Process node processing efficiency monitoring method based on decision flow

Abstract

The invention belongs to the technical field of decision flow node efficiency monitoring, and particularly discloses a flow node processing efficiency monitoring method based on decision flow, which comprises the following steps of: performing efficiency rigidity analysis by collecting input data volume and processing time consumption of decision flow nodes to quantify the efficiency vulnerability of the nodes; meanwhile, on the basis of identifying efficiency fragile nodes based on efficiency rigidity, downstream nodes of each efficiency fragile node are determined according to a data transmission relation between the nodes, and an upstream and downstream association group is formed. By implementing bidirectional coupling evaluation of forward load conduction from upstream to downstream and reverse back pressure retardation from downstream to upstream on node pairs in the group, efficiency analysis can be deduced from time-consuming observation of a surface layer to causal logic deduction of a bottom layer, a risk conduction path is generated, an efficiency degradation propagation link is provided for operation and maintenance personnel, and the efficiency of the operation and maintenance personnel is improved. And the risk source node can be quickly positioned, so that the repair time is greatly shortened.

Description

Technical Field

[0001] This invention belongs to the field of decision process node efficiency monitoring technology, and specifically discloses a method for monitoring the processing efficiency of process nodes based on decision flow. Background Technology

[0002] As enterprises deepen their digital transformation, organizing business rules, decision nodes, and branch nodes through decision flow orchestration to automate business processes has become an important means of improving operational efficiency. However, modern enterprise decision flows are typically characterized by high complexity and large scale, often containing dozens or even hundreds of heterogeneous decision nodes. The processing efficiency of these nodes directly determines the response speed of the overall business process.

[0003] Existing methods for monitoring the processing efficiency of decision flow nodes primarily rely on threshold alarms for a single performance metric, such as triggering an alarm when the node's processing time reaches its limit. However, these methods have the following technical drawbacks: First, a single metric cannot comprehensively depict the true health status of a node. Node processing time is significantly affected by fluctuations in the amount of input data; focusing solely on processing time cannot distinguish whether a node's slowdown is due to an increase in input data volume or a decrease in its own processing capacity. For example, when the amount of input data surges, a moderate increase in processing time is normal, but using a fixed threshold alarm can easily generate a large number of false alarms. Conversely, when a node experiences performance degradation, processing time may gradually increase even with a constant amount of data, but a single metric cannot capture such subtle changes, leading to missed alarms.

[0004] Secondly, existing monitoring methods treat each node as an isolated entity, lacking correlation analysis of the mutual influence between nodes. In the decision-making process, nodes have data transmission relationships; efficiency fluctuations in upstream nodes can be transmitted to downstream nodes through data flow, and conversely, processing bottlenecks in downstream nodes can also affect upstream nodes through backpressure. Traditional single-point monitoring cannot identify this efficiency coupling relationship between nodes, making it difficult to locate the root cause node when risks propagate along the data flow, thus limiting the timeliness and accuracy of fault diagnosis. Summary of the Invention

[0005] To solve the above-mentioned technical problems, or at least partially solve them, the present invention provides a method for monitoring the processing efficiency of process nodes based on decision flow.

[0006] The objective of this invention can be achieved through the following technical solution: a process node processing efficiency monitoring method based on decision flow, comprising the following steps: real-time acquisition of the input data volume and processing time of each node in the decision flow within a continuous time window, and determination of the instantaneous efficiency stiffness of the node based on the degree of change of the input data volume and the degree of change of the processing time within each window.

[0007] The instantaneous efficiency stiffness of each node is compared with the efficiency stiffness benchmark based on historical data statistics to identify nodes with weak efficiency.

[0008] Based on the data flow direction between nodes defined in the decision requirements graph, the downstream nodes of each inefficiency-vulnerable node are determined, forming an upstream and downstream association group.

[0009] A bidirectional efficiency impact coupling analysis was performed on the node pairs in the upstream-downstream and downstream-upstream association groups, respectively.

[0010] Based on the results of the bidirectional efficiency impact coupling analysis, the efficiency coupling deterioration of node pairs in the upstream and downstream related groups is determined.

[0011] Nodes identified as being in a state of deteriorating efficiency coupling are aggregated according to data flow direction to form inefficient transmission paths, and hierarchical early warning information is generated based on path length and node deterioration type.

[0012] Combining all the above technical solutions, the positive effects of this invention are as follows: 1. This invention quantifies the efficiency vulnerability of nodes by collecting the input data volume and processing time of decision flow nodes to perform efficiency stiffness analysis. This efficiency assessment realizes the dynamic decoupling of performance monitoring and input load, which can maximize the accuracy of efficiency status assessment, thereby overcoming the limitations of solely relying on processing time for efficiency assessment under load fluctuations, and thus significantly reducing the missed detection rate of efficiency degradation caused by the high load masking effect.

[0013] 2. Based on the identification of efficiency-vulnerable nodes based on efficiency stiffness, this invention determines the downstream nodes of each efficiency-vulnerable node according to the data transmission relationship between nodes, forming an upstream-downstream association group. By implementing bidirectional coupling evaluation of positive load transmission from upstream to downstream and reverse back pressure blockage from downstream to upstream for nodes within the group, this invention effectively overcomes the limitations of isolated perspective and causal blind spots in single-point monitoring, enabling efficiency analysis to go from superficial time consumption observation to underlying causal logic deduction.

[0014] 3. This invention utilizes the results of bidirectional efficiency coupling analysis to form an inefficient transmission path, and generates hierarchical early warning information based on the path length and node deterioration type, providing maintenance personnel with an efficiency degradation propagation link, enabling them to quickly locate the risk source node, thereby significantly shortening the repair time and ensuring business continuity. Attached Figure Description

[0015] The present invention will be further described with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present invention. For those skilled in the art, other drawings can be obtained based on the following drawings without creative effort.

[0016] Figure 1This is a diagram illustrating the implementation steps of the method of the present invention.

[0017] Figure 2 This is a flowchart of the efficiency impact coupling analysis from upstream to downstream in this invention.

[0018] Figure 3 This is a flowchart of the efficiency impact coupling analysis from downstream to upstream in this invention. Detailed Implementation

[0019] 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.

[0020] See Figure 1 As shown, this invention proposes a process node processing efficiency monitoring method based on decision flow, including the following steps: S1, real-time acquisition of the input data volume and processing time of each node in the decision flow within a continuous time window, and determination of the instantaneous efficiency stiffness of the node based on the degree of change of the input data volume and the degree of change of the processing time within each window.

[0021] In enterprise decision-making processes, complex business processes are typically broken down into specific nodes, such as data cleaning nodes, model inference nodes, and external service call nodes. During operation, these nodes receive request data from upstream sources, perform calculations, and incur corresponding processing time. The longer the processing time, the lower the node's efficiency, which in turn slows down the overall business process response speed, resulting in longer user wait times.

[0022] Traditionally, node efficiency assessments often rely solely on processing time as a single outcome metric. However, processing time is essentially the result of the combined effects of input data and node performance: as the amount of input data increases, resource competition within the node intensifies and queue waiting times lengthen, inevitably leading to increased processing time. Judging solely by processing time fails to distinguish whether the increase stems from node performance degradation or simply from normal peak business loads, resulting in a misjudgment of the true efficiency degradation.

[0023] To address the aforementioned issues, this invention integrates the dynamic changes in the amount of input data at a node with the response fluctuations in processing time, introduces efficiency stiffness as an efficiency characterization index, and achieves dynamic decoupling between the node's internal processing capabilities and external input load by quantifying the sensitivity of processing time to changes in input data.

[0024] In a specific embodiment of the present invention, determining the instantaneous efficiency stiffness of a node includes the following: S11, for each node, calculating the total amount of input data received by the node and the average time spent processing the data within each sliding time window.

[0025] S12. Take the absolute value of the difference between the total amount of input data in the current time window and the previous adjacent time window, and divide it by the total amount of input data in the previous adjacent time window to obtain the input data change rate, which is used to characterize the degree of change of input data. Similarly, compare the average processing time of the current time window with the previous adjacent time window to calculate the processing time change rate, which is used to characterize the degree of change of processing response.

[0026] S13. Calculate the ratio of the rate of change of input data to the rate of change of processing time for each node in the current sliding time window to obtain the instantaneous efficiency stiffness of the node. This reflects the magnitude of the change in input data corresponding to a unit change in processing time. The higher the efficiency stiffness value, the lower the sensitivity of the node's processing time to changes in input data, that is, the smaller the change in processing time caused by a unit change in input data, and therefore the stronger the stability of the processing time. Conversely, it indicates that the node is sensitive to changes in input data and has the risk of efficiency vulnerability.

[0027] S2. Compare the instantaneous efficiency stiffness of each node with the efficiency stiffness benchmark based on historical data statistics to identify nodes with weak efficiency.

[0028] Since the introduced efficiency stiffness index can quantify the sensitivity of node processing time to changes in the amount of input data, and the intrinsic performance characteristics of nodes usually exhibit relative statistical stability and regularity in historical operation, it is possible to compare the instantaneous efficiency stiffness of nodes with the adaptive statistical benchmark generated based on historical data, thereby identifying nodes with low efficiency.

[0029] As one possible implementation method of this invention, the identification of efficiency-vulnerable nodes is carried out as follows: S21, collect node operation logs of the decision flow in multiple historical operation cycles. Considering that the input data volume of the enterprise decision flow is not randomly and uniformly distributed, but exhibits periodic tidal characteristics driven by business patterns such as morning and evening peak hours on weekdays, each historical operation cycle is divided into time segments that match the business tidal characteristics, such as the morning peak period from 09:00 to 11:00, the afternoon stable period from 14:00 to 16:00, and the nighttime trough period from 02:00 to 05:00. This ensures that subsequent efficiency comparisons are limited to homogeneous business scenarios, effectively eliminating misjudgments caused by differences in the amount of input data.

[0030] S22. Utilize the running logs of each node in each time segment to calculate the instantaneous efficiency stiffness of the historical time window, and aggregate them to form a historical efficiency stiffness dataset of the node in the corresponding time segment.

[0031] S23. Given that under stable system conditions, the efficiency stiffness of a node should follow a certain statistical distribution within a specific time segment, such as a normal distribution, a statistical feature of the efficiency stiffness distribution, such as a histogram, is constructed for each node's historical efficiency stiffness dataset within each time segment. The efficiency stiffness confidence interval at a pre-set confidence level (e.g., 95% confidence) is extracted, which can characterize the normal efficiency stiffness boundary of the node in that scenario. Since lower efficiency stiffness means the node is more sensitive to increases in input data volume—that is, a small increase in data volume leads to a sharp increase in processing time—it represents greater fragility. Therefore, the lower limit of the confidence interval represents the acceptable critical efficiency stiffness for the corresponding time segment. This lower limit is used as the efficiency stiffness benchmark for the node in that time segment. Once the efficiency stiffness falls below the benchmark, it indicates that the node's resilience has fallen below the historical normal level, posing a risk of efficiency degradation.

[0032] S24. Real-time acquisition of the current time segment and retrieval of the corresponding efficiency stiffness benchmark. Considering that instantaneous jitters such as network interruptions are inevitable during the decision flow operation, these are usually instantaneous, while true performance degradation often manifests as a continuous state deviation. Therefore, by introducing the observation window length as a duration threshold, typically the length of two sliding time windows can be taken. If the current instantaneous efficiency stiffness of a node is lower than the benchmark and the duration exceeds the observation window length, the node is determined to be an efficiency-vulnerable node. This can filter out occasional anomalies to a certain extent and improve the signal-to-noise ratio and accuracy of efficiency-vulnerable node determination.

[0033] S3. Based on the data flow direction between nodes defined in the decision requirements diagram, determine the downstream nodes of each efficiency-vulnerable node to form an upstream and downstream association group.

[0034] After initially identifying efficiency-vulnerable nodes based on single-point efficiency stiffness analysis, given the data flow relationship between nodes in the decision flow, the efficiency degradation of upstream nodes often generates a cascading transmission effect through the data link, leading to a decrease in the operating efficiency of downstream nodes. In order to overcome the limitations of single-node efficiency analysis and track the transmission of node efficiency degradation, this invention aggregates the identified efficiency-vulnerable nodes and their direct downstream nodes to construct upstream and downstream association groups to carry out subsequent efficiency impact coupling analysis.

[0035] In a preferred embodiment, the upstream and downstream association group is composed of the following: extracting the direct downstream nodes of each inefficiency-vulnerable node according to the data flow direction between nodes defined in the decision demand graph.

[0036] Each inefficiency-vulnerable node is grouped with its direct downstream nodes into an upstream-downstream association group.

[0037] Furthermore, given that the topology of the decision flow has multi-path distribution characteristics, the data flow between nodes is not a single linear link, and the amount of data carried by different downstream nodes is often uneven. If all downstream nodes are analyzed in parallel, it will consume a lot of computing resources and cause delays in the analysis of high-priority paths. In this context, the downstream nodes are sorted from largest to smallest according to the proportion of data received. The larger the proportion of data received, the higher the dependence and coupling of the downstream node on the upstream node. The wider the coverage of business interruption caused by the deterioration of upstream efficiency and the greater the potential loss, the more important it is to analyze such related groups first.

[0038] Based on the sorting results, the inefficient vulnerable nodes are paired with each direct downstream node to form multiple upstream and downstream association groups. Each association group is marked with a priority label, which can ensure that, under the condition of limited resources, the inefficient transmission path with the largest impact and the highest business risk is locked first.

[0039] S4. Perform bidirectional efficiency impact coupling analysis on the node pairs in the upstream-downstream and downstream-upstream association groups, respectively.

[0040] When constructing upstream and downstream related groups and performing efficiency impact coupling analysis, given that the interaction between nodes is not a one-way transmission, downstream nodes will not only passively bear the input delay from upstream nodes, but will also trigger back pressure due to their own processing capacity bottlenecks, thereby generating a reverse blocking effect on upstream nodes. Therefore, the efficiency impact coupling analysis carried out in this invention adopts a two-way perspective from upstream to downstream and from downstream to upstream, which can more comprehensively understand the transmission of efficiency degradation between nodes.

[0041] See Figure 2 As shown, the first type of efficiency influence coupling analysis applied to the above scheme is implemented as follows: (11) For each node in the upstream and downstream association group, record its instantaneous efficiency stiffness sequence in a continuous time window, and at the same time record the data flow from the upstream node to the downstream node in each time window.

[0042] (12) Considering that when the downstream node passively bears the input delay from the upstream node, the efficiency degradation time series of the upstream node and the efficiency degradation time series of the downstream node exhibit a co-evolutionary characteristic. Since it takes physical time to propagate from the upstream node to the downstream node, the efficiency degradation of the two is not completely synchronized on the time axis, but there is a time lag. Therefore, the maximum exploration lag step size is set (usually taken as 1.2 to 1.5 times the maximum network delay in the link theory), and the lag step size is traversed to calculate the cross-correlation coefficient between the instantaneous efficiency stiffness sequence of the upstream node and the instantaneous efficiency stiffness sequence of the downstream node. This coefficient is used to quantify the statistical correlation strength of the efficiency fluctuation of the upstream and downstream nodes under different time lags.

[0043] (13) Among the cross-correlation numbers under different lag step lengths, if there is a cross-correlation number under a certain lag step length that exceeds the positive coupling threshold, it indicates that there is a statistical correlation of efficiency degradation between the upstream node and the downstream node. On this basis, it is necessary to verify whether the physical time corresponding to the lag step length is less than the maximum tolerable transmission delay. This delay reflects the theoretical maximum physical time limit required for the fault signal to be transmitted from the upstream node to the downstream node under normal network topology and protocol mechanism. Correspondence exceeding this time limit is considered as indirect causal transmission. If the physical time corresponding to the lag step length is less than the maximum tolerable transmission delay, it is preliminarily determined that there is a positive efficiency coupling effect.

[0044] (14) Given that when upstream and downstream efficiency deterioration occurs in coordination, the decline in the processing efficiency of the upstream node as the data source will directly lead to a decrease in output capacity, which in turn will cause a decrease in the amount of data flowing downstream. This decrease in data volume constitutes physical evidence of insufficient upstream power supply. Therefore, within the lag window where a positive coupling effect is initially determined, the data flow volume of the current window is compared with the data flow volume of the previous adjacent window to calculate the decrease in data flow. The decrease in data flow can be calculated by dividing the difference between the data flow volume of the previous adjacent window and the current window by the data flow volume of the previous adjacent window. If the decrease in data flow exceeds the flow attenuation limit, it indicates that the data link has been blocked. At this time, it is confirmed that there is a positive efficiency coupling effect. This step uses the change in data flow volume as a secondary verification to ensure that the observed correlation is not just a numerical coincidence, but is supported by real data flow obstruction.

[0045] In the implementation of positive efficiency impact coupling analysis, the positive coupling threshold is a critical value used to determine whether there is a statistical correlation between the efficiency degradation sequences of upstream and downstream nodes. Since the cross-correlation coefficient typically ranges from -1 to 1, the focus in the positive propagation scenario is on positive correlation. Therefore, the positive coupling threshold is usually set to a high value between (0, 1), such as 0.7.

[0046] The flow attenuation limit is a critical proportion for determining whether the output data volume of the upstream node is blocked. For example, it can be determined by statistically analyzing the 95th percentile of the decrease in data flow volume of adjacent windows within a historical normal operation cycle, and is used to exclude normal fluctuation interference.

[0047] See Figure 3As shown, the second type of efficiency impact coupling analysis applied to the above scheme is implemented as follows: (21) When the downstream node's processing capacity is insufficient, the queue backlog length increases, and the upstream node enters a resource waiting state through counter-pressure, causing the upstream node's blocking waiting time to increase. Therefore, for the node pairs in the upstream and downstream association group, the queue backlog length of the downstream node in each time window and the blocking waiting time of the upstream node due to waiting for the downstream node to release resources are collected, and the normalized product of the queue backlog length and the blocking waiting time is used as the downstream blocking degree. This index describes the intensity of the reverse pressure exerted by the downstream on the upstream by combining the congestion space scale and the congestion duration. The normalization is introduced to eliminate the difference in the units of the original data, and the product form has a nonlinear suppression effect. Only when the spatial congestion and the time stagnation are both at a high level will the downstream blocking degree increase significantly, effectively filtering out the pseudo counter-pressure scenario where there is only backlog but no blocking or only blocking time but no backlog.

[0048] (22) For upstream and downstream related groups, the instantaneous efficiency stiffness sequence of upstream nodes, the blockage sequence of downstream nodes, and the local data backlog sequence of upstream nodes are recorded synchronously within a continuous time window. The local data backlog is used to verify whether the back pressure is truly effective. If the upstream data inflow is greater than the downstream data outflow and there is blockage, the flow difference will inevitably be converted into the upstream local data backlog increment.

[0049] (23) When back pressure occurs from downstream to upstream, the downstream blockage degree increases, and the upstream node, due to frequent entry into a blocking waiting state, should show a decreasing trend in its instantaneous efficiency stiffness. Therefore, the downstream blockage degree sequence and the upstream efficiency stiffness sequence exhibit a negative correlation time series characteristic, that is, the higher the downstream blockage, the lower the upstream efficiency. Since the transmission of back pressure signal takes time, the upstream only shows efficiency degradation after the downstream blockage lasts for a period of time. In order to capture this reverse causal lag effect, by setting the maximum exploration lag step, the cross-correlation coefficient between the downstream node blockage degree sequence and the upstream node instantaneous efficiency stiffness sequence under different lag step lengths is calculated sequentially. This coefficient is used to quantify the statistical correlation strength between the downstream back pressure behavior and the upstream efficiency degradation under different time lags.

[0050] (24) If there is a cross-correlation coefficient less than the reverse coupling threshold under a certain lag step and the physical time corresponding to the lag step is less than the maximum response delay, where the maximum response delay refers to the theoretical maximum time taken for the back pressure signal to be transmitted to the upstream node after the downstream blockage occurs, it indicates that there is a strong statistical negative correlation between the upstream and downstream, which is consistent with the timing logic of the downstream being blocked first and the upstream being slow. Moreover, the delay of the correlation is within the physical reach range, excluding the accidental coincidence under long time delay. At this time, it is preliminarily determined that there is a reverse efficiency coupling effect.

[0051] (25) Given that relying solely on statistical correlation may be affected by external factors and produce statistical spurious correlation, within the lag window where it is initially determined that there is a reverse coupling effect, the difference between the local data backlog of the upstream node in the current window and the previous adjacent window is calculated and then divided by the local data backlog of the previous adjacent window to obtain the data backlog increase rate. This is used as a physical empirical verification. If it is confirmed to be a real back pressure transmission, the upstream node will be unable to send data to the downstream, which will lead to an increase in local data backlog. If the data backlog increase rate exceeds the backlog deterioration limit, it confirms that the observed correlation has a physical material basis and effectively eliminates spurious correlation caused by non-back pressure factors. At this time, it is confirmed that there is a reverse efficiency coupling effect.

[0052] In the implementation of reverse efficiency impact coupling analysis, the focus in the reverse propagation scenario is on negative correlation. Therefore, the reverse coupling threshold is usually set as a negative number with a large absolute value within the interval (-1, 0) (such as -0.7) to define the critical level of strong negative correlation. Since in the negative number domain, the smaller the value, the stronger the negative correlation, and the larger the value, the weaker the negative correlation, the judgment logic follows the principle of being less than the reverse coupling threshold.

[0053] The backlog worsening limit is the critical increase in the amount of local data backlog at the upstream node to determine whether there is an abnormal surge. Specifically, it can be determined by the 95th percentile of the increase in backlog between adjacent windows during the historical normal operation cycle.

[0054] It should be noted that in this invention, both cross-correlation coefficient and cross-correlation coefficient refer to the correlation coefficient after normalization. They are calculated in the same way, both being generalizations of the Pearson correlation coefficient under lag. To facilitate the distinction between different analytical objects, the lag correlation analysis of homogeneous variables (efficiency stiffness and efficiency stiffness) is called the cross-correlation coefficient, while the lag correlation analysis of heterogeneous variables (blockage degree and efficiency stiffness) is called the cross-correlation coefficient, thus reflecting the differences in the analytical objects.

[0055] S5. Based on the bidirectional efficiency impact coupling analysis results, determine the efficiency coupling deterioration of node pairs in the upstream and downstream association groups, including the following: If there is only a positive efficiency coupling impact in an upstream and downstream association group, then the downstream node is determined to be in a passive coupling deterioration state affected by the upstream transmission.

[0056] If a certain upstream and downstream related group has both positive efficiency coupling effects and negative efficiency coupling effects, then the upstream and downstream nodes are determined to be in a state of mutually aggravated bidirectional coupling deterioration.

[0057] It should be noted that, given the unidirectional transmission principle from upstream to downstream between decision flow nodes, reverse efficiency coupling is essentially a feedback-driven blockade of the upstream's positive output behavior from the downstream. This means that the existence of reverse coupling requires the active flow of positive data as a physical prerequisite. Physically, pure reverse efficiency coupling cannot occur independently of the positive transmission mechanism. Therefore, in practical decision logic, a situation where only reverse efficiency coupling has an impact is theoretically unattainable; hence, this scheme does not establish such an independent decision branch.

[0058] S6. Aggregate nodes that are determined to be in a state of deteriorating efficiency coupling according to the data flow direction to form inefficient transmission paths, and generate hierarchical early warning information based on path length and node deterioration type.

[0059] In a preferred embodiment of the present invention, the formation of an inefficient transmission path in the above steps includes the following: S61, using a node in a passive coupling deterioration state or a bidirectional coupling deterioration state as a path seed node.

[0060] S62. Starting from the path seed node, backtrack upstream in reverse order of data transmission direction. If an adjacent upstream node is in any of the following states: efficiency vulnerability, passive coupling deterioration, or bidirectional coupling deterioration, then include that upstream node in the current path and continue backtracking upstream until a node not determined to be in any deterioration state is encountered or the decision flow source is reached, thus determining the starting boundary of the path.

[0061] S63. Starting from the path seed node, traverse downstream in the order of data transmission. If the adjacent downstream node is in a passive coupling deterioration state or a bidirectional coupling deterioration state, then include the downstream node in the current path and continue traversing downstream until a node that has not been determined to be in any deterioration state is encountered or the end of the decision flow is reached, and the termination boundary of the path is determined.

[0062] S64. Record the sequence of consecutive nodes between the starting boundary and the ending boundary as an inefficient propagation path, and record the starting node identifier and the list of nodes passed through the path.

[0063] The inefficient transmission paths formed by the above operations ensure that each generated path represents a continuous and uninterrupted transmission process of efficiency degradation.

[0064] To proactively intervene in the spread of efficiency degradation, after constructing the inefficient transmission path, a tiered early warning operation is initiated. The specific details are as follows: the total number of nodes in each inefficient transmission path is counted as the path length, which is used to characterize the spatial coverage of the efficiency degradation impact, and the deterioration type of each node in the path is summarized.

[0065] If there are nodes in the path that are in a bidirectional coupling deterioration state, it means that the upstream and downstream nodes are trapped in a positive feedback vicious cycle. This state has strong self-reinforcing and instability, and is very likely to cause decision flow deadlock in a short period of time. Under this condition, regardless of the length of the path, it is considered to be a high-risk level.

[0066] For high-risk levels, data flow between nodes can be immediately cut off.

[0067] If a path contains only nodes in a passively coupled and deteriorated state, and the path length exceeds a length threshold (e.g., 3 nodes), it indicates that although the transmission between nodes is only unidirectional, the long path suggests that efficiency degradation has cascaded and spread over a large scale. If not addressed promptly, the number of affected business nodes will significantly increase, requiring close attention. In this case, the path should be marked as medium-risk.

[0068] For medium-risk levels, traffic scheduling is initiated to prioritize the protection of end nodes in the link, and the operations and maintenance team is notified to conduct root cause investigation to prevent the spread from expanding further.

[0069] If the path contains only nodes in a passively coupled and deteriorated state and the path length is less than the length threshold, it indicates that the impact of efficiency degradation is limited and the evolution mode is unidirectional, and the overall stability is still controllable. In this case, the path is marked as low-risk.

[0070] For low-risk levels, they are added to a watchlist, allowed to retry, and only escalated to higher levels after the duration expires, to prevent operational resources from being overwhelmed by minor fluctuations.

[0071] The path identifier, path length, node deterioration type, and risk level are combined to generate graded early warning information.

[0072] By implementing the aforementioned tiered early warning system, inefficient transmission paths are transformed into actionable decision-making instructions. This not only solves the problem of traditional alarms failing to differentiate between priorities due to a lack of context, but also ensures that the most destructive efficiency risks are addressed first, given limited operational resources.

[0073] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0074] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0075] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0076] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0077] Finally, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for monitoring the processing efficiency of process nodes based on decision flow, characterized in that, Includes the following steps: The system collects the amount of input data and the time spent processing data for each node in the decision flow within a continuous time window in real time. Based on the degree of change in the amount of input data and the degree of change in the time spent processing data within each window, the instantaneous efficiency stiffness of the node is determined. The instantaneous efficiency stiffness of each node is compared with the efficiency stiffness benchmark based on historical data statistics to identify nodes with weak efficiency. Based on the data flow direction between nodes defined in the decision requirements diagram, the downstream nodes of each efficiency-vulnerable node are determined, forming an upstream and downstream association group; A bidirectional efficiency impact coupling analysis was performed on the node pairs in the upstream-downstream and downstream-upstream directions, respectively. Based on the results of the bidirectional efficiency impact coupling analysis, the efficiency coupling deterioration of node pairs in the upstream and downstream related groups is determined. Nodes identified as being in a state of deteriorating efficiency coupling are aggregated according to data flow direction to form inefficient transmission paths, and hierarchical early warning information is generated based on path length and node deterioration type.

2. The method for monitoring the processing efficiency of process nodes based on decision flow as described in claim 1, characterized in that: The instantaneous efficiency stiffness of the determined node includes the following: Within each sliding time window, the total amount of input data received and the average time spent processing the data are statistically analyzed for each node. Compare the total amount of input data in the current time window with the previous adjacent time window to calculate the rate of change of input data, and compare the average processing time of the current time window with the previous adjacent time window to calculate the rate of change of processing time. The ratio of the rate of change of input data volume to the rate of change of processing time is defined as the instantaneous efficiency stiffness of the node in the current time window.

3. The method for monitoring the processing efficiency of process nodes based on decision flow as described in claim 1, characterized in that: The process for identifying nodes with low efficiency is as follows: Collect node operation logs of the decision flow in multiple historical operation cycles, and divide each historical operation cycle into several time segments according to the business tidal characteristics; By utilizing the running logs of each node in each time segment, the instantaneous efficiency stiffness of the historical time window is calculated, and the logs are aggregated to form a historical efficiency stiffness dataset of the node in the corresponding time segment. For each node's historical efficiency stiffness dataset in each time segment, construct statistical features of efficiency stiffness distribution, extract confidence intervals of efficiency stiffness at a pre-set confidence level, and use the lower limit of the confidence interval as the efficiency stiffness benchmark of the node in the corresponding time segment. The system obtains the current time segment in real time, retrieves the corresponding efficiency stiffness benchmark, and determines that a node is an efficiency-vulnerable node if the current instantaneous efficiency stiffness of a node is lower than the benchmark and the duration exceeds the observation window length.

4. The method for monitoring the processing efficiency of process nodes based on decision flow as described in claim 1, characterized in that: The upstream and downstream related groups are constructed as follows: Based on the data flow between nodes defined in the decision requirements graph, extract the direct downstream nodes of each inefficiency-vulnerable node; Each inefficiency-vulnerable node is grouped with its direct downstream nodes into an upstream-downstream association group.

5. The method for monitoring the processing efficiency of process nodes based on decision flow as described in claim 1, characterized in that: The upstream and downstream association group also includes the following formation process: When a vulnerable node has multiple direct downstream nodes, they are sorted from largest to smallest according to the proportion of data received by each downstream node. Based on the sorting results, the efficiency-vulnerable nodes are paired with each direct downstream node to form multiple upstream and downstream association groups, and each association group is marked with a priority label.

6. The method for monitoring the processing efficiency of process nodes based on decision flow as described in claim 1, characterized in that: The upstream-to-downstream efficiency impact coupling analysis in the bidirectional efficiency impact coupling analysis is implemented as follows: For node pairs in the upstream and downstream association group, extract the instantaneous efficiency stiffness sequences of the upstream and downstream nodes in multiple consecutive time windows, and record the data flow from the upstream node to the downstream node in each time window. Set a maximum exploration lag step size, and calculate the cross-correlation coefficient between the instantaneous efficiency stiffness sequence of the upstream node and the instantaneous efficiency stiffness sequence of the downstream node under different lag step sizes; If the cross-correlation coefficient at a certain lag step exceeds the positive coupling threshold, and the physical time corresponding to that lag step is less than the maximum tolerable propagation delay, then it is preliminarily determined that there is a positive efficiency coupling effect. Within the lag window where a positive coupling effect is initially determined, the data flow rate of the current window is compared with the data flow rate of the previous adjacent window, and the decrease in data flow rate is calculated. If the decrease in data flow rate exceeds the flow attenuation limit, a positive efficiency coupling effect is confirmed.

7. The method for monitoring the processing efficiency of process nodes based on decision flow as described in claim 6, characterized in that: The downstream-to-upstream efficiency impact coupling analysis in the bidirectional efficiency impact coupling analysis is implemented as follows: For node pairs in the upstream and downstream association groups, the queue backlog length of the downstream node in each time window and the blocking waiting time of the upstream node due to waiting for the downstream node to release resources are collected, and the normalized product of the queue backlog length and the blocking waiting time is used as the downstream blocking degree. For upstream and downstream related groups, the instantaneous efficiency stiffness sequence of upstream nodes, the blockage degree sequence of downstream nodes, and the local data backlog sequence of upstream nodes are recorded synchronously within a continuous time window. Set a maximum exploration lag step size, and calculate the cross-correlation coefficient between the downstream node blockage degree sequence and the upstream node instantaneous efficiency stiffness sequence under different lag step sizes; If the cross-correlation coefficient at a certain lag step is less than the reverse coupling threshold, and the physical time corresponding to that lag step is less than the maximum response delay, then it is preliminarily determined that there is a reverse efficiency coupling effect. Within the lag window where a reverse coupling effect is initially determined, the local data backlog of the upstream node in the current window is compared with that of the previous adjacent window to calculate the increase in data backlog. If the increase in data backlog exceeds the backlog deterioration limit, then a reverse efficiency coupling effect is confirmed.

8. The method for monitoring the processing efficiency of process nodes based on decision flow as described in claim 7, characterized in that: The determination of efficiency coupling degradation refers to the following process: If a certain upstream and downstream related group only has a positive efficiency coupling effect, then the downstream node is determined to be in a passive coupling deterioration state affected by the upstream transmission. If a certain upstream and downstream related group has both positive efficiency coupling effects and negative efficiency coupling effects, then the upstream and downstream nodes are determined to be in a state of mutually aggravated bidirectional coupling deterioration.

9. The method for monitoring the processing efficiency of process nodes based on decision flow as described in claim 1, characterized in that: The inefficient conduction path includes the following formation process: Nodes in a state of deteriorated passive coupling or deteriorated bidirectional coupling are used as path seed nodes. Starting from the seed node of the path, backtrack upstream in reverse order of the data transmission direction. If an adjacent upstream node is in any of the following states: inefficiently vulnerable state, passively coupled deteriorated state, or bidirectionally coupled deteriorated state, then the upstream node is included in the current path, and backtracking continues upstream until a node not determined to be in any deteriorated state is encountered or the source of the decision flow is reached, thus determining the starting boundary of the path. Starting from the path seed node, traverse downstream in the order of data transmission. If an adjacent downstream node is in a passive coupling deterioration state or a bidirectional coupling deterioration state, then include the downstream node in the current path and continue traversing downstream until a node that has not been determined to be in any deterioration state is encountered or the end of the decision flow is reached, and the termination boundary of the path is determined. A sequence of consecutive nodes between the starting and ending boundaries is recorded as an inefficient propagation path, and the starting node identifier and the list of nodes passed through the path are recorded for this path.

10. The method for monitoring the processing efficiency of process nodes based on decision flow as described in claim 1, characterized in that: The generation process for the tiered early warning information is as follows: The total number of nodes in each inefficient propagation path is counted as the path length, and the deterioration type of each node in the path is summarized. If there are nodes in the path that are in a state of deteriorated bidirectional coupling, then the path is marked as high-risk. If a path contains only nodes in a passively coupled deteriorating state and the path length exceeds a length threshold, then the path is marked as a medium-risk level. If a path contains only nodes in a passively coupled deteriorated state and the path length is less than the length threshold, then the path is marked as low-risk. The path identifier, path length, node deterioration type, and risk level are combined to generate graded early warning information.

Images