present invention provides a method and system for handling abnormal handover of communication links. The system includes a monitoring module, an analysis module, a positioning module, a decision-making module, and a compensation module.

[0066] The monitoring module is responsible for acquiring real-time operational data from multiple communication links in the bank's debt collection process. This real-time operational data includes connection status data and quality indicator data for each communication link. Connection status data reflects the connectivity of the link, such as connection success rate and disconnection frequency; quality indicator data includes parameters such as bandwidth, latency, and packet loss rate. The analysis module calculates anomaly assessment parameters for each communication link based on the real-time operational data. These parameters quantify the degree of anomaly in the connection status of the corresponding communication link. The location module identifies abnormal communication areas based on the anomaly assessment parameters. Abnormal communication areas refer to the parts of the communication link that require switching operations. The decision-making module determines the primary target link based on the abnormal communication areas. The primary target link is the main communication link that needs to be prioritized for handling. The compensation module performs transmission compensation adjustments on the primary target link, adjusting transmission parameters to bring the communication quality indicators of the abnormal communication area to a preset standard threshold, thereby restoring normal communication.

[0067] Example 1: See Figure 2In the analysis module of the communication link anomaly handover handling system, the parameter acquisition unit dynamically selects one link from all communication links as the second target link. The selection process follows a polling mechanism or a priority rule based on link load status to ensure comprehensive evaluation coverage without over-concentrating resources on a single link. The time-series processing unit then obtains historical and real-time quality indicator data for the second target link from the monitoring module. This data is sorted by timestamp to form a continuous time-series sequence, containing key indicators such as bandwidth utilization, transmission delay, and packet loss rate. The data acquisition frequency is dynamically adjusted according to network load to balance accuracy and system overhead. The difference calculation unit scans the time-series sequence point by point, calculates the differences in indicator values ​​between adjacent time points, and generates a first-level difference set. This calculation process uses real-time streaming processing to minimize latency and ensure that the timeliness of the anomaly assessment keeps pace with changes in network status.

[0068] The evaluation generation unit performs statistical analysis on the first-level difference set. First, it takes the absolute value of all differences to eliminate the positive and negative offsetting effects. Then, it calculates the arithmetic mean of these absolute values ​​as the first evaluation factor, which reflects the overall fluctuation intensity of the link quality index. The parameter output unit further introduces a secondary calculation subunit to perform secondary differencing on the first-level difference sequence to generate a second-level difference set, thereby capturing the fluctuation characteristics of the quality index change rate. The parameter generation subunit then calculates the average of the absolute values ​​of the second-level difference set as the second evaluation factor, which is used to quantify the accelerating trend of link instability. The weighted processing subunit inputs the first and second evaluation factors into a preset weight calculation model. This model dynamically allocates the contribution ratio of the two types of factors according to the link type and application scenario, and finally outputs a comprehensive anomaly evaluation parameter as a quantitative index of the anomaly degree of the second target link. The entire analysis process adopts a pipelined architecture to achieve parallel processing. The anomaly evaluation parameter generation cycle for each link is controlled within milliseconds to ensure that the system can quickly respond to changes in network status. All intermediate calculation results are temporarily stored in a distributed cache for other modules to call, and evaluation logs are generated for subsequent auditing and model optimization. This implementation method effectively distinguishes between random fluctuations and real abnormal trends through multi-level differential analysis of link quality indicators, providing highly reliable input data for subsequent positioning modules.

[0069] Example 2: See Figure 3The calculation process is based on statistical analysis of the first-level difference set. First, the absolute values ​​of all differences in the sequence are summed, and then divided by the total number of differences to obtain the first average value. This average value serves as the first evaluation factor, directly reflecting the average intensity of link quality fluctuations. The parameter output unit then initiates the processing flow. The secondary calculation subunit receives the first-level difference sequence and performs a second difference calculation, that is, it calculates the change between adjacent first-level differences to generate the second-level difference set. This step is used to capture the fluctuations in the rate of change of quality indicators. The parameter generation subunit applies a similar processing method to the second-level difference set, calculating the arithmetic mean of its absolute values ​​as the second evaluation factor. This factor quantifies the acceleration characteristics of link instability changes.

[0070] The weighted processing subunit employs a dynamic weight allocation mechanism. The system automatically adjusts the weight coefficients of the first and second evaluation factors based on the real-time load status and historical performance data of the communication link. These weight coefficients are determined through a pre-configured lookup table or a real-time calculation model. The weighted calculation process uses a linear combination method, multiplying the two evaluation factors by their corresponding weights and then summing them to obtain a comprehensive anomaly assessment parameter. This parameter integrates the instantaneous fluctuations and trends in link quality. The entire calculation process is completed in real-time in memory, with intermediate results temporarily stored in a circular buffer for quick access by subsequent modules. Simultaneously, timestamp logs generated by all calculation steps are written to an audit database for subsequent analysis. Considering the balance between computational efficiency and resource consumption, the system uses a sliding window mechanism to process time-series data. The window size is dynamically adjusted according to the network environment to ensure the real-time performance and accuracy of the assessment. For outlier handling, the algorithm uses a mean-pruning method to remove significant outliers before calculating the average, thereby improving the robustness of the evaluation factors. The determination of the weighting coefficients depends on the link type and historical performance data. For high-load links, the second evaluation factor will be given a higher weight to better capture its changing trend, while for links with higher stability, more attention will be paid to the instantaneous fluctuations reflected by the first evaluation factor.

[0071] The system also incorporates an anomaly handling mechanism during computation. When input data is insufficient or invalid values ​​exist, it automatically switches to a degraded processing mode, using the most recent valid calculation result or default value to continue execution and ensure system continuity. All computation modules employ a parallel pipeline architecture, supporting simultaneous computation across multiple links without resource conflicts. Computational resource allocation is dynamically adjusted based on link priority. The final anomaly assessment parameters are standardized and mapped to a unified numerical range for easy comparison and use by subsequent modules. Standardized parameters also include confidence indicators for decision-making modules. A balance between computational efficiency and result reliability is emphasized, employing multi-level differential and weighted fusion methods to comprehensively assess link anomalies, providing the system with accurate and timely anomaly quantification indicators. All intermediate data and final results generated during computation are timestamped and version-identified, ensuring data traceability and consistency. The system periodically self-checks the health status of computation modules and generates operational status reports for maintenance personnel to monitor.

[0072] Taking the actual operation of a bank's debt collection business as an example, a communication link connecting the Beijing data center and the Shanghai branch's collection system experienced intermittent quality fluctuations. The monitoring module collected a latency index data sequence for this link at 10 consecutive time points: [45,52,63,58,71,65,84,79,96,112] (unit: milliseconds). The parameter acquisition unit marked this link as the second target link, and the timing processing unit organized these data into a timing sequence according to the collection time and sent it to the difference calculation unit. The difference calculation unit began to calculate the latency difference between adjacent time points, generating the first-level difference sequence: [7,11,-5,13,-6,19,-5,17,16]. These values ​​reflect the change in latency index within adjacent sampling intervals. The evaluation generation unit preprocesses the sequence, taking the absolute values ​​of all differences to obtain a new sequence: [7,11,5,13,6,19,5,17,16]. The arithmetic mean of this absolute value sequence is calculated as the first evaluation factor, resulting in (7+11+5+13+6+19+5+17+16) / 9≈11.11 milliseconds. The secondary calculation unit of the parameter output unit performs a second difference calculation on the first-level difference sequence, obtaining the second-level difference sequence: [4,-16,18,-19,25,-24,22,-1]. These values ​​reflect the changes in the rate of change of delay. The parameter generation unit calculates the average of the absolute values ​​of this sequence as the second evaluation factor, resulting in (4+16+18+19+25+24+22+1) / 8≈16.12 milliseconds. The weighted processing subunit employs a dynamic weight allocation mechanism. Based on the service type (collection voice call) and historical stability data of the link, a weight of 0.6 is assigned to the first evaluation factor, and a weight of 0.4 is assigned to the second evaluation factor. The weighted calculation process is: 11.11 × 0.6 + 16.12 × 0.4 ≈ 13.11 milliseconds. This weighted result is output as the anomaly assessment parameter for the link.

[0073] The system simultaneously records auxiliary data during the calculation process, including timestamps of the original data sequence, generation time of the difference sequence, and metadata such as the basis for weight determination. All of this data is version-marked and stored in a distributed database. When new delay data is collected at subsequent time points, the system uses a sliding window mechanism to update the time series, removing the oldest data point and adding the latest data, and re-executing the entire calculation process to update the anomaly assessment parameters. Throughout the processing, the system uses a mean-pruning method for outlier values. When significant outliers appear in the difference sequence (such as exceeding three standard deviations), these values ​​are automatically excluded before calculating the average. The calculation module adopts a parallel pipeline architecture, supporting the simultaneous execution of the same analysis process on multiple links. Computational resources are dynamically allocated according to link priority to ensure that high-priority collection business links are processed in a timely manner. The final generated anomaly assessment parameters, along with the confidence index, are transmitted to the location module. A parameter value of 13.11 milliseconds indicates that the delay fluctuation of this link is relatively high, requiring further attention and processing by the system. Intermediate results and final parameters of all calculation steps are recorded in the audit log for operation and maintenance personnel to query and analyze. The system automatically generates a performance report of the calculation module every week, and statistically analyzes the accuracy of parameter calculation and processing timeliness of each link.

[0074] Example 3: The sorting unit first receives a set of anomaly assessment parameters from the analysis module. These parameters correspond to the quantified anomaly levels of each communication link. The sorting unit uses a quicksort or heapsort algorithm to sort all parameters in descending order. The sorting process is based on numerical value and considers timestamps to ensure that the newest data is processed first. The sorted results are stored in a dynamic array for use by subsequent units. The filtering unit extracts the top K anomaly assessment parameters from the sorted array, where K is a preset number that is dynamically adjusted according to system configuration and network scale, usually based on historical anomaly patterns or operation and maintenance strategies. The filtering unit then maps these parameters to corresponding communication link identifiers to determine the third target link set, which represents the part of the current network with the highest anomaly level.

[0075] The prediction unit receives real-time operational data from the third target link, including connection status data and quality indicator data. The data is organized in time-series format. The prediction unit inputs this data into a pre-trained prediction model. The prediction model employs a machine learning-based regression algorithm, such as random forest or gradient boosting tree. During model training, historical anomaly data and network topology information are used to learn anomaly propagation patterns. The prediction model outputs anomaly prediction values ​​for each location in the communication network. These values ​​represent the probability or intensity of an anomaly occurring at that location. A key formula is used during model calculation to generate the prediction values:

[0076]

[0077] in: Indicates abnormal predicted values. These are the weight coefficients of the i-th input feature, obtained through model training. It is a feature function that processes the i-th feature data. At time t, It is the total number of features. It is the time decay factor. This is the time gradient term, reflecting the trend of abnormal changes. Feature data includes bandwidth utilization, latency jitter, and packet loss rate, etc. The time gradient term is calculated based on the difference of historical data. The meanings of the characters in the formula are as follows: This represents the final anomaly prediction value, used to quantify the anomaly risk of network locations; The contribution weight representing each feature is obtained through optimization using training data; It is a non-linear transformation function, such as sigmoid or ReLU, used to process eigenvalues; It is the specific value of the i-th feature, such as real-time bandwidth data; It is the current timestamp; It is the feature dimension, which is usually a fixed value; Control the impact of time factors and adjust according to network stability; It is the time derivative term, calculated as the rate of change of outliers within the most recent time window.

[0078] The region determination unit receives a set of anomalous predicted values ​​output by the prediction model. These values ​​correspond to physical or logical locations within the network. The unit uses clustering algorithms such as DBSCAN or K-means to group these values ​​to identify high-anomaly regions. The clustering process is based on the similarity and spatial proximity of the predicted values, and the threshold setting is dynamically optimized according to the network topology and historical anomaly data. The unit then determines the anomalous communication regions based on the clustering results, typically defined as continuous regions where the predicted values ​​exceed a preset threshold. The region boundaries are refined using interpolation or contour detection methods. Finally, the unit outputs region coordinates and severity indicators for use by the decision-making module. The entire localization process runs in real time, with data streams processed asynchronously through a message queue to ensure low latency. The module also integrates a monitoring mechanism to continuously evaluate and adjust the performance of the prediction model, such as updating model parameters through online learning to adapt to network changes.

[0079] The computational resources for the sorting and filtering units are allocated based on a load balancing strategy, prioritizing data from high-priority links. The sorting algorithm selection considers time complexity, typically using an O(nlogn) method to handle large-scale data. The prediction unit's model inference employs a distributed computing framework, such as Spark or TensorFlow Serving, to process prediction requests from multiple links in parallel. The model input data undergoes standardization and cleaning to remove noise and outliers, improving prediction accuracy. The clustering algorithm parameters for the region determination unit, such as distance thresholds and minimum sample size, are optimized through grid search or Bayesian methods to ensure accurate region identification. Outputs include geographic information or logical identifiers of outlier regions, along with confidence scores for subsequent modules. The system also handles edge cases; for example, when the number of third-target links is insufficient, the filtering unit automatically adjusts the K value or uses a set of backup links. The prediction model falls back to simpler heuristics, such as nearest-neighbor-based prediction, when there is insufficient data. All intermediate data and production logs are recorded in persistent storage for auditing and troubleshooting. Communication between modules uses REST APIs or message buses to ensure decoupling and scalability.

[0080] Taking the network operation and maintenance center for bank debt collection as an example, the system detected abnormalities in 15 major communication links nationwide. The analysis module generated a set of abnormality assessment parameters for each link: [23.6,18.9,35.2,27.4,41.8,19.3,32.7,38.5,22.1,29.8,36.4,24.7,40.1,31.5,26.9] (the larger the value, the higher the degree of abnormality). The sorting unit used the quicksort algorithm to sort these parameters in descending order, resulting in an ordered sequence: [41.8,40.1,38.5,36.4,35.2,32.7,31.5,29.8,27.4,26.9,24.7,23.6,22.1,19.3,18.9]. The screening unit selects the communication links corresponding to the top 5 abnormal evaluation parameters as the third target links based on the system's preset threshold parameters. These links are connected to the core nodes of Beijing-Shanghai, Guangzhou-Shenzhen, Chengdu-Chongqing, Wuhan-Zhengzhou, and Xi'an-Lanzhou, respectively.

[0081] The prediction unit immediately initiates the processing flow, extracting operational data for these five links from the real-time database. This includes time-series data such as bandwidth utilization, transmission latency, and packet loss rate over the past 30 minutes, as well as network topology information. The prediction model employs a deep learning architecture based on time-series analysis. Input data, after standardized preprocessing, is fed into the model inference engine. The model first embeds the operational characteristics of each link, then uses a multi-head attention mechanism to capture the spatiotemporal correlations between different links, and finally outputs the abnormal prediction values ​​for each grid point in the network (with 50 km × 50 km as the basic unit). After receiving these prediction values, the region determination unit uses a density clustering algorithm to identify continuous regions where abnormal prediction values ​​exceed the threshold of 0.85. The system identified a significant abnormal communication region in East China, covering network nodes in cities such as Shanghai, Hangzhou, and Nanjing, with abnormal prediction values ​​fluctuating between 0.87 and 0.92. The region determination unit maps the boundary coordinates of the region to a specific list of network devices, including 3 core routers, 12 aggregation switches and 28 access devices, and generates a region anomaly intensity map for the decision-making module to use.

[0082] Throughout the entire process, the computation time for the sorting and filtering units is controlled within 200 milliseconds, the inference time of the prediction model is approximately 1.2 seconds, and the execution time of the clustering algorithm for the region determination unit is 800 milliseconds. The system re-executes the complete localization process every 5 minutes to ensure the timeliness of abnormal region identification. All intermediate results, including the sorted parameter sequence, the third target link list, the predicted value matrix, and the region boundary coordinates, are written to the distributed database in real time and pushed to the operation and maintenance monitoring dashboard. Operation and maintenance personnel can view the evolution trend of abnormal regions through a visual interface. The system retains the location history data for the most recent 24 hours and supports viewing the abnormal distribution at any point in time using a time slider. When a newly detected abnormal region deviates significantly from the historical pattern, the system automatically triggers an early warning mechanism to notify network engineers for manual verification. This implementation method ensures the rapid and accurate localization of abnormal regions in the bank's debt collection business communication network, providing a reliable spatial basis for subsequent switching decisions and compensation adjustments.

[0083] Example 4: The prediction unit identifies adjacent communication links from the third target link set through the link selection subunit. For example, in the network topology of a bank's debt collection business, the link from the Beijing data center to the Shanghai branch is selected as the fourth target link, and the link from the Shanghai branch to the Hangzhou sub-branch is selected as the fifth target link. These two links are directly adjacent in physical routing and share a core network node. The point generation subunit generates several virtual monitoring points on the network path between the two links. These monitoring points are evenly distributed across the routing hop count, and each monitoring point corresponds to an exit interface or logical transmission segment of a network device. The system assigns a unique identifier to each monitoring point and records its network address information. The prediction execution subunit inputs the network status data of each monitoring point into a pre-trained anomaly prediction model. The model is based on a deep neural network architecture, and the input features include real-time throughput, routing hop count, and historical anomaly records. The output is the anomaly probability value of each monitoring point. The quality detection unit of the compensation module collects quality indicator data of all monitoring points in the abnormal communication area, including parameters such as transmission delay, packet loss rate, and bandwidth utilization. It calculates the absolute deviation of each monitoring point's indicator from the preset standard value and selects the largest deviation value as the basis for regional quality assessment.

[0084] Table 1: Monitoring Data of Quality Indicators in Abnormal Communication Areas

[0085] Monitoring point ID Network address Delay deviation (ms) Packet loss rate deviation (%) Bandwidth utilization deviation (%) MP001 192.168.1.1 125 8.2 25 MP002 192.168.1.2 83 5.1 18 MP003 192.168.1.3 156 9.8 31 MP004 192.168.1.4 92 6.4 22 MP005 192.168.1.5 178 11.5 37

[0086] The judgment unit compares the monitored maximum quality deviation value with a preset standard threshold. The standard threshold is dynamically set according to the real-time requirements of the bank's debt collection business. When the maximum delay deviation exceeds 100 milliseconds or the packet loss rate deviation exceeds 5%, the system determines that a compensation adjustment mechanism needs to be activated. The adjustment amount calculation unit calculates the difference between the maximum deviation value and the standard threshold. For example, if the delay deviation of monitoring point MP005 is 178 milliseconds and the standard threshold is 100 milliseconds, then the delay deviation difference is 78 milliseconds. This difference will be used as the baseline value for transmission compensation.

[0087] The compensation execution unit adjusts parameters of the first target link based on the deviation difference. For latency deviations, a strategy of dynamically adjusting the transmission protocol window size and retransmission timeout parameters is adopted. For packet loss rate deviations, compensation is achieved through forward error correction coding strength and redundant data packet injection ratio. Bandwidth utilization deviations are optimized through traffic shaping and load redistribution. The system monitors the adjustment effect in real time during the compensation process, collecting post-compensation quality index data every 5 seconds and comparing it with standard thresholds until the quality indexes of all monitoring points reach the preset standard range. A proportional-integral-derivative relationship is established between the compensation adjustment amount and the quality deviation difference. The system dynamically adjusts the compensation intensity according to the deviation change trend to avoid over-compensation or under-compensation. All compensation operations are recorded in the system log, including detailed information such as adjustment time, adjustment parameters, quality before adjustment, and quality after adjustment. This data is used for subsequent system optimization and audit tracking. Network maintenance personnel can view the compensation process and effect in real time through a graphical interface and manually intervene to adjust the compensation strategy or parameter settings when necessary.

[0088] Example 5: In the communication link management scenario of bank debt collection, the optimization and grading module implements differentiated processing strategies based on customer risk type. When the system detects an anomaly in the collection communication link of a high-risk customer (such as a customer overdue for more than 90 days), the duration monitoring unit immediately begins recording the duration of the abnormal link failure and obtains the customer's risk rating label through the bank's customer management system. The strategy selection unit compares the duration of the anomaly with a preset threshold (such as 30 seconds). If the anomaly time is short, a first-level optimization strategy is activated, which focuses on quickly restoring basic communication functions. If the duration of the anomaly exceeds the threshold, a second-level optimization strategy is activated, which enhances transmission stability while restoring communication.

[0089] The parameter adjustment unit dynamically configures the control parameters of the compensation module according to the strategy level. Under the first-level optimization strategy, a smaller bandwidth adjustment step size (e.g., 5Mbps) and a higher quality compensation frequency (twice per second) are set. The second-level optimization strategy uses a larger bandwidth adjustment step size (10Mbps) and enables redundant transmission channels. The correlation analysis unit of the strategy execution module analyzes the operation parameters in the current transmission compensation adjustment process, identifies core parameters with high correlation to customer risk level (such as the maximum guaranteed bandwidth value in bandwidth allocation parameters and the minimum latency requirement in quality compensation parameters), and distinguishes auxiliary parameters with less impact (such as log recording frequency and monitoring data sampling interval). The weight allocation unit establishes a mapping relationship between customer risk type and parameter weight. The weight of the associated parameter for high-risk customer type is set to 0.8, and the weight of the non-associated parameter is 0.2; the weight ratio of 0.6 to 0.4 is used for medium-risk customer type; and the weight configuration of 0.4 to 0.6 is used for low-risk customer type. The execution control unit implements multi-level optimization control based on weight configuration. When dealing with high-risk customer link anomalies, the system prioritizes adjusting bandwidth allocation parameters (increasing the guaranteed bandwidth from 50Mbps to 80Mbps) and quality compensation parameters (reducing the maximum allowable latency from 100ms to 50ms), while maintaining the original settings for non-critical parameters such as log sampling frequency.

[0090] In practice, the system continuously monitors the optimization effect and dynamically adjusts the weight allocation. If it finds that the recovery speed of an abnormal link for a medium-risk customer does not meet expectations under the weight configuration (0.6 / 0.4), the system will automatically increase the weight of the associated parameter to 0.7 and correspondingly reduce the weight of the unassociated parameter. All weight adjustment decisions are recorded in the system audit log, including detailed information such as adjustment time, customer risk level, duration of abnormality, weight configuration used, and adjustment effect. This data is used for subsequent iteration and improvement of optimization strategies. Bank operations and maintenance personnel can view the execution status of optimization strategies for customers with different risk levels through the management interface. In special circumstances, they can manually intervene in the weight allocation scheme, such as temporarily adopting a higher-level optimization strategy for certain important customers. The system automatically generates a strategy execution effect analysis report every week, which statistically analyzes indicators such as abnormal recovery time, compensation success rate, and resource consumption for customer links of each risk level.

[0091] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A communication link anomaly handover handling system, characterized in that, include: The monitoring module is used to acquire real-time operational data of multiple communication links in the bank's debt collection business; The real-time operational data includes connection status data and quality index data for each communication link; The analysis module is used to obtain anomaly evaluation parameters for each communication link based on the real-time operating data; the anomaly evaluation parameters are used to characterize the degree of anomaly in the connection status of the corresponding communication link. The positioning module is used to determine the abnormal communication area based on various anomaly assessment parameters; the abnormal communication area is the area in the communication link that requires a handover operation. The decision module is used to determine a first target link based on the abnormal communication area; the first target link is the primary communication link that needs to be prioritized among all communication links. The compensation module is used to perform transmission compensation adjustment on the first target link so that the communication quality index of the abnormal communication area reaches a preset standard threshold. The analysis module includes: The parameter acquisition unit is used to select a second target link from each communication link; the second target link is any one of the communication links. A timing processing unit is used to obtain timing quality index data of the second target link based on the real-time running data; The difference calculation unit is used to obtain multiple first-level differences based on the time-series quality index data; the first-level difference is the difference between two adjacent time-series quality index data. An evaluation generation unit is used to obtain the first evaluation factor based on each first-level difference; The parameter output unit is used to obtain the anomaly evaluation parameters of the second target link based on the first evaluation factor.

2. The communication link anomaly handover processing system according to claim 1, characterized in that, The evaluation generation unit is specifically used for: A first average value is obtained based on each first-level difference; the first average value is the average of the absolute values ​​of each first-level difference. The first average value is denoted as the first evaluation factor.

3. The communication link anomaly handover processing system according to claim 2, characterized in that, The parameter output unit includes: The secondary computation subunit is used to obtain multiple second-level differences based on each first-level difference; the second-level difference is the difference between two temporally adjacent first-level differences; The parameter generation subunit is used to obtain the second evaluation factor based on each second-level difference; The weighted processing subunit is used to weight the first evaluation factor and the second evaluation factor through a preset weight coefficient to obtain the abnormal evaluation parameters of the second target link.

4. The communication link anomaly handover processing system according to any one of claims 1 to 3, characterized in that, The positioning module includes: The sorting unit is used to sort all anomaly evaluation parameters from largest to smallest. The filtering unit is used to select the communication links corresponding to the top-ranked preset number of abnormal evaluation parameters in the sorting results as the third target links. The prediction unit is used to input the real-time operating data of each third target link into the prediction model; The region determination unit is used to obtain the abnormal prediction values ​​of each location in the communication network based on the prediction model, and to determine the abnormal communication region based on each abnormal prediction value.

5. The communication link anomaly handover processing system according to claim 4, characterized in that, The prediction unit includes: The link selection subunit is used to select a fourth target link and a fifth target link based on each third target link; the fourth target link and the fifth target link are any two adjacent communication links among the third target links; The point generation subunit is used to obtain a preset number of predicted points between the fourth target link and the fifth target link; The prediction execution subunit is used to obtain the abnormal prediction values ​​of each prediction point based on the prediction model.

6. The communication link anomaly handover processing system according to any one of claims 1 to 3, characterized in that, The compensation module includes: A quality detection unit is used to obtain the maximum quality deviation value of the abnormal communication area; The judgment unit is used to not initiate compensation adjustment when the maximum quality deviation value is less than or equal to the preset standard threshold, and to initiate compensation adjustment when the maximum quality deviation value is greater than the preset standard threshold. The adjustment calculation unit is used to obtain the quality deviation difference based on the maximum quality deviation value and a preset standard threshold. The compensation execution unit is used to perform transmission compensation adjustment on the first target link based on the quality deviation difference.

7. The communication link anomaly handover processing system according to claim 1, characterized in that, Also includes: The tiered system module is optimized to determine tiered optimization strategies based on bank customer type information. The bank customer type information includes high-risk customer type information, medium-risk customer type information, and low-risk customer type information; The strategy execution module is used to perform multi-level optimization control on the transmission compensation adjustment process of the compensation module based on the hierarchical optimization strategy.

8. The communication link abnormal handover processing system according to claim 7, characterized in that, The optimization and hierarchical module includes: The duration monitoring unit is used to obtain the duration of communication link abnormalities; The strategy selection unit is used to adopt a first-level optimization strategy when the duration of the communication link abnormality is less than the preset duration, and to adopt a second-level optimization strategy when the duration of the communication link abnormality is greater than or equal to the preset duration. The parameter adjustment unit is used to adjust the control parameters of the compensation module based on different levels of optimization strategies. The strategy execution module includes: The correlation analysis unit is used to obtain correlation parameters and non-correlation parameters during the transmission compensation adjustment process; the correlation parameters include bandwidth allocation parameters and quality compensation parameters. The weight allocation unit is used to determine the adjustment weights corresponding to the associated and unassociated parameters based on the bank customer type information. The execution control unit is used to perform multi-level optimization control based on the adjustment weights corresponding to the associated and non-associated parameters, respectively.

9. A method for handling abnormal handover of a communication link, characterized in that, The method flow applied to all modules of the communication link abnormal handover handling system as described in any one of claims 1 to 8.

Images