Method for automatic collection and visualization of multi-subject telecommunications billing data

Abstract

The application relates to the field of data processing and discloses a multi-subject telecom service billing data automatic collection and visual analysis method, which comprises the following steps: collecting network resource consumption data of a service sub-subject, performing alignment processing on missing values to construct a time sequence resource state matrix; disassembling an invoice file into independent cost items, matching data extraction operators according to a billing rule and applying the data extraction operators to the time sequence resource state matrix, calculating an allocation weight and a sub-item amount; generating an idle resource identifier in combination with control plane and user plane data; performing fund conservation verification on the aggregated sub-item amount to generate a comprehensive billing data vector; establishing a directed acyclic graph, mapping the sub-item amount into a connected line pixel width, and rendering a service sub-subject node according to the idle resource identifier. Through the application, accurate mapping of underlying resource consumption and financial cost is realized, the placeholder cost of idle equipment is separated, the fund flow direction and the equipment state are intuitively displayed, and the auditing efficiency of billing data is improved.

Description

Technical Field

[0001] This invention relates to the field of data processing, specifically to a method for the automatic collection, visualization, and analysis of invoicing data for multi-entity telecommunications services. Background Technology

[0002] In the shared network architecture of telecommunications services, multiple business sub-entities typically share the underlying physical communication equipment. When financial settlement occurs, the operator issues a consolidated communication invoice for the overall resource consumption. Existing data processing methods mainly rely on the macro-level business volume declared by each business sub-entity or a pre-agreed fixed percentage to statically break down the total invoice amount and allocate various expenses to the corresponding business sub-entities, thereby completing the collection and settlement of invoicing data.

[0003] Because the billing rules in consolidated invoices often encompass multi-dimensional composite models such as peak traffic billing or port minimum guarantee billing, existing static allocation methods only summarize data at the business level and fail to extract and align the instantaneous consumption data of underlying network resources in a time-series dimension. This makes it difficult to establish a precise mapping between the operating status of underlying devices and upper-level financial cost items, resulting in inaccurate calculation of allocation weights. Furthermore, due to the lack of a linkage and comparison mechanism between control plane status indicators and actual user plane load data, the system struggles to proactively identify idle physical ports that continuously generate occupancy fees despite having no actual business load. This makes it easy for the occupancy costs of idle devices to be directly allocated to entities that are generating normal business.

[0004] After cost allocation and data collection are completed, the existing outputs are mostly structured data tables or comprehensive detailed invoices. Faced with multi-level entity relationships and complex settlement indicators, auditors find it difficult to directly clarify the fund transfer path through plain text or two-dimensional tables. Due to the lack of a graphical presentation mechanism that integrates the fund allocation flow with the idle status of underlying communication equipment, the readability of invoice data is limited, resulting in time-consuming manual auditing of billing data and low efficiency in identifying non-performing assets. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an automatic collection and visualization analysis method for invoicing data of multi-entity telecommunications services. This method solves the problems of inaccurate mapping between multi-entity resource consumption and billing allocation in existing telecommunications services, as well as the difficulty in identifying and intuitively auditing the occupancy costs of idle communication equipment.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A method for automatically collecting, visualizing, and analyzing invoicing data from multiple entities in telecommunications services includes the following steps:

[0008] S100: Collect instantaneous network resource consumption data of each service sub-sub-subject. The instantaneous network resource consumption data includes at least resource indicator data, control plane status data, and user plane load data. Align missing values ​​in the time-series data sequence and construct a set of time-series resource snapshot matrices within the specified billing period.

[0009] S200 parses the received merged invoice source file, breaks down the total amount into independent cost items, and attaches the corresponding mathematical extraction operators according to the billing rules of the cost items, and establishes the mapping relationship between each cost item and the target resource indicators, the set of participating entities, and the abnormal rollback rules.

[0010] S300 applies the mounted mathematical extraction operator to the time-series resource submatrix corresponding to the target resource indicator corresponding to the billing cost item, calculates the allocation weight and item amount of each business sub-entity under different billing cost items, and synchronously generates a zombie resource identifier to represent the idle status of assets during the corresponding operator calculation process.

[0011] S400 aggregates and calculates the various itemized amount data, performs a global capital conservation check, and generates a comprehensive financial bill vector containing all business sub-entities.

[0012] S500 establishes a directed acyclic graph structure with the total invoice amount as the root node, connects cost item nodes and business sub-sub-entity nodes to the directed acyclic graph according to hierarchical relationships, maps each fund transfer amount to the rendering width of directed edges, and draws node views combined with extreme value characteristics in the leaf node area representing the business sub-sub-entity. Based on the zombie resource identifier, the graphic rendering style of the target node is changed to expose idle physical communication assets.

[0013] Furthermore, in step S100 above, the step of aligning missing values ​​in the time-series data sequence includes: for continuous indicators representing resource consumption rate, calculating interpolation filling values ​​based on the most recent valid sampled value before and after the missing point, combined with the time span ratio between the missing point and the previous valid time point; for state indicators representing equipment operating characteristics, extracting the parameter state of the previous valid time moment as the filling value for the current missing time point; if the initial sampling point is missing, extracting the most recent valid record value after the time axis as the initial state filling value.

[0014] Furthermore, in step S100 above, the step of constructing a set of time-series resource snapshot matrices within a specified billing period includes: organizing the data sequences of each business sub-subject after missing value filling preprocessing in a structured manner according to the indicator type; constructing a row dimension by business sub-subject identifier and a column dimension by time point, and encapsulating the one-dimensional time-series sequence under the same indicator type into a two-dimensional time-series resource sub-matrix; and collecting the time-series resource sub-matrices corresponding to all indicator types to form a set of time-series resource snapshot matrices.

[0015] Furthermore, in step S200 above, the step of attaching the corresponding mathematical extraction operator according to the billing rule definition of the cost item includes: when the billing rule depends on the network attachment status of the device port and is not affected by the traffic volume, a Boolean existence operator is attached to the corresponding cost item; when the billing rule depends on the total volume of data consumed within the billing period, a time integration operator is attached to the corresponding cost item; when the billing rule depends on the extreme value of data traffic generated within a specific time window, a sliding window quantile operator with a time delay tolerance parameter is attached to the corresponding cost item; when the billing rule does not belong to the above categories, it is attached to the static amortization operator by default.

[0016] Furthermore, in step S300 above, the step of applying the mounted mathematical extraction operator to the corresponding time-series resource submatrix for calculation includes: for the Boolean existence operator, extracting the row features of the target time-series resource submatrix and performing logical judgment to determine whether each service sub-subject has a network attachment state value greater than zero within the billing period, so as to determine the Boolean existence apportionment weight.

[0017] Furthermore, in step S300 above, the step of synchronously generating a zombie resource identifier to characterize the idle state of an asset during the calculation of the corresponding operator includes: based on the architectural characteristics of the separation of the network control plane and the user plane, within the observation time window, counting the cumulative number of sampling points where the control plane status indication of the service sub-sub ...

[0018] Furthermore, in step S300 above, the step of applying the mounted mathematical extraction operator to the corresponding time-series resource sub-matrix for calculation also includes: for the time integration operator, performing discrete numerical integration accumulation on the effective instantaneous resource consumption data of each business sub-sub-body matrix row vector on the time axis of the specified billing cycle, combined with a fixed time step, to obtain the total consumption of a single entity, and calculating the time integration allocation weight based on the ratio of the total consumption of a single entity to the total consumption of all entities.

[0019] Furthermore, in step S300 above, the step of applying the mounted mathematical extraction operator to the corresponding time-series resource submatrix for calculation also includes: performing local mean slice projection calculation for a sliding window quantile operator with a time delay tolerance parameter; specifically including: accumulating the total concurrent consumption of the target time-series resource submatrix along the column dimension to obtain the total concurrent consumption at each sampling time point; determining the quantile threshold according to the preset quantile ratio, and retrieving the earliest occurrence time point where the total concurrent consumption is not less than the quantile threshold as the extreme value timestamp; converting the time delay tolerance parameter into the corresponding sampling point offset step, extracting a local region matrix with the extreme value timestamp as the center and the offset step, calculating the mean slice vector of each business sub-subject within the local region matrix, and obtaining the corresponding extreme value amortization weight based on each mean slice component.

[0020] Furthermore, in step S300 above, the step of applying the mounted mathematical extraction operator to the corresponding time-series resource submatrix for calculation also includes: for the static amortization operator, determining whether the business sub-subject belongs to the preset set of participating amortization subjects; if it does, then allocating equal amortization weights according to the total number of members in the set of participating amortization subjects; if the set of participating amortization subjects is empty, then temporarily storing the independent cost item in the unallocated amount pool according to the preset standby rules.

[0021] Furthermore, in step S400 above, the steps of aggregating and calculating the output of each item amount data and performing a global capital conservation check include: summing up the values ​​of each item amount belonging to the same business sub-sub-entity to generate the aggregated total cost corresponding to the business sub-entity; and structurally binding the aggregated total cost with the underlying network state parameters and the accumulated traffic volume, and encapsulating it into a comprehensive financial bill vector with physical audit attributes.

[0022] Furthermore, in step S400 above, the step of performing global capital conservation verification specifically includes: summarizing the total collection cost of all business sub-entities, calculating the absolute value of the error between the total amount output by the system and the total amount of the merged invoice sources; determining whether the absolute value of the error is less than the system's preset tolerance threshold, and if it does not exceed it and is within the compensable range, then locating the corresponding cost item that caused the difference, and compensating the error value to the item amount of the business sub-entity with the largest allocation weight under the corresponding cost item, and recalculating the total collection cost; if the absolute value of the error exceeds the safety compensation limit, then generating an abnormal alarm log.

[0023] Furthermore, in step S500 above, the step of mapping each fund transfer amount to the rendering width of the directed edge includes: extracting the preset minimum connection width and maximum connection width, performing linear space conversion based on the ratio of the fund transfer amount from the upstream node to the downstream node in the total invoice amount, and generating the final pixel width of the directed edge on the display terminal.

[0024] Furthermore, in step S500 above, the step of drawing a node view combining extreme value features in the leaf node region representing the business sub-subject includes: drawing a miniature time-series line chart with the sampling time point as the horizontal axis and the resource consumption value as the vertical axis in the view associated with the leaf node; extracting the window start time point and window end time point generated in the local mean slice projection calculation step; and drawing a semi-transparent rectangular vertical color band that runs through the vertical axis and corresponds to the window time interval in the miniature time-series line chart based on the scaling of the time-series coordinates.

[0025] Furthermore, in step S500 above, the step of changing the graphic rendering style of the target node according to the zombie resource identifier includes: traversing the comprehensive financial bill vector corresponding to each business sub-sub-entity and extracting the zombie resource identifier; when the zombie resource identifier is identified as being in a valid state, changing the rendering material of the boundary wireframe of the corresponding leaf node and all the in-degree directed edges pointing to the leaf node from a solid line form to a hollow dot matrix form.

[0026] Furthermore, in the operation steps of the above mathematical operators: if the result of the denominator calculation is zero, causing the normal allocation logic to fail, the calculation of the current operator is terminated, and the abnormal rollback rule bound to the independent cost item is triggered; the abnormal rollback rule includes switching to the static allocation logic to allocate the cost item proportionally.

[0027] A second aspect of the present invention provides an automatic collection, visualization, and analysis system for multi-entity telecommunications service invoicing data, comprising:

[0028] The data preprocessing module is used to obtain the instantaneous physical network operating parameters of each business sub-sub-subject, perform time-series missing value filling and alignment, and generate a set of time-series resource snapshot matrices;

[0029] The invoice parsing module is used to decompose the total amount of the consolidated invoice into multiple independent cost items, attach mathematical extraction operators to the cost items, and establish underlying mapping relationships.

[0030] The heterogeneous computing module is used to call different mathematical operators to perform matrix space projection calculations on the time-series resource snapshot matrix set, calculate the allocated weights and amounts, and synchronously generate zombie resource identifiers based on the control plane and user plane running parameters.

[0031] The financial aggregation module is used to perform horizontal merging of various item amounts, implement multi-dimensional state encapsulation and capital tolerance conservation verification, and generate a comprehensive financial bill vector.

[0032] The graphics rendering module is used to parse the comprehensive financial bill vector and calculate parameters to construct a directed acyclic graph. It uses the screen border line width to map the amount of funds, uses the time sequence color band to map the extreme values ​​of the tolerance window, and uses the hollow dot matrix material to mark the ineffective costs associated with zombie resources.

[0033] This invention provides a method for automatically collecting, visualizing, and analyzing invoicing data from multiple entities in telecommunications services. It offers the following advantages:

[0034] 1. This invention constructs a time-series resource state matrix by performing alignment processing on missing values ​​of network resource consumption data, and calculates the allocation weight by matching the corresponding data extraction operator according to the billing rules of independent cost items. This realizes the data mapping between the instantaneous consumption records of the underlying devices and the financial cost items of the upper layer, and improves the accuracy of cost allocation calculation among multiple entities.

[0035] 2. This invention compares control plane status data and user plane load data, counts the number of consecutive sampling points where the control plane status data indicates validity and the user plane load data is equal to zero, and generates idle resource identifiers accordingly. This allows physical ports without actual service loads to be separated in the calculation process, avoiding the incorrect allocation of the occupancy cost of idle communication equipment to entities that normally generate services.

[0036] 3. This invention generates a comprehensive billing data vector by summarizing the itemized amounts and performing a fund conservation check. It also establishes a directed acyclic graph containing cost item nodes and business sub-entity nodes, directly mapping the amount values ​​to the pixel width of the lines connecting the nodes. At the same time, it changes the rendering style of the nodes according to the idle resource identifier, transforming the complex structured bill into a graphical interface that intuitively displays the fund flow between the entities and the idle status of the underlying equipment, thus improving the efficiency of manual auditing of billing data. Attached Figure Description

[0037] Figure 1 This is a system framework diagram of the present invention;

[0038] Figure 2 This is a flowchart of the method of the present invention;

[0039] Figure 3 This is a data processing principle diagram for constructing the time-series resource snapshot matrix of the present invention;

[0040] Figure 4 This is a flowchart of the invoice dimensionality reduction parsing and operator mounting process of the present invention;

[0041] Figure 5 This is a flowchart of the matrix heterogeneous projection calculation of the present invention;

[0042] Figure 6 This is a flowchart illustrating the multidimensional financial data aggregation process of the present invention.

[0043] Figure 7 This is a flowchart of the directed acyclic graph rendering and visual effects mapping process of the present invention;

[0044] Figure 8 This is a three-dimensional surface mapping diagram of cross-dimensional fund flow and responsibility weight according to the present invention;

[0045] Figure 9 This is a sunk cost map tracking and mapping diagram based on a force-guided layout according to the present invention;

[0046] Figure 10 This is a joint rendering diagram of the temporal resource envelope and extreme value color band of the present invention.

[0047] Among them, 100 is the data preprocessing module; 200 is the invoice parsing module; 300 is the heterogeneous computing module; 400 is the financial aggregation module; and 500 is the graphics rendering module. Detailed Implementation

[0048] The technical solutions in 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.

[0049] See attached document Figure 1 , Figure 1 This is an architecture diagram of an automatic collection, visualization, and analysis system for multi-entity telecommunications service invoicing data according to an embodiment of the present invention. The present invention provides an automatic collection, visualization, and analysis system for multi-entity telecommunications service invoicing data, comprising: a data preprocessing module 100, an invoice parsing module 200, a heterogeneous computing module 300, a financial aggregation module 400, and a graphics rendering module 500.

[0050] The data preprocessing module 100 connects to an external network management system to obtain the instantaneous physical network operating parameters of each service sub-entity. These instantaneous physical network operating parameters include at least resource consumption indicators, control plane status indicators, and user plane load indicators corresponding to the billing rules. The data preprocessing module 100 incorporates time alignment logic and a hole-filling algorithm to standardize the continuously received asynchronously sampled time-series data sequences, generating missing-item time-series resource sub-matrices according to indicator types. These sub-matrices collectively constitute a time-series resource snapshot matrix set. The data preprocessing module 100 transmits the generated time-series resource snapshot matrix set to the heterogeneous computing module 300.

[0051] The invoice parsing module 200 receives merged invoice source file data from external input, parses the billing rule field in the source file, and decomposes the total invoice amount into multiple independent billing cost items. Based on the basic attributes of each billing cost item, the invoice parsing module 200 matches the corresponding mathematical extraction operator, and further establishes a mapping relationship between the billing cost item and the target resource indicator, the set of participating entities, and the exception rollback rule, generating an operator mapping instruction set and outputting it to the heterogeneous computing module 300.

[0052] The heterogeneous computing module 300 receives a set of time-series resource snapshot matrices from the data preprocessing module 100 and a set of operator mapping instructions from the invoice parsing module 200. Based on different operator types, the heterogeneous computing module 300 calls the corresponding computation path to perform Boolean judgment, discrete integration, local window mean slicing, or static amortization calculations on the time-series resource sub-matrix corresponding to the target resource indicator. For Boolean existence operators, the heterogeneous computing module 300 simultaneously performs a two-dimensional state comparison on the control plane state sub-matrix and the user plane load sub-matrix, extracting time-series segments with network attachment status but no actual business byte transmission, and generating zombie resource identifier vectors. The heterogeneous computing module 300 obtains the amortization weight, item amount, and status identifier of each business sub-entity under different billing cost items and sends the results to the financial aggregation module 400.

[0053] The financial aggregation module 400 performs horizontal aggregation of the itemized amounts of each business sub-entity under different billing cost items, generates a comprehensive financial bill vector, and performs unallocated amount replenishment, capital conservation verification, and terminal formatted output.

[0054] The graphics rendering module 500 receives a comprehensive financial bill vector from the financial aggregation module 400, as well as cost item hierarchy relationships, operator parameters, peak window parameters, and zombie resource identifiers from the invoice parsing module 200 and the heterogeneous computing module 300. Based on this data, it generates a directed acyclic graph data structure. The graphics rendering module 500 converts the numerical values ​​corresponding to the fund transfer amount into directed edge width parameters in the graph model and generates a node view containing a miniature time-series line chart and a tolerance window color band on the front-end interface. For business sub-entity nodes carrying zombie resource identifiers, the graphics rendering module 500 renders their node borders and in-degree directed edges as hollow dot matrix shapes to expose idle physical communication assets that generate network port occupancy fees while completing financial aggregation.

[0055] See attached document Figure 2 , Figure 2 This is a flowchart of a method for automatically collecting, visualizing, and analyzing invoicing data of multi-entity telecommunications services according to an embodiment of the present invention. The present invention provides a method for automatically collecting, visualizing, and analyzing invoicing data of multi-entity telecommunications services, comprising the following steps:

[0056] S100: Collect instantaneous network resource consumption data of each service sub-sub-subject. The instantaneous network resource consumption data includes at least resource index data, control plane status data, and user plane load data. Align the holes in the time-series data sequence through linear interpolation, forward padding, or backward padding logic to construct a set of time-series resource snapshot matrices within the specified billing period.

[0057] S200 parses the received merged invoice source file, breaks down the total amount into independent cost items, and attaches Boolean existence operators, time integration operators, sliding window quantile operators with time delay tolerance parameters or static amortization operators according to the billing rules definition of cost items, and establishes the mapping relationship between each cost item and the target resource indicators, the set of participating entities, and the abnormal rollback rules.

[0058] S300 applies the mounted mathematical extraction operators to the time-series resource submatrix corresponding to the target resource indicator for the billing cost item. For Boolean existence and time integration operators, it extracts row features of the target time-series resource submatrix and performs logical judgment and numerical integration. For Boolean existence operators, it simultaneously performs a two-dimensional state comparison between the control plane state submatrix and the user plane payload submatrix, generating zombie resource identifiers for matrix sequences that only maintain device heartbeat or port open states but do not generate effective payloads. For sliding window quantile operators, it extracts a local region matrix of the target time-series resource submatrix based on the delay tolerance parameter, calculates the mean slice vector, and obtains the allocation weight and item amount corresponding to the business sub-subject. For static allocation operators, it performs proportional allocation calculations within a preset set of participating allocation subjects.

[0059] S400 aggregates the output of various itemized amount data and performs cost item-level compensation for unallocated amounts due to zero denominators or floating-point rounding, generating a complete financial aggregate bill containing all business sub-entities.

[0060] S500 establishes a directed acyclic graph structure with the total invoice amount as the root node. Cost item nodes and business sub-sub-entity nodes are connected to the directed acyclic graph according to the hierarchical relationship. The fund transfer amount corresponding to each edge is mapped to the rendering width of the directed edge. A time-series line graph corresponding to the leaf node region representing the business sub-sub-entity is drawn. The corresponding window color band is rendered in the time-series line graph in combination with the time delay tolerance parameter used in the sliding window quantile operator. According to the zombie resource identifier, the corresponding directed edge path is rendered as a hollow dot matrix primitive to present the abnormal cost flow that has shared the minimum monthly communication fee but has not generated business network benefits.

[0061] The steps described above will be explained in detail below with reference to specific embodiments.

[0062] See Figure 3 , Figure 3 This is a schematic diagram illustrating the data processing principle for constructing a time-series resource snapshot matrix according to an embodiment of the present invention. In this embodiment, the data preprocessing module 100 performs the construction operation of the time-series resource snapshot matrix set, which may include the following steps:

[0063] S110 performs asynchronous polymorphic network data acquisition. The data preprocessing module 100 establishes a communication connection with the streaming telemetry gateway of the external network management system via an application programming interface. This occurs within the specified financial billing cycle. Inside, the data preprocessing module 100 operates at a configured fixed time step. A polling sampling request is sent to the network management system to obtain instantaneous network resource consumption data. The instantaneous network resource consumption data includes at least: resource indicator data to characterize the actual consumption level, control plane status data to characterize the network attachment or link connection status, and user plane payload data to characterize the effective transmission of service bytes.

[0064] As a preferred method, a fixed time step The value is set to a time interval of minutes or seconds based on the performance of the underlying physical probe. The total number of service sub-entities sharing basic telecommunications resources within the system is set to... The theoretical total number of sampling time points within the billing cycle is Mark any business sub-entity as ,in … Mark any sampling time point as ,in … Assume that the system collects a total of [number] data. Network metrics, which label any type of network metric as ,in … The data preprocessing module 100 at time points... Obtain business sub-entities The Sampled values ​​of the original network metrics and recorded them as Due to communication delays and signaling jitter caused by the underlying physical probes, various raw time-series data sequences may have data gaps due to the lack of valid return values. For different indicator types, the data preprocessing module 100 establishes corresponding time-series data sequences and performs alignment and filling according to indicator type in subsequent steps.

[0065] S120, perform alignment and hole cleaning preprocessing of the time-series data sequence. To provide missing matrix input required for subsequent operator calculations, the data preprocessing module 100 executes branch processing logic based on the business attribute characteristics of the instantaneous network resource consumption data.

[0066] The following is based on any one of the first Taking the time-series data sequence corresponding to the category indicator as an example, based on the attribute differences of each indicator, for continuous indicators representing resource consumption rates, the data preprocessing module 100 calls a linear interpolation algorithm to numerically fill in the data gaps. This is set at the effective time point. and subsequent effective time points There are missing time points in the data. ,satisfy ,and and For valid sampled records, the data preprocessing module 100 calculates missing time points. fill value The calculation formula is as follows:

[0067] ;

[0068] In the formula, The effective values ​​output are calculated using linear interpolation. This represents the most recent valid sample value before the missing point. This represents the most recent valid sample value after the missing point. The parameter representing the time span between the missing point and the preceding valid time point.

[0069] For state-type indicators characterizing equipment operating features, the data preprocessing module 100 calls the forward filling algorithm to preserve the state of data gaps. At time point... When sampled values ​​are missing, the data preprocessing module 100 extracts the time points. The valid recorded value is used as the fill value for the current time point. In a physical sense, this represents maintaining the parameter state unchanged from the previous valid time point. The calculation formula is as follows:

[0070] ;

[0071] In the formula, This represents the current state value output by the forward padding algorithm. This represents the valid state value from the previous time point. To avoid logical dead zones in forward padding due to missing samples at the first sampling point, if the time point... The data at that point is missing. The data preprocessing module 100 executes backward filling logic to extract the most recent valid record value from the timeline as the state value of the initial time point. For the data acquisition of the specific interface protocol of the network management system and the underlying code language implementation of the basic missing value interpolation algorithm, those skilled in the art can refer to the relevant open-source data processing component standard library for execution and calling. Its basic function configuration is a well-known technology in this field and will not be elaborated upon here.

[0072] S130, Generate and encapsulate a set of time-series resource snapshot matrices. The data preprocessing module 100 structures the data sequences of each business sub-subject after processing by linear interpolation, forward padding, or backward padding logic according to their indicator types. To facilitate the vectorized reading operation of the heterogeneous computing module 300, the data preprocessing module 100 constructs row dimensions based on business sub-subject identifiers and column dimensions based on time points, encapsulating one-dimensional time-series sequences of the same indicator type into a set with a dimension size of [missing information]. The two-dimensional time-series resource submatrices. The two-dimensional time-series resource submatrices corresponding to all indicator types collectively constitute the time-series resource snapshot matrix set. .

[0073] As a preferred approach, the set of time-series resource snapshot matrices can be represented as:

[0074] ;

[0075] Among them, any All are of dimension size A two-dimensional matrix, the elements of which are composed of valid sampled values ​​after hole cleaning preprocessing:

[0076] ;

[0077] In the formula, This refers to the total number of business sub-entities sharing basic telecommunications resources within the system. The total theoretical number of sampling time points within the billing period is represented by the matrix elements, which represent the effective sampled values ​​after hole cleaning preprocessing. The preferred denot is denoted as . The row vectors of the matrix represent the specific business sub-entity within the complete billing cycle regarding the [number]th [item]. The time-series change sequence of the class of indicators, the matrix column vectors represent the changes of all business sub-entities at a specific instant regarding the first... The cross-sectional state of the index. The heterogeneous computing module 300 selects the corresponding time-series resource sub-matrix from the time-series resource snapshot matrix set and performs calculations based on the target resource index given in the operator mapping instruction set.

[0078] See Figure 4 , Figure 4 This is a flowchart of invoice dimensionality reduction parsing and operator mounting according to an embodiment of the present invention. In this embodiment, the invoice parsing module 200 performs invoice file disassembly and operator configuration operations, which may include the following steps:

[0079] S210, perform the breakdown of independent cost items for the total amount of consolidated invoices. The invoice parsing module 200 receives the consolidated invoice source file data from external input. In actual business scenarios, the billing data structure issued by telecom operators is presented as a comprehensive file containing multiple billing dimensions. The invoice parsing module 200 uses its built-in text parsing engine to identify the billing detail data table in the invoice file and extract the total invoice amount and the corresponding billing rule fields.

[0080] Based on the account classification attribute in the billing rules field, the invoice parsing module 200 decomposes the total invoice amount into multiple orthogonal and non-overlapping independent scalar cost items. The total amount of the merged invoices is set to... The set of independent cost items after decomposition is ,in This represents the total number of cost items. The total invoice amount and each individual cost item satisfy a linear summation mathematical relationship, calculated using the following formula:

[0081] ;

[0082] In the formula, For the first The specific monetary value of each independent scalar cost item. For optical character recognition and structured detail field extraction of text characters in the invoice source file, those skilled in the art can use mature document parsing components for processing. The document parsing process is well-known in the field and will not be elaborated upon here.

[0083] S220 performs matching and mounting of extraction operators based on billing physical attributes. After obtaining individual cost items, in order to establish an accurate computational mapping relationship between financial cost data and physical network resource status, the invoice parsing module 200 matches the corresponding mathematical extraction operator based on the basic attributes of each decomposed billing cost item. The invoice parsing module 200 traverses the set of independent cost items. Each cost item The corresponding billing form identifier is extracted. The billing form identifier can be determined by matching the account classification field, billing description text, and fee code in the invoice with a preset rule mapping table. Furthermore, the target resource indicator, the set of participating entities, and the abnormal rollback rules corresponding to the cost item are determined.

[0084] When the billing form identifier represents a guaranteed monthly fee, the operator's billing logic depends only on whether the device port is in a network attached active state, and is not affected by the actual business data traffic volume. The invoice parsing module 200 is used as the corresponding cost item to instantiate and attach a Boolean existence operator, and the control plane status indicator is determined as the target resource indicator.

[0085] When the billing form identifier represents the linear total traffic, that is, the operator's billing logic depends on the total volume of data packets consumed within a specific billing cycle, the invoice parsing module 200 instantiates the corresponding cost item and attaches the time integration operator, and determines the traffic rate indicator or the cumulative byte count indicator as the target resource indicator.

[0086] When the billing type identifier represents shared peak bandwidth, the industry-standard percentile extreme value billing mode is adopted. The billing logic depends on the extreme value of data traffic generated concurrently by each business sub-sub-entity within a specific time window. The invoice parsing module 200 instantiates the corresponding cost item and attaches a sliding window quantile operator with a time delay tolerance parameter, and determines the bandwidth utilization rate indicator or concurrent traffic indicator as the target resource indicator.

[0087] As a preferred approach, the delay tolerance parameter is set as follows: Its specific value range is configured from hundreds of milliseconds to several seconds, and the specific value is directly determined by the timestamp synchronization accuracy of the underlying hardware. Because there is a slight misalignment in timestamps between the instantaneous consumption data collected by the enterprise side and the extreme value data statistically collected by the operator side, a latency tolerance parameter is set. It can establish a buffer judgment interval near the extreme time point to avoid the distortion of the calculated weight due to the failure of strict time point alignment.

[0088] To ensure the integrity of the system's operational logic and avoid interruptions in the parsing process due to unknown billing types, when the billing type identifier does not belong to the three preset categories mentioned above, the invoice parsing module 200 defaults to mounting it as a static allocation operator. As a preferred method, the static allocation operator divides the weights proportionally within a preset set of participating entities; the set of participating entities can be the entire set of business sub-entities or a subset of business sub-entities that meet the preset participation conditions.

[0089] After completing the instantiation and configuration of the operators, the invoice parsing module 200 encapsulates all independent cost items and their bound mathematical extraction operators, target resource indicators, participating allocation subject sets, exception rollback rules, and corresponding configuration parameters into a standardized set of operator mapping instructions. The invoice parsing module 200 outputs the generated set of operator mapping instructions to the heterogeneous computing module 300 as the underlying execution criterion for driving matrix calculations.

[0090] See Figure 5 , Figure 5This is a flowchart of a matrix heterogeneous projection calculation according to an embodiment of the present invention. In this embodiment, the heterogeneous calculation module 300 receives a set of time-series resource snapshot matrices from the data preprocessing module 100, and performs multi-dimensional financial space mapping and matrix calculation based on the set of operator mapping instructions generated by the invoice parsing module 200. The term "heterogeneous calculation" as used herein refers to the use of different mathematical extraction operators corresponding to different billing cost items, calling different calculation paths such as Boolean decision, discrete integration, local window mean slicing, and static amortization to complete the aggregation calculation. Specific operations may include the following steps:

[0091] S310 performs row feature logical projection and zombie resource identification based on Boolean existence operators. For independent cost items in the operator mapping instruction set that carry Boolean existence operators, the heterogeneous computing module 300 extracts row features of the target time-series resource submatrix and performs logical judgment. At the underlying physical layer, the minimum cost is related to whether the device is powered on or whether the port is active. Based on this, the heterogeneous computing module 300 traverses the row vectors of the target time-series resource submatrix to determine whether each service sub-subject has network attachment status within the billing period.

[0092] As a preferred approach, the Boolean existence state of the business sub-subject is defined as follows: If the state value of the business sub-entity is greater than zero at any sampling time point, then it is determined that... ,otherwise The heterogeneous computing module 300 calculates the average weight accordingly. The calculation formula is:

[0093] ;

[0094] In the formula, For specific business sub-entities The corresponding Boolean values ​​have allocated weights. As a business sub-entity The Boolean existence state value, This represents the total number of business sub-entities within the system. For the first The Boolean existence state value of each business sub-entity.

[0095] To prevent the logical anomaly of a denominator being zero due to the entire system failing to generate an attached state during the billing cycle, if If the heterogeneous computing module 300 does not directly invalidate the independent cost item, it triggers the abnormal rollback rule corresponding to the independent cost item. As a preferred approach, the abnormal rollback rule is to switch to the static amortization operator and perform proportional amortization on the independent cost item within a preset set of participating amortization entities.

[0096] During the execution of Boolean existence operator logical projection, the heterogeneous computing module 300 simultaneously performs a two-dimensional state comparison of the network control plane and user plane to examine the actual usage efficiency of communication physical assets. Based on the architectural characteristics of modern telecommunications networks where control plane signaling and user plane service payloads are separated, the heterogeneous computing module 300 extracts the control plane state sequence and user plane payload sequence of the service sub-subject respectively.

[0097] Set the control plane status to When the device completes RADIUS authentication or the port is in a link-connected state, the control plane status value is 1; otherwise, it is 0. The user plane payload is set to... Characterized at a time point The actual number of valid communication bytes generated. Heterogeneous computing module 300 within an observation time window. Internal execution logic matching, setting the observation time window The size is configured to be several hours or several days depending on the audit granularity requirements. To improve the enforceability of the judgment rules, the heterogeneous computing module 300 uses an observation time window... Internal statistics simultaneously satisfy and Number of sampling points and compare it with a preset threshold. Compare; when At that time, the judgment Otherwise, it will be judged. As a preferred method, the preset threshold It can be set as the minimum number of consecutive sampling points that meet the above conditions within the observation time window, or as a threshold for the cumulative number of sampling points that meet the above conditions, and its expression is:

[0098] ;

[0099] ;

[0100] In the formula, Indicates business sub-entity The number of sampling points that meet the condition of "control surface effective and user surface zero load" within the observation time window. This indicates the preset judgment threshold.

[0101] S320 performs linear consumption area projection based on the time integration operator. For independent cost items in the operator mapping instruction set that are equipped with the time integration operator, the heterogeneous computing module 300 performs surface accumulation and addition operations on the row vectors of each service sub-subject matrix on the specified billing cycle time axis. The discrete integration process is physically equivalent to calculating the cumulative traffic volume covered by discrete sampling points. The heterogeneous computing module 300 calls the discrete numerical integration algorithm to calculate the total cumulative data consumption volume corresponding to a specific service sub-subject, and the calculation formula is as follows:

[0102] ;

[0103] In the formula, As a business sub-entity Total points consumed throughout the entire billing cycle For the corresponding time point in the time-series resource snapshot matrix Instantaneous resource consumption data, For the configured fixed time step, This represents the theoretical total number of sampling time points within the billing cycle.

[0104] After obtaining the total consumption of each individual, the heterogeneous computing module 300 summarizes the total point consumption of all business sub-entities, and calculates the allocation weight corresponding to each business sub-entity by calculating the ratio of individual consumption to total consumption. The calculation formula is as follows:

[0105] ;

[0106] In the formula, As a business sub-entity The corresponding time integral amortization weight, As a business sub-entity Total points consumed throughout the entire billing cycle This represents the total number of business sub-entities within the system. For the first The total points consumed by each business sub-entity. To avoid logical anomalies where the denominator is zero due to no traffic generated by the system as a whole, if... If the heterogeneous computing module 300 triggers the abnormal rollback rule corresponding to the independent cost item, it will perform standby allocation processing on the independent cost item instead of directly resetting all allocation weights to zero. The heterogeneous computing module 300 will allocate weights... Multiply by the corresponding independent scalar cost item amount to obtain the itemized amount of the business sub-entity under the corresponding billing rule. For the writing of the loop accumulation code in the underlying computer system for basic discrete integral operations, those skilled in the art can refer to the standard algebraic operation logic implementation; its basic implementation logic is well-known technology in this field and will not be elaborated here.

[0107] S330 performs local mean slice projection calculation based on the time-delay-tolerant quantile operator. For independent cost items in the operator mapping instruction set that are equipped with sliding window quantile operators with time-delay tolerance parameters, the heterogeneous computing module 300 performs local window slicing to reduce extreme value amortization distortion caused by the asynchronous time delay deviation between the physical probe and the billing gateway.

[0108] The heterogeneous computing module 300 accumulates the target temporal resource submatrix along the column dimension to calculate the total concurrent consumption of the system at each sampling time point, forming the total system consumption set. Let the total system consumption sequence be... ,in:

[0109] ;

[0110] The heterogeneous computing module 300 determines the quantile threshold according to the preset quantile ratio. As a preferred method, if the preset quantile ratio is configured to be 95%, then:

[0111] ;

[0112] After obtaining the quantile threshold, the heterogeneous computing module 300 retrieves a set of candidate time points from the original time series that satisfy the condition that the total system consumption is not less than the quantile threshold. :

[0113] ;

[0114] If the candidate time point set If there are multiple time points, the earliest time point is taken as the extreme value timestamp. ,Right now:

[0115] ;

[0116] After locating the extreme timestamp, the heterogeneous computing module 300 uses the time delay tolerance parameter. A local region matrix of the target temporal resource submatrix is ​​extracted. To convert the temporal parameters into the index span of the matrix column dimensions, the heterogeneous computing module 300 converts the temporal tolerance parameter into the corresponding sampling point offset steps. The sampling point offset step number is the rounded-up value of the ratio of the delay tolerance parameter to the fixed time step size, and its expression is:

[0117] ;

[0118] Heterogeneous computing module 300 in time interval Inner extraction matrix slices, where:

[0119] ;

[0120] ;

[0121] To avoid algorithmic errors caused by matrix out-of-bounds errors due to extreme time points being too close to the beginning or end of the billing cycle, the above boundary constraints are automatically applied when extracting slices.

[0122] Within the truncated local region matrix, the heterogeneous computing module 300 calculates the mean slice vector for each business sub-subject along the time dimension. Since the actual number of sampling points participating in the mean calculation may be less than [a certain value] in the boundary truncation scenario, [the calculation may be less than a certain value]. Therefore, the actual window length is defined. for:

[0123] ;

[0124] Then business sub-entity Mean slice components The expression is:

[0125] ;

[0126] Using the calculated mean components, the heterogeneous computing module 300 uses the mean slice vector as a benchmark to calculate the allocated weights corresponding to the business sub-sub-entities. When the sum of the mean slice components of all business sub-sub ...

[0127] ;

[0128] Based on the allocation weight The heterogeneous computing module 300 further multiplies the corresponding extreme value billing cost item amount and outputs the item amount of the business sub-sub ...

[0129] S340 performs proportional weight allocation based on the static amortization operator. For independent cost items in the operator mapping instruction set that are equipped with the static amortization operator, the heterogeneous computing module 300 extracts the set of participating entities corresponding to that cost item. Let the number of entities in the set of participating entities be... When the business sub-entity When a participant belongs to the aforementioned set of entities participating in the cost-sharing, it is assigned a static average weight. :

[0130] ;

[0131] When the business sub-sub-entity If an entity does not belong to the aforementioned set of entities participating in the cost-sharing, its static cost-sharing weight is assigned to 0. Alternatively, it can be uniformly written as:

[0132] ;

[0133] In the formula, For the first Each independent cost item corresponds to a set of participating entities for allocation. The heterogeneous calculation module 300 is based on static allocation weights. Multiply by the corresponding independent cost item amount to output the itemized amount for the business sub-entity under that cost item. If the set of participating entities is empty, the independent cost item will be temporarily stored in the unallocated amount pool according to the preset standby rules; if the system does not configure the corresponding standby rules, the independent cost item will be marked as an item awaiting manual review.

[0134] See Figure 6 , Figure 6 This is a flowchart of a multi-dimensional financial data aggregation process according to an embodiment of the present invention. In this embodiment, the financial aggregation module 400 receives various calculation results from the heterogeneous computing module 300, performs horizontal cost overlay and multi-dimensional data alignment and encapsulation, and the specific operation may include the following steps:

[0135] S410 performs horizontal cost aggregation in orthogonal dimensions. The financial aggregation module 400 extracts the itemized amounts independently generated by each business sub-entity under different operator mapping instructions within the system, and horizontally merges the discrete billing data based on a unified entity identifier. From a business logic perspective, the various billing rules parsed in the early stages are mathematically orthogonal, and the itemized costs generated by each need to be ultimately aggregated to the single physical entity that triggered the cost, thereby achieving single-entity convergence of multi-dimensional costs. The heterogeneous computing module 300 is set as the business sub-entity within the billing cycle. The calculated set of independent item amounts is ,in This is the total number of independent cost items derived from the total invoice amount. The financial aggregation module 400 calls scalar addition logic to calculate the number of business sub-entities. Total allocated cost The calculation formula is as follows:

[0136] ;

[0137] In the formula, As a business sub-entity The final total cost collected during the current billing cycle. The total number of independent cost items. As a business sub-entity Corresponding to the The allocated amount for each independent scalar cost item. Based on the above algebraic summation calculation, the financial aggregation module 400 realizes the aggregation of financial costs from the perspective of a single entity.

[0138] S420 generates a comprehensive financial billing vector with physical audit attributes. To meet the enterprise's internal network management center's requirements for reviewing communication asset performance, the financial aggregation module 400 structurally binds the total allocated cost with the status identifiers extracted during the underlying physical detection process, constructing a joint data structure containing both financial and network dimensions. The financial aggregation module 400 extracts various status parameters output by the heterogeneous computing module 300, including Boolean existence states. and zombie resource identifiers As a preferred approach, the financial aggregation module 400 serves as a sub-entity of the business. Concatenate multi-dimensional comprehensive financial statement vectors Its structure can be represented as:

[0139] ;

[0140] In the formula, This is the encapsulated comprehensive financial statement vector. For the system's globally unique identifier for the business sub-entity, The total cost of horizontally stacked outputs. As a business sub-entity The set of itemized amounts corresponding to each independent cost item A Boolean existence value representing the active state of a device. A zombie resource identifier to represent the idle state of assets. This represents the total volume of data packets consumed within the billing cycle. By retaining the set of itemized amounts in the comprehensive financial bill vector, the subsequent graphics rendering module 500 can construct a directed acyclic layer structure based on the cost item dimension, from the total invoice amount to the cost item and then to the business sub-subject.

[0141] S430 performs global capital conservation verification and terminal formatted output. Before the final billing data is entered into the database or issued, the financial aggregation module 400 performs system-level data integrity checks to prevent discrepancies between the allocated total and the invoice source total caused by the truncation of floating-point arithmetic precision at the computer's underlying level. The financial aggregation module 400 summarizes the total allocated cost of all business sub-entities and calculates the total amount output by the system and the total amount of the merged invoice source. Error value between Its expression is:

[0142] ;

[0143] In the formula, This is the global funding error value calculated by the system. The total amount of the consolidated invoices entered from external sources. This represents the total number of business sub-entities within the system. (400 error value for the financial aggregation module) Is it less than the system's preset tolerance threshold? Tolerance threshold The value is set based on the minimum currency settlement unit of the enterprise's financial system, configured as the minimum circulation unit of the corresponding currency or the set micro floating-point tolerance range.

[0144] like The financial aggregation module 400 determines that the global capital conservation check has passed, and then vectorizes the comprehensive financial statements of each business sub-entity. Formatted output is sent to the associated database component or external display terminal. To ensure the system's automated workflow and avoid process dead zones caused by minor floating-point errors, if any are detected during the verification process... The financial aggregation module 400 further determines the error value. Is it within the compensable range?

[0145] When errors are caused by multi-step division truncation or rounding, the financial aggregation module 400 prioritizes error compensation at the cost item level. This means it first identifies the corresponding cost item causing the difference and compensates the error value of that cost item to the sub-item amount of the business sub-entity with the largest allocation weight under that cost item. Then, it recalculates the total allocated cost of the corresponding business sub-entity. This ensures that error compensation remains within the original cost allocation logic range; only when the error value exceeds the safety compensation limit will the system generate an abnormal alarm log and trigger a manual review process.

[0146] For the persistent storage mechanism of the final bill vector in the relational database and the report rendering and display based on the front-end browser page, those skilled in the art can call conventional persistence frameworks and chart components for adaptation. The specific implementation mechanism is a well-known technology in this field and will not be elaborated here.

[0147] See Figure 7 , Figure 7 This is a flowchart of directed acyclic graph rendering and visual effects mapping according to an embodiment of the present invention. In this embodiment, the graphics rendering module 500 receives a comprehensive financial bill vector from the financial aggregation module 400, a set of cost item hierarchy and operator mapping instructions from the invoice parsing module 200, and rendering parameters such as peak timestamps, window offset steps, and zombie resource identifiers from the heterogeneous computing module 300. It transforms the abstract financial aggregation results and underlying physical state into a multi-dimensional visual front-end interface. Specific operations may include the following steps:

[0148] S510 constructs a directed acyclic graph topology layer-level structure and edge width mapping based on financial aggregation results. In enterprise communication settlement scenarios, fund flows exhibit a unidirectional convergence characteristic, decomposing from consolidated invoice sources to specific billing rules, and then allocating funds down from the billing rules to various business sub-entities. The graphics rendering module 500 transforms the acyclic unidirectional flow attribute into a graph model data structure. In this structure, the graphics rendering module 500 extracts the total amount of the merged invoice sources to establish a unique root node, establishes intermediate nodes based on the set of independent cost items output by the invoice parsing module 200, and establishes leaf nodes based on the comprehensive financial bill vector output by the financial aggregation module 400. After constructing the basic hierarchical nodes, the graphics rendering module 500 establishes a set of directed edges based on the cost item amounts and the sets of sub-item amounts. The edge amount from the root node to each cost item node is the amount of the corresponding independent cost item, and the edge amount from each cost item node to the leaf node of the business sub-sub-entity is the sub-item amount of the corresponding business sub-entity under that independent cost item.

[0149] To visually represent the proportion of each expense component, the graphics rendering module 500 maps the absolute value of the fund transfer amount to the geometric rendering width of directed edges. The graphics rendering module 500 calculates the directed edges. Screen pixel width The calculation formula is as follows:

[0150] ;

[0151] In the formula, The width of the directed edge in pixels after mapping. and These are the minimum and maximum connection widths set according to the resolution of the front-end display terminal, respectively. To start from the upstream node Flowing to downstream nodes The amount of funds transferred, To consolidate the total amount of invoice sources, the graphics rendering module 500 completed the visualization topology assembly of the basic funding dependency network through a mapping mechanism.

[0152] S520 performs vertical indicator color band rendering based on the latency tolerance parameter. The graphics rendering module 500 renders a miniature time-series line chart in the pop-up view or sidebar associated with the leaf node. The horizontal axis of the miniature time-series line chart represents the theoretical total sampling time points within the billing cycle. The vertical axis represents the instantaneous consumption value of the target resource indicator corresponding to the current cost item. The graphics rendering module 500 extracts the extreme value timestamps output by the heterogeneous computing module 300 during slice calculation. Window start time point Window end time point Or the equivalent number of sampling point offset steps In the plotting coordinate system of the miniature time-series line chart, the graphics rendering module 500 uses intervals... To represent the tolerance window range corresponding to this cost item, a rectangular semi-transparent color band is drawn running through the vertical axis. Let the total pixel length of the horizontal axis be... Then the graphics rendering module 500 calculates the x-coordinates of the start and end pixels of the color band on the screen. and Their expressions are as follows:

[0153] ;

[0154] ;

[0155] After obtaining the start and end coordinates, the graphics rendering module 500 calls the graphics drawing instructions to generate fill transparency parameters. A rectangular color band primitive. As a preferred method, the transparency parameter... The value range is set to 0.3 to 0.5 to ensure that the color band indicates the range of extreme value influence without obscuring the underlying time series line data.

[0156] S530 performs dimensionality reduction rendering of hollowed-out dot matrix primitives based on zombie resource identifiers. The graphics rendering module 500 calls the state parsing logic to expose invalid fund flows based on the underlying physical state in the comprehensive financial statement vector. The graphics rendering module 500 traverses the comprehensive financial statement vector corresponding to each leaf node. Extract the zombie resource identifiers encapsulated within. .

[0157] As a preferred method, when detected When the corresponding business sub-entity is identified as a placeholder asset with no business load, the graphics rendering module 500 determines that it is an idle asset. Upon triggering this determination, the graphics rendering module 500 invokes the graphics dimensionality reduction rendering pipeline, modifying the graphic properties of the boundary wireframe of the corresponding leaf node and all directed edges pointing to the corresponding leaf node from solid lines to a hollow dot matrix. Through the visual contrast between solid lines and hollow dot matrices, the graphics rendering module 500, without altering the topology of the fund flow, achieves a rendering output that intuitively links the financial allocation flow with the ineffective load of the physical network, assisting enterprise network administrators in identifying and clearing idle communication resources.

[0158] For the specific WebGL or Canvas instructions encapsulated in the front-end browser for the directed graph primitive style, those skilled in the art can call the graph rendering engine component to implement it. The basic drawing logic is a well-known technology in this field and will not be elaborated here.

[0159] In an enterprise-level hybrid cloud network resource settlement scenario, the business sub-entity A carrying streaming media services, the business sub-entity B carrying the core database, and the business sub-entity C carrying the testing environment are all designated to participate in the cost-sharing. The billing cycle is set at thirty calendar days, and the total amount of the consolidated invoices received by the system is one million yuan.

[0160] The invoice parsing module 200 breaks down the total amount into three independent cost items: a basic guaranteed bandwidth fee of 300,000 yuan, an extreme bandwidth fee of 500,000 yuan, and an excess traffic fee of 200,000 yuan. The heterogeneous computing module 300 matches corresponding data extraction operators for different independent cost items. During the calculation of the state detection logic operator, business sub-entity C did not generate actual user plane payload data, but its control plane status data indicated that the physical port was in a valid state; the heterogeneous computing module 300 determined that business sub-entity C was in an idle and occupies space state, generated an idle resource identifier for it, and calculated its sub-item amount under the basic guaranteed bandwidth fee to be 50,000 yuan.

[0161] During the calculation process of the extreme value filtering logic operator, business sub-entity A generates a traffic extreme value that lasts for two hours on the fifteenth day of the billing cycle; the heterogeneous computing module 300 extracts the local region matrix corresponding to the extreme value timestamp for calculation and determines that the itemized amount of business sub-entity A under the extreme value bandwidth fee is 400,000 yuan.

[0162] The financial aggregation module 400 aggregates the amounts of each item, performs a capital conservation check, and generates a comprehensive billing data vector containing the data of each business sub-entity. The graphics rendering module 500 constructs a directed acyclic graph (DAG) based on the comprehensive billing data vector, using the total amount as the root node and mapping the item amounts to the pixel width of the DAG lines. For business sub-entity C, which is in an idle state, the graphics rendering module 500, based on the idle resource identifier, changes the boundary frame of its corresponding business sub-entity node from a solid line to a dotted outline. For business sub-entity A, which generates extreme traffic values, the graphics rendering module 500 extracts the start and end times of the time window from the view associated with the node and draws a rectangular vertical color band covering the corresponding time interval to identify the time series interval for extreme value billing.

[0163] Backtracking tests were conducted using network time-series monitoring snapshots and historical billing statements from a real cloud platform. A traditional static algorithm based on absolute linear amortization of total traffic bytes was used as a benchmark. This traditional static algorithm ignores the resident costs of equipment docking stations and penalties for sudden peak traffic. To avoid data deviations from engineering realities, the collected samples were normalized. The experimental comparison data between the multi-dimensional heterogeneous amortization scheme and the traditional static algorithm are shown in the table below:

[0164] Subsidiary A's share of the total amount 600,000 yuan 620,000 yuan Accurately achieve cross-source tracing of extreme costs and traffic consumption Subsidiary B's share of the total amount 400,000 yuan 330,000 yuan Excluding extreme peak interference and reverting to baseline water consumption level Subsidiary C's share of the total amount 0 yuan 50,000 yuan Hidden sunk costs caused by the persistent presence of physical ports for capture Zombie resource discovery mechanism Relying on underlying physical periodic inspections The financial topology diagram is presented intuitively using a dot matrix material. Improve the efficiency of clearing out idle network assets Global billing reconciliation error 0.25 yuan 0.00 yuan Tolerance checking and automatic compensation mechanisms eliminate computing power dead zones

[0165] Experimental data demonstrates that the technical solution improves the accuracy of physical attribution in multi-dimensional composite billing scenarios. Numerical comparisons show that the combination of orthogonal dimension reduction computation and force-guided topology graph rendering not only achieves lossless convergence of funds but also provides the auditing capability of reverse reviewing the underlying network performance of financial bills, with data fluctuation range conforming to the real physical laws of enterprise-level applications.

[0166] See Figure 8 , Figure 8 The three-dimensional surface mapping diagram of cross-dimensional capital flow and responsibility weight according to the present invention shows the morphological differences in cost spatial distribution between the comparison algorithm and the multi-dimensional heterogeneous allocation scheme.

[0167] See Figure 9 , Figure 9 The sunk cost map based on the force-oriented layout of the present invention shows the topological matrix of unidirectional capital flow and the actual visual penetration effect of the hollow dot matrix material.

[0168] See Figure 10 , Figure 10 According to the joint rendering diagram of the temporal resource envelope and extreme color band of the present invention, the underlying computer graphics construction output of superimposing a transparency indicator matrix on a non-stationary flow waveform is depicted.

[0169] 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 method for automatically collecting, visualizing, and analyzing invoicing data of multi-entity telecommunications services, characterized in that: Includes the following steps: S100 collects network resource consumption data of the service sub-sub-subject. The network resource consumption data includes resource indicator data, control plane status data and user plane load data. It performs alignment processing on the time series missing values ​​of the network resource consumption data and constructs the time series resource status matrix within the billing period. S200, parse the consolidated invoice file, break down the total amount of the consolidated invoice file into independent cost items, and match the corresponding data extraction operator according to the billing rules of the independent cost items; S300 applies the data extraction operator to the time-series resource status matrix, calculates the allocation weight and item amount of the business sub-entity under the independent cost item, and synchronously generates idle resource identifiers based on control plane status data and user plane load data. S400 aggregates the itemized amounts and performs a cash conservation check to generate a comprehensive billing data vector containing the business sub-entities; S500: Establish a directed acyclic graph with the total amount as the root node, connect the cost item nodes and business sub-sub-entity nodes corresponding to the independent cost items to the directed acyclic graph, map the item amount to the pixel width of the line connecting the directed acyclic graph, draw the node view in the business sub-sub-entity node area, and render the business sub-sub-entity node to the preset style according to the idle resource identifier. In step S200, matching the corresponding data extraction operator based on the billing rules for independent cost items includes: Under the condition that the billing rule belongs to the fixed occupation fee of the device port, the status detection logic operator is matched for the independent cost item; Under the condition that the billing rule is based on cumulative traffic consumption, match the integral accumulation logic operator for the independent cost item; Under the condition that the billing rule is based on the extreme value billing of data traffic generated within a specific time window, an extreme value filtering logic operator with a time delay parameter is set for matching independent cost items; In S300, the process of synchronously generating idle resource identifiers based on control plane status data and user plane payload data includes: Within the observation time window, count the number of consecutive sampling points where the control surface status data is valid and the user surface load data is equal to zero; If the number of consecutive sampling points reaches the judgment threshold, the physical port of the business sub-sub ... In step S300, applying the data extraction operator to the time-series resource state matrix to calculate the allocation weight and item amount of the business sub-entity under independent cost items further includes: For the extreme value filtering logic operator, the time-series resource state matrix is ​​accumulated along the column dimension to obtain the total concurrent consumption at each sampling time point; The quantile threshold is determined according to the quantile ratio, and the earliest time point when the total concurrent consumption is not less than the quantile threshold is retrieved as the extreme value timestamp. The time delay parameter is converted into the sampling point offset step number, and the local region matrix is ​​extracted with the extreme value timestamp as the center and the sampling point offset step number. Calculate the mean slice vector of each business sub-entity within the local region matrix, and obtain the third allocation weight based on the mean slice components; In S500, drawing a node view in the business sub-subject node area includes: In the view associated with the business sub-sub-node area, draw a time-series line chart with the sampling time point as the horizontal axis and the network resource consumption data value as the vertical axis; Extract the start and end times of the time window that triggers extreme value billing. Based on coordinate scaling, a rectangular vertical color band is drawn in the time series line chart, which runs through the vertical axis and covers the time window from the start time point to the end time point. In S500, rendering the business sub-sub-nodes to a preset style based on the idle resource identifier includes: Traverse the comprehensive billing data vector of the business sub-entities and extract the idle resource identifiers from the comprehensive billing data vector; When the idle resource identifier is active, the boundary frame of the business sub-sub-node will be changed from a solid line to a hollow dot matrix.

2. The method for automatic collection, visualization, and analysis of multi-entity telecommunications service invoicing data according to claim 1, characterized in that, In step S100, the step of aligning missing time-series values ​​of network resource consumption data to construct a time-series resource state matrix within the billing period includes: Based on the effective sample values ​​adjacent to the missing values ​​of continuous indicators in network resource consumption data and the proportion of time span, calculate the interpolated filling values ​​of continuous indicators. Extract the parameter status from the previous valid time point as the filling value for the status-type indicators in the network resource consumption data; The network resource consumption data and interpolated values ​​of the business sub-subjects, as well as the filled values ​​of status indicators, are arranged according to indicator type. A row dimension is constructed using the business sub-subject identifier, and a column dimension is constructed using time points to generate a time-series resource status matrix.

3. The method for automatic collection, visualization, and analysis of multi-entity telecommunications service invoicing data according to claim 1, characterized in that, In step S300, applying the data extraction operator to the time-series resource state matrix to calculate the allocation weight and item amount of the business sub-entity under independent cost items includes: For the state detection logic operator, determine the value of each business sub-entity having a valid network attachment state within the billing period, and determine the first apportionment weight; For the integral accumulation logic operator, during the billing period, the instantaneous consumption data of the row vectors corresponding to the business sub-sub ...

4. The method for automatic collection, visualization, and analysis of multi-entity telecommunications service invoicing data according to claim 1, characterized in that, In S400, the aggregation of itemized amounts and the execution of a cash conservation check include: Summarize the itemized amounts of all business sub-entities to generate the total output from the system, and calculate the absolute value of the error between the total output from the system and the total amount in the merged invoice files; If the absolute value of the error is less than the tolerance threshold, the independent cost item that caused the difference is located, and the value of the absolute value of the error is compensated to the item amount of the business sub-entity with the largest allocation weight under the independent cost item.

5. The method for automatic collection, visualization, and analysis of multi-entity telecommunications service invoicing data according to claim 1, characterized in that, In step S300, applying the data extraction operator to the time-series resource state matrix to calculate the allocation weight and item amount of the business sub-entity under independent cost items includes: In the operation of the data extraction operator, the calculation process of the data extraction operator is terminated when the denominator of the allocated weight is equal to zero. Trigger the abnormal rollback rule bound to the independent cost item, and perform static amortization of the independent cost item proportionally according to the number of business sub-entities.

Images