Processing method and processing device for dynamic threshold adjustment, equipment and medium

Abstract

The embodiment of the invention discloses a processing method and a processing device for dynamic threshold adjustment, equipment and a medium, which can obtain a target risk assessment result of a to-be-assessed project and a target fluctuation assessment result of distribution characteristics of risk weights, and when the target risk assessment result is that the running state of the to-be-assessed project is in a continuous abnormal state, the to-be-assessed project is subjected to dynamic threshold adjustment. And when the target fluctuation evaluation result is that the current risk weight is in an abnormal fluctuation state, performing semantic correction on the current risk threshold. The dynamic change of the operating environment of the project to be evaluated is found accurately and timely, and the risk threshold is corrected, so that the risk threshold can adapt to the dynamic change of the project environment, the risk missing report is reduced, and the response of a computer system is prevented from being slowed down.

Description

Technical Field

[0001] This invention relates to the field of intelligent project management technology, and in particular to a processing method, processing device, equipment and medium for dynamic threshold adjustment. Background Technology

[0002] Existing project risk response threshold management methods often employ static thresholds. Static thresholds are fixed and cannot adapt to dynamic changes in the project environment. When risk situations are complex and varied while thresholds remain fixed, the computer system may need to constantly reassess and recalculate various risk factors to determine whether to trigger risk response measures. This process consumes system computing resources, leading to slower system response. For example, in a software development project, the development environment and requirements are constantly changing, and static thresholds cannot reflect new risk situations in a timely manner. The computer needs to frequently perform risk assessments for new code changes, requirement changes, and other factors, which will slow down the system's response when processing other tasks, such as extending the time for opening files and running programs.

[0003] This not only leads to a higher rate of false negatives, but also slows down system response. Summary of the Invention

[0004] The main objective of this invention is to provide a dynamic threshold adjustment processing method, processing device, equipment, and medium, which can solve the problems in the prior art where static thresholds lead to a high false negative rate and slow system response.

[0005] To achieve the above objectives, a first aspect of the present invention provides a method for dynamic threshold adjustment, the method comprising:

[0006] Obtain the risk weights, risk thresholds, and preset risk weight distribution thresholds for the project under evaluation, based on its current operational status.

[0007] The risk weights and a preset distribution feature extraction algorithm are used to extract features to obtain the distribution features of the risk weights. The distribution features are used to reflect the fluctuation of the risk weights of the project to be evaluated.

[0008] Based on the risk threshold and risk weight, a risk assessment is performed to obtain the target risk assessment result of the project to be assessed.

[0009] Based on the distribution characteristics and distribution threshold, a volatility assessment is performed to determine the target volatility assessment result for the distribution characteristics.

[0010] If the target risk assessment result indicates that the operating status of the project to be assessed is in a continuous abnormal state, and the target fluctuation assessment result indicates that the risk weight is in an abnormal fluctuation state, then based on the target risk assessment result and the target fluctuation assessment result, target early warning information is output to a preset terminal, and semantic analysis processing is performed using the target early warning information and the preset semantic analysis model to obtain a threshold correction value.

[0011] The threshold is corrected using the threshold correction value and the risk threshold to obtain the corrected risk threshold, and then the process returns to the step of obtaining the risk weight, risk threshold and preset risk weight distribution threshold of the current operating status of the project to be evaluated.

[0012] In one feasible implementation, obtaining the risk threshold under the current operating state includes:

[0013] Obtain the preset baseline risk threshold and the status parameters under the current operating state;

[0014] The risk threshold is obtained using the state parameters, the baseline risk threshold, and a preset dynamic threshold calibration algorithm.

[0015] In one feasible implementation, after obtaining the preset baseline risk threshold, the method further includes:

[0016] Obtain a preset stratification threshold, which includes a first threshold and a second threshold, wherein the first threshold is greater than the second threshold;

[0017] The target comparison result is obtained by comparing the risk weight with the first threshold and the second threshold respectively.

[0018] If the target comparison result is that the risk weight is greater than the first threshold, then the benchmark risk threshold is increased according to a preset first proportional coefficient to obtain the increased benchmark risk threshold;

[0019] If the target comparison result is that the risk weight is less than the second threshold, then the benchmark risk threshold is reduced according to the preset second proportional coefficient to obtain the reduced benchmark risk threshold.

[0020] In one feasible implementation, the target early warning information includes at least a risk description text and an occurrence time series of the current risk event, and the semantic analysis model includes a BERT model and an LSTM model. Then, the step of performing semantic analysis processing using the target early warning information and the preset semantic analysis model to obtain the threshold correction value includes:

[0021] The risk description text is semantically analyzed using the BERT model to obtain its semantic features.

[0022] The occurrence time series is analyzed using the LSTM model to obtain time series characteristics;

[0023] The threshold correction value is obtained by weighting the text semantic features and temporal features.

[0024] In one feasible implementation, the preset distribution feature extraction algorithm includes a standard deviation algorithm. Then, the step of using the risk weights and the preset distribution feature extraction algorithm to extract features and obtain the distribution features of the risk weights includes:

[0025] Using a preset time window and the first time point corresponding to the risk weight, a risk weight sequence is obtained. The risk weight sequence includes the risk weight and the historical risk weights of each historical second time point within the time window before the first time point.

[0026] The standard deviation of the risk weights is obtained using the risk weight sequence and the standard deviation algorithm, and the distribution characteristics include the standard deviation.

[0027] In one feasible implementation, the state parameters include schedule deviation rate, resource volatility rate, learning rate coefficient, resource sensitivity coefficient, and environmental complexity factor. The step of obtaining the risk threshold using the state parameters, a baseline risk threshold, and a preset dynamic threshold calibration algorithm includes:

[0028] The schedule deviation rate, resource volatility, learning rate coefficient, resource sensitivity coefficient, environmental complexity factor, and baseline risk threshold are input into the dynamic threshold calibration algorithm to obtain the risk threshold.

[0029] In one feasible implementation, the step of performing risk assessment based on the risk threshold and risk weight to obtain the target risk assessment result of the project to be assessed includes:

[0030] If the risk weight is greater than or equal to the risk threshold, the preliminary risk assessment result is determined to be an abnormal state, and the number of consecutive abnormalities N is incremented by 1, with N initially set to 0.

[0031] Determine whether N is greater than or equal to a preset number of times threshold;

[0032] If N is greater than or equal to a preset number of times threshold, then the target risk assessment result is determined to be that the operating status of the project to be assessed is in a continuous abnormal state;

[0033] If N is less than a preset number of times threshold, then the target risk assessment result is determined to be that the operating status of the project to be assessed is in a discontinuous abnormal state.

[0034] If the risk weight is less than the risk threshold, the preliminary risk assessment result is determined to be in a normal state, and N = 0 is set.

[0035] To achieve the above objectives, a second aspect of the present invention provides a processing apparatus for dynamic threshold adjustment, the apparatus comprising:

[0036] Data acquisition unit: used to acquire the risk weights, risk thresholds, and preset risk weight distribution thresholds of the project under evaluation in its current operating status;

[0037] Feature extraction unit: used to extract features using the risk weights and a preset distribution feature extraction algorithm to obtain the distribution features of the risk weights, which are used to reflect the fluctuation of the risk weights of the project to be evaluated;

[0038] First assessment unit: used to conduct risk assessment based on the risk threshold and risk weight, and obtain the target risk assessment result of the project to be assessed;

[0039] The second evaluation unit is used to perform fluctuation evaluation based on the distribution characteristics and distribution threshold, and to determine the target fluctuation evaluation result of the distribution characteristics.

[0040] Semantic analysis unit: If the target risk assessment result indicates that the operating status of the project to be assessed is in a continuous abnormal state, and the target fluctuation assessment result indicates that the risk weight is in an abnormal fluctuation state, then based on the target risk assessment result and the target fluctuation assessment result, it outputs target early warning information to a preset terminal, and uses the target early warning information and a preset semantic analysis model to perform semantic analysis processing to obtain a threshold correction value.

[0041] Threshold correction unit: Used to correct the threshold using the threshold correction value and the risk threshold to obtain the corrected risk threshold, and return to execute the step of obtaining the risk weight, risk threshold and preset risk weight distribution threshold of the current operating status of the project to be evaluated.

[0042] To achieve the above objectives, a third aspect of the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps shown in the first aspect and any feasible implementation.

[0043] To achieve the above objectives, a fourth aspect of the present invention provides a computer device including a memory and a processor, the memory storing a computer program, which, when executed by the processor, causes the processor to perform the steps shown in the first aspect and any feasible implementation.

[0044] The embodiments of the present invention have the following beneficial effects:

[0045] This invention provides a method for dynamic threshold adjustment, comprising: obtaining the risk weight, risk threshold, and preset risk weight distribution threshold of the project under current operating status; performing feature extraction using the risk weight and a preset distribution feature extraction algorithm to obtain the distribution feature of the risk weight, which reflects the fluctuation of the risk weight of the project under evaluation; performing risk assessment based on the risk threshold and risk weight to obtain the target risk assessment result of the project under evaluation; performing fluctuation assessment based on the distribution feature and distribution threshold to determine the target fluctuation assessment result of the distribution feature; if the target risk assessment result indicates that the operating status of the project under evaluation is in a continuous abnormal state, and the target fluctuation assessment result indicates that the risk weight is in an abnormal fluctuation state, then outputting a target early warning message to a preset terminal based on the target risk assessment result and the target fluctuation assessment result, and performing semantic analysis processing using the target early warning message and a preset semantic analysis model to obtain a threshold correction value; performing threshold correction using the threshold correction value and the risk threshold to obtain the corrected risk threshold, and returning to execute the step of obtaining the risk weight, risk threshold, and preset risk weight distribution threshold of the project under current operating status.

[0046] The above method yields the target risk assessment results and the target fluctuation assessment results of the risk weight distribution characteristics of the project under evaluation. When the target risk assessment result indicates that the project's operating status is in a continuous abnormal state, and the target fluctuation assessment result indicates that the current risk weight is in an abnormal fluctuation state, semantic corrections are made to the current risk threshold. This enables accurate and timely detection of dynamic changes in the operating environment of the project under evaluation, allowing for adjustments to the risk threshold to adapt to these changes, reducing underreporting of risks, and preventing slowdowns in computer system response. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] in:

[0049] Figure 1 This is a flowchart of a dynamic threshold adjustment processing method according to an embodiment of the present invention;

[0050] Figure 2 This is a structural block diagram of a dynamic threshold adjustment processing device according to an embodiment of the present invention;

[0051] Figure 3 This is a structural block diagram of a computer device in an embodiment of the present invention. Detailed Implementation

[0052] 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.

[0053] Please see Figure 1 , Figure 1 This is a flowchart of a dynamic threshold adjustment processing method according to an embodiment of the present invention. This method can be applied to both terminals and servers. The terminal can be a desktop terminal or a mobile terminal; a mobile terminal can be at least one of a mobile phone, tablet computer, or laptop computer. The server can be a standalone server or a server cluster composed of multiple servers. This embodiment uses a terminal application as an example. Figure 1 The method shown includes the following steps:

[0054] 101. Obtain the risk weights, risk thresholds, and preset risk weight distribution thresholds for the project under current operating status;

[0055] It should be noted that the project to be evaluated (hereinafter referred to as the project) can be an IT project or an AI project, such as an e-commerce platform development project or a medical AI development project, etc.

[0056] Among them, the risk weight w i The essence of risk weighting is the quantification of a project's risk at a specific point in time, reflecting the severity of specific risk events; for example: requirement changes (weight 0.35), technical debt (weight 0.28), etc. Risk weight w i Source: Risk event data collected in real time through project management tools (such as JIRA / Confluence), combined with manual or algorithmic quantification. Risk weight w i Purpose: Used to calculate the standard deviation σ t It measures the volatility of risk distribution.

[0057] Among them, the risk threshold T k The essence of this is that the threshold is a dynamically adjusted critical value used to determine whether the risk weight is abnormal. After semantic correction, the final threshold is generated. The method of semantic correction will be described later.

[0058] The calculation formula is as follows:

[0059]

[0060] Risk threshold T k Its function is to trigger risk response strategy adjustments or alarms when the risk weight exceeds the threshold (such as raising the risk weight for high-risk and lowering the risk weight for low-risk in a tiered threshold system).

[0061] The preset risk weight distribution threshold is used to measure whether the distribution characteristics of the risk weight fluctuate abnormally or exceed the normal fluctuation range. The distribution characteristics are used to reflect the fluctuation of the risk weight of the project to be evaluated, and can be expressed by statistical indicators, such as standard deviation. Therefore, the distribution threshold is the standard deviation threshold.

[0062] In one feasible implementation, obtaining the risk threshold under the current operating state includes: obtaining a preset baseline risk threshold T. base and the status parameters under the current operating state; using the aforementioned status parameters and the benchmark risk threshold T base The risk threshold is obtained by using a preset dynamic threshold calibration algorithm. The dynamic threshold calibration algorithm can refer to equation (1) above, and incorporates the schedule deviation rate ΔP, resource volatility ΔR, learning rate coefficient α, resource sensitivity coefficient β, environmental complexity factor ρ, and the baseline risk threshold T. base The input is fed into the dynamic threshold calibration algorithm to obtain the risk threshold T. k .

[0063] Among them, the state parameters are used to reflect the parameters of the project's operating environment, including but not limited to schedule deviation rate ΔP, resource volatility rate ΔR, learning rate coefficient α, resource sensitivity coefficient β, and environmental complexity factor ρ.

[0064] For example, the schedule deviation rate, resource volatility, learning rate coefficient, resource sensitivity coefficient, and environmental complexity factor are explained as follows:

[0065] 1) Definition of schedule deviation rate ΔP: The ratio of the actual progress to the planned progress, calculated using the following formula:

[0066]

[0067] For example: if the plan is to complete 30% and the actual completion is 20%, and the total cycle is 100 days, then ΔP = (20-30) / 100 = -0.1 (a negative value indicates a delay in progress).

[0068] The schedule deviation rate ΔP is collected by extracting schedule data in real time from the Gantt chart or task status of project management tools (such as JIRA and Confluence).

[0069] Method for determining the schedule deviation rate ΔP: Automatically monitor milestone completion times and obtain the schedule deviation rate by integrating system data through API.

[0070] 2) Definition of resource volatility ΔR: The ratio of resource consumption deviating from the plan, calculated using the following formula:

[0071]

[0072] For example: if the budget is 1 million yuan and the actual consumption is 1.2 million yuan, then ΔR = (120 - 100) / 100 = 0.2.

[0073] Resource volatility ΔR is collected by obtaining resource usage data through financial systems (such as SAP) or resource management platforms (such as Microsoft Project).

[0074] Methods for determining resource volatility ΔR: Regularly compare budgets with actual expenditures, and correct outliers through manual review.

[0075] 3) Definition of learning rate coefficient α: It controls the sensitivity of the threshold to progress and resource fluctuations, and can be set in the range of [0.05, 0.2]. The larger α is, the more aggressive the threshold adjustment; the smaller α is, the more conservative.

[0076] Methods for determining the learning rate coefficient α:

[0077] Initial value: Preset according to project type (0.1 for IT projects, 0.15 for medical projects).

[0078] Dynamic optimization: trained using historical data or adjusted based on Monte Carlo simulations.

[0079] 4) Definition of resource sensitivity coefficient β: It amplifies or suppresses the impact of resource fluctuations on the threshold, and its range is [0.1, 0.3]. The larger β is, the more significant the impact of resource fluctuations on the threshold.

[0080] Methods for determining the resource sensitivity coefficient β:

[0081] Preset rules: Set resource-intensive projects (such as hardware development) to 0.3 and software projects to 0.2.

[0082] Ablation experiment: Compare the changes in false negative rate after removing resource-sensitive items.

[0083] 5) Definition of environmental complexity factor ρ: It reflects the complexity of a multi-stage project, and the calculation formula is as follows:

[0084]

[0085] For example: if the number of stages = 6, then ρ = 1 + log 10 (6 / 3) = 1.3.

[0086] The role of the environmental complexity factor ρ: the more stages there are, the larger ρ becomes, and the impact of resource fluctuations is diluted (ΔR divided by ρ).

[0087] The logarithmic base of the environmental complexity factor ρ: Based on actual calculation results, it conforms to logarithm. 10 (For example, when the number of stages is 8, ρ = 1.43), so it can be set according to the base of 10.

[0088] The definition of project phase number is: the total number of independent phases in the project lifecycle (such as requirements, design, development, testing, and delivery).

[0089] Project phase number collection method: Extract phase division information from project plan documents or WBS (Work Breakdown Structure).

[0090] The method for determining the number of project phases is as follows: the project manager defines them during the planning stage and stores them in a structured manner using tools such as Microsoft Project.

[0091] In one feasible implementation, when the risk weight exceeds a threshold, a risk response strategy adjustment or alarm is triggered. An alarm is triggered when the risk weight exceeds a certain number of consecutive occurrences; other situations may trigger risk response strategy adjustments, such as raising the threshold for high-risk risks and lowering it for low-risk risks, to ensure the threshold aligns with the current operating environment. This involves pre-setting tiered thresholds to handle different levels of risk differently, such as raising the threshold for high-risk risks and lowering it for low-risk risks within a tiered threshold system. Specifically, after obtaining the preset baseline risk threshold, steps A01 to A04 are also included:

[0092] A01. Obtain a preset stratification threshold, wherein the stratification threshold includes a first threshold and a second threshold, and the first threshold is greater than the second threshold;

[0093] It should be noted that a tiered threshold system is pre-constructed, and tiered thresholds are set. The tiered thresholds include a first threshold and a second threshold. The first threshold is greater than the second threshold. The first threshold can be 0.3 and the second threshold can be 0.1, representing the critical values ​​of two different risk levels, high risk and low risk, respectively.

[0094] A02. The risk weight is compared with the first threshold and the second threshold respectively to obtain the target comparison result;

[0095] A03. If the target comparison result is that the risk weight is greater than the first threshold, then the benchmark risk threshold is increased according to the preset first proportional coefficient to obtain the increased benchmark risk threshold.

[0096] A04. If the target comparison result is that the risk weight is less than the second threshold, then the benchmark risk threshold is reduced according to the preset second proportional coefficient to obtain the reduced benchmark risk threshold.

[0097] Furthermore, the risk weight is compared with the first threshold and the second threshold respectively. If the risk weight is greater than the first threshold, the benchmark risk threshold is increased by a preset first proportional coefficient (e.g., 0.2), resulting in an increased benchmark risk threshold. If the risk weight is less than the second threshold, the benchmark risk threshold is decreased by a preset second proportional coefficient (e.g., 0.15), resulting in a decreased benchmark risk threshold. Other cases: a dynamic threshold is maintained. That is, the risk threshold is dynamically adjusted by adjusting the benchmark risk threshold.

[0098] 102. Using the risk weights and a preset distribution feature extraction algorithm, feature extraction is performed to obtain the distribution features of the risk weights. The distribution features are used to reflect the fluctuation of the risk weights of the project to be evaluated.

[0099] Furthermore, in order to improve the sensitivity to changes in the operating environment, after obtaining the risk weights, it is also necessary to determine their distribution characteristics. Specifically, the risk weights and a preset distribution feature extraction algorithm are used to extract features to obtain the distribution characteristics of the risk weights. The distribution characteristics are used to reflect the fluctuation of the risk weights of the project to be evaluated. The risk weights are the risk quantification of the state of each link, which can reflect the fluctuation of the operating environment.

[0100] For example, if the preset distribution feature extraction algorithm includes, but is not limited to, statistical analysis algorithms such as the standard deviation algorithm, then step 102 includes B01 to B02:

[0101] B01. Using a preset time window and the first time point corresponding to the risk weight, a risk weight sequence is obtained. The risk weight sequence includes the risk weight and the historical risk weights of each historical second time point within the time window before the first time point.

[0102] B02. Using the risk weight sequence and the standard deviation algorithm, the standard deviation of the risk weight is obtained, and the distribution characteristics include the standard deviation.

[0103] Understandably, in order to achieve statistical analysis of risk weights and clarify weight changes, this application uses a sliding time window to perform statistical analysis on the risk weights collected at each time point to obtain their distribution characteristics. The length of the time window can be 5 time points, with the first time point being t, the second time points including t-1, t-2, t-3, and t-4, and the risk weight sequence including the risk weight at time point t, the risk weight at time point t-1, the risk weight at time point t-2, the risk weight at time point t-3, and the risk weight at time point t-4, etc. There are five risk weights. Each time a new risk weight is collected, the distribution characteristics of the current risk weight will be recalculated.

[0104] The distribution characteristic can be the standard deviation. Taking a sliding window of 5 time units as an example, the formula for calculating the standard deviation is as follows:

[0105]

[0106] In the formula, σ t : Standard deviation of the risk weight distribution at the current time t; ∑_(i=t-5)^t: Summation over 5 consecutive time points from t-5 to t; w i : Risk weight value at time point i; μ: Average weight value over 5 time points.

[0107] 103. Conduct a risk assessment based on the risk threshold and risk weight to obtain the target risk assessment result of the project to be assessed;

[0108] It should be noted that risk assessment is performed based on the aforementioned risk threshold and risk weight to obtain the target risk assessment result for the project to be assessed. If the risk weight is greater than or equal to the risk threshold, the risk assessment result is considered to be in an abnormal state; otherwise, it is considered not abnormal. For greater accuracy, this application combines the risk assessment result with whether the anomaly is continuous. If the anomaly occurs consecutively a certain number of times, it is considered a continuous anomaly and an alarm is triggered. The target risk assessment result includes the result that the operating status of the project to be assessed is in a continuous abnormal state.

[0109] Specifically, step 103 may include steps C01 to C05:

[0110] C01. If the risk weight is greater than or equal to the risk threshold, the preliminary risk assessment result is determined to be an abnormal state, and the number of consecutive abnormalities N is incremented by 1, with N initially set to 0.

[0111] C02. Determine whether N is greater than or equal to a preset number threshold;

[0112] C03. If N is greater than or equal to a preset number of times threshold, then the target risk assessment result is determined to be that the operating status of the project to be assessed is in a continuous abnormal state.

[0113] C04. If N is less than a preset number of times threshold, then the target risk assessment result is determined to be that the operating status of the project to be assessed is in a discontinuous abnormal state.

[0114] C05. If the risk weight is less than the risk threshold, the preliminary risk assessment result is determined to be in a normal state, and N = 0 is set.

[0115] It should be noted that this application maintains a counter N to record the number of consecutive anomalies. The counter N is adjusted based on the comparison result between each risk weight and risk threshold. If the comparison result shows that the risk weight is greater than or equal to the risk threshold, the counter N is incremented; otherwise, the counter is cleared. This means that any instance without an anomaly implies a discontinuity of anomalies, thus clearing N. This allows for real-time assessment of anomaly continuity, enabling more accurate and efficient control over the adjustment of risk thresholds based on semantic analysis. The consecutive occurrence threshold can be three times; that is, if an anomaly occurs more than three times consecutively, it is considered a continuous anomaly state.

[0116] 104. Based on the distribution characteristics and distribution threshold, perform fluctuation assessment to determine the target fluctuation assessment result of the distribution characteristics;

[0117] Furthermore, semantic analysis to adjust the risk threshold also requires assessing the fluctuation of the operating environment. If the fluctuation is not significant, semantic correction is not necessary at this time. Specifically, fluctuation assessment is performed based on the distribution characteristics and distribution threshold to determine the target fluctuation assessment result for the distribution characteristics. Wherein, when the distribution characteristic σ... t A distribution threshold of 0.15 indicates possible abnormal fluctuations. The target fluctuation assessment result is that the risk weight is in an abnormal fluctuation state, otherwise it is a normal fluctuation.

[0118] If the target risk assessment result indicates that the operating status of the project to be assessed is in a continuous abnormal state, then the environment is considered abnormal and the threshold needs to be adjusted to execute step 105; otherwise, execute step 106.

[0119] 105. If the target risk assessment result indicates that the operating status of the project to be assessed is in a continuous abnormal state, and the target fluctuation assessment result indicates that the risk weight is in an abnormal fluctuation state, then based on the target risk assessment result and the target fluctuation assessment result, target early warning information is output to a preset terminal, and semantic analysis processing is performed using the target early warning information and the preset semantic analysis model to obtain a threshold correction value.

[0120] It should be noted that if the target risk assessment result indicates that the project under assessment is in a continuously abnormal operating state, and the target fluctuation assessment result indicates that the risk weight is in an abnormal fluctuation state, then a semantic correction threshold is required. Specifically, based on the target risk assessment result and the target fluctuation assessment result, a target early warning message is output to a preset terminal, and semantic analysis processing is performed using the target early warning message and a preset semantic analysis model to obtain a threshold correction value. In other words, the abnormal result is output to the terminal so that the user is aware, and the threshold is corrected through semantic analysis. The early warning message includes, but is not limited to, prompts composed of analytical terms and indicators related to project anomalies. Thus, the threshold correction value can be obtained through semantic analysis of the information. Furthermore, the threshold adjustment process can be included as part of the project report, generating a project report including the threshold adjustment (details omitted here), so that the user is aware of the changes in the threshold and risk response measures throughout the process.

[0121] For example, the target warning information includes at least a risk description text of the current risk event and a semantic analysis model of the occurrence time series, including a BERT model and an LSTM model. The step of using the target warning information and the preset semantic analysis model to perform semantic analysis processing to obtain the threshold correction value includes steps D01 to D03:

[0122] D01. Use the BERT model to perform semantic analysis on the risk description text to obtain the text semantic features V_text;

[0123] D02. Use the LSTM model to perform time series analysis on the occurrence time series to obtain the time series feature V_time;

[0124] D03. Based on the text semantic features and temporal features, a weighted calculation is performed to obtain the threshold correction value V_final.

[0125] It should be noted that semantic analysis drives the correction: a correction factor V_final is generated using a multimodal fusion model (BERT+LSTM): V_final = 0.6V_text + 0.4V_time, with a correction range of ±0.08; this correction factor is a threshold correction value.

[0126] 1) V_text (text semantic features): Text feature vectors generated by semantic parsing of risk description texts (such as "critical path delay" and "CT annotation delay") using the BERT model.

[0127] V_text (text semantic features) is generated by BERT extracting contextual semantic information (such as entities and event severity) from the text and outputting a 768-dimensional vector.

[0128] The role of V_text (text semantic features): quantifies the urgency of risks implied in the text (e.g., "delay" corresponds to a high correction value).

[0129] 2) V_time (Time Series Features): A time feature vector generated by performing time series analysis on the occurrence time series of risk events (such as risk weights at 5 consecutive time points) using the LSTM model.

[0130] V_time (time series features) is generated by LSTM capturing the time dependence and periodicity of risk events (such as sudden risk clusters) and outputting a 128-dimensional vector.

[0131] The role of V_time (time series feature) is to reflect the dynamic trend of risk events (e.g., the risk weight has been rising continuously recently).

[0132] The specific process of semantic analysis is illustrated below:

[0133] Text parsing (generated by V_text):

[0134] (1) Input: Risk description text (such as problem description in JIRA / Confluence).

[0135] (2) Processing:

[0136] BERT word segmentation and encoding are used to extract sentence vectors marked with [CLS].

[0137] Key semantic features are further extracted using a CNN network;

[0138] (3) Output: V_text value (e.g., “critical path delay” corresponds to V_text = 0.05).

[0139] Time series analysis (V_time generation):

[0140] (1) Input: Risk event timestamp sequence (e.g., risk weights of 5 consecutive time points).

[0141] (2) Processing:

[0142] A time segment is captured using a sliding window (from time t-5 to time t).

[0143] Bidirectional LSTM encoding is used to generate the hidden state at the last time step.

[0144] (3) Output: V_time value (e.g., V_time = 0.02 corresponding to a continuous increase in risk weight).

[0145] Multimodal fusion:

[0146] (1) Weighted calculation: V_final = 0.6V_text + 0.4V_time.

[0147] (2) Limiting correction: Limit V_final to ±0.08.

[0148] 106. Use the threshold correction value and risk threshold to correct the threshold, obtain the corrected risk threshold, and return to the step of obtaining the risk weight, risk threshold and preset risk weight distribution threshold of the current operating status of the project to be evaluated.

[0149] After obtaining the threshold correction value, the threshold is corrected using the threshold correction value and the risk threshold to obtain the corrected risk threshold, and then the process returns to the step of obtaining the risk weight, risk threshold and preset risk weight distribution threshold of the current operating status of the project to be evaluated.

[0150] For example, the corrected formula can be: final threshold = T_k + V_final.

[0151] For example, if the dynamic threshold T_k = 0.223, V_text = 0.05, and V_time = 0.02, then V_final = 0.6 × 0.05 + 0.4 × 0.02 = 0.038, and the final threshold = 0.223 + 0.038 = 0.258.

[0152] The modified formula is only a simplified one and may vary depending on the actual situation and needs. It will not be exhaustive here. As long as the modified threshold is obtained by combining the above two parameters, it should be considered as not deviating from the concept of this case and included in the scope of protection of this application.

[0153] The role of semantic analysis: V_text enhances semantic sensitivity: key risk descriptions in the text (such as "delay") directly raise the threshold. V_time adapts to dynamic changes: recent risk trends (such as continuous exceedances) adjust the threshold magnitude.

[0154] In summary, the risk threshold can be dynamically adjusted according to changes in the operating environment. The dynamic threshold T_k is the base value before semantic correction, calculated from the progress, resources, and environmental parameters. V_text can be adjusted based on the semantic analysis results of the risk description text to reflect the urgency implied in the text. V_time can be adjusted based on the time series characteristics of the risk event to adapt to the risk trend (e.g., the adjustment range needs to be increased for sudden risks). The correction range is ±0.08 to prevent the semantic correction from deviating excessively from the baseline threshold and to ensure system stability.

[0155] Its advantages include: Multimodal fusion: Combining textual semantics (What) and temporal dynamics (When) to avoid single-dimensional bias. Closed-loop control: The corrected threshold is fed back to the risk strategy engine in real time, forming a closed loop of "analysis → correction → execution → re-analysis". Noise robustness: The ±0.08 correction range limit avoids drastic threshold fluctuations caused by individual risk event descriptions.

[0156] V_text and V_time quantify risk features from the two dimensions of text content and time series, respectively, and generate a correction factor V_final through weighted fusion. The core of semantic analysis is to achieve multimodal risk understanding through the BERT+LSTM model, and finally dynamically optimize the threshold to improve the accuracy and adaptability of risk detection.

[0157] The experiment verified that the mechanism can reduce the false negative rate from 38% to 12%, and reduce the adverse impact on computer response speed, thus improving the speed.

[0158] This invention provides a dynamic threshold adjustment method. Through the above-described method, the target risk assessment result and the target fluctuation assessment result of the risk weight distribution characteristics of the project to be evaluated can be obtained. When the target risk assessment result indicates that the operating state of the project to be evaluated is in a continuous abnormal state, and the target fluctuation assessment result indicates that the current risk weight is in an abnormal fluctuation state, a semantic correction is made to the current risk threshold. This achieves accurate and timely detection of dynamic changes in the operating environment of the project to be evaluated, and corrects the risk threshold so that it can adapt to the dynamic changes in the project environment, reducing risk underreporting and preventing slowdowns in computer system response.

[0159] Please see Figure 2 , Figure 2 This is a structural block diagram of a dynamic threshold adjustment processing device according to an embodiment of the present invention, as shown below. Figure 2 The apparatus shown includes:

[0160] Data acquisition unit 201: used to acquire the risk weights, risk thresholds, and preset risk weight distribution thresholds of the project under evaluation in its current operating state;

[0161] Feature extraction unit 202: used to extract features using the risk weights and a preset distribution feature extraction algorithm to obtain the distribution features of the risk weights, the distribution features being used to reflect the fluctuation of the risk weights of the project to be evaluated;

[0162] First assessment unit 203: used to perform risk assessment based on the risk threshold and risk weight, and obtain the target risk assessment result of the project to be assessed;

[0163] Second evaluation unit 204: used to perform fluctuation evaluation based on the distribution characteristics and distribution threshold, and determine the target fluctuation evaluation result of the distribution characteristics;

[0164] Semantic analysis unit 205: If the target risk assessment result is that the operating status of the project to be assessed is in a continuous abnormal state, and the target fluctuation assessment result is that the risk weight is in an abnormal fluctuation state, then based on the target risk assessment result and the target fluctuation assessment result, it outputs target early warning information to a preset terminal, and uses the target early warning information and the preset semantic analysis model to perform semantic analysis processing to obtain a threshold correction value.

[0165] Threshold correction unit 206: Used to perform threshold correction using the threshold correction value and risk threshold to obtain the corrected risk threshold, and return to execute the step of obtaining the risk weight, risk threshold and preset risk weight distribution threshold of the current operating status of the project to be evaluated.

[0166] It should be noted that, Figure 2 The functions of each module in the device shown are as follows: Figure 1 The steps in the method shown are similar, and will not be repeated here to avoid repetition. Please refer to the previous sections for details. Figure 1 The content of each step in the method shown.

[0167] This invention provides a dynamic threshold adjustment processing device. Through this device, the target risk assessment result and the target fluctuation assessment result of the risk weight distribution characteristics of the project to be evaluated can be obtained. When the target risk assessment result indicates that the operating state of the project to be evaluated is in a continuous abnormal state, and the target fluctuation assessment result indicates that the current risk weight is in an abnormal fluctuation state, semantic correction is made to the current risk threshold. This achieves accurate and timely detection of dynamic changes in the operating environment of the project to be evaluated, and corrects the risk threshold so that it can adapt to the dynamic changes in the project environment, reducing risk underreporting and preventing slowdowns in computer system response.

[0168] Figure 3 An internal structural diagram of a computer device in one embodiment is shown. This computer device can specifically be a terminal or a server. Figure 3 As shown, the computer device includes a processor, memory, and a network interface connected via a system bus. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and may also store a computer program, which, when executed by the processor, causes the processor to perform the aforementioned methods. The internal memory may also store a computer program, which, when executed by the processor, causes the processor to perform the aforementioned methods. Those skilled in the art will understand that… Figure 3The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0169] In one embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform actions such as... Figure 1 The steps are shown.

[0170] In one embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, causes the processor to perform the following actions: Figure 1 The steps are shown.

[0171] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0172] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0173] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for dynamic threshold adjustment, characterized in that, The method includes: Obtain the risk weights, risk thresholds, and preset risk weight distribution thresholds for the project under evaluation, based on its current operational status. The risk weights and a preset distribution feature extraction algorithm are used to extract features to obtain the distribution features of the risk weights. The distribution features are used to reflect the fluctuation of the risk weights of the project to be evaluated. Based on the risk threshold and risk weight, a risk assessment is performed to obtain the target risk assessment result of the project to be assessed. Based on the distribution characteristics and distribution threshold, a volatility assessment is performed to determine the target volatility assessment result for the distribution characteristics. If the target risk assessment result indicates that the operating status of the project to be assessed is in a continuous abnormal state, and the target fluctuation assessment result indicates that the risk weight is in an abnormal fluctuation state, then based on the target risk assessment result and the target fluctuation assessment result, target early warning information is output to a preset terminal, and semantic analysis processing is performed using the target early warning information and the preset semantic analysis model to obtain a threshold correction value. The threshold is corrected using the threshold correction value and the risk threshold to obtain the corrected risk threshold, and then the process returns to the step of obtaining the risk weight, risk threshold and preset risk weight distribution threshold of the current operating status of the project to be evaluated.

2. The method according to claim 1, characterized in that, Obtaining the risk thresholds for the current operating status includes: Obtain the preset baseline risk threshold and the status parameters under the current operating state; The risk threshold is obtained using the state parameters, the baseline risk threshold, and a preset dynamic threshold calibration algorithm.

3. The method according to claim 2, characterized in that, After obtaining the preset baseline risk threshold, the following steps are also included: Obtain a preset stratification threshold, which includes a first threshold and a second threshold, wherein the first threshold is greater than the second threshold; The target comparison result is obtained by comparing the risk weight with the first threshold and the second threshold respectively. If the target comparison result is that the risk weight is greater than the first threshold, then the benchmark risk threshold is increased according to a preset first proportional coefficient to obtain the increased benchmark risk threshold; If the target comparison result is that the risk weight is less than the second threshold, then the benchmark risk threshold is reduced according to the preset second proportional coefficient to obtain the reduced benchmark risk threshold.

4. The method according to claim 1, characterized in that, The target warning information includes at least a risk description text of the current risk event and its occurrence time series. The semantic analysis model includes a BERT model and an LSTM model. The step of using the target warning information and the preset semantic analysis model to perform semantic analysis processing to obtain the threshold correction value includes: The risk description text is semantically analyzed using the BERT model to obtain its semantic features. The occurrence time series is analyzed using the LSTM model to obtain time series characteristics; The threshold correction value is obtained by weighting the text semantic features and temporal features.

5. The method according to claim 1, characterized in that, The preset distribution feature extraction algorithm includes the standard deviation algorithm. Therefore, the step of using the risk weights and the preset distribution feature extraction algorithm to extract features and obtain the distribution features of the risk weights includes: Using a preset time window and the first time point corresponding to the risk weight, a risk weight sequence is obtained. The risk weight sequence includes the risk weight and the historical risk weights of each historical second time point within the time window before the first time point. The standard deviation of the risk weights is obtained using the risk weight sequence and the standard deviation algorithm, and the distribution characteristics include the standard deviation.

6. The method according to claim 2, characterized in that, The state parameters include schedule deviation rate, resource volatility rate, learning rate coefficient, resource sensitivity coefficient, and environmental complexity factor. The step of obtaining the risk threshold using the state parameters, a baseline risk threshold, and a preset dynamic threshold calibration algorithm includes: The schedule deviation rate, resource volatility, learning rate coefficient, resource sensitivity coefficient, environmental complexity factor, and baseline risk threshold are input into the dynamic threshold calibration algorithm to obtain the risk threshold.

7. The method according to claim 1, characterized in that, The step of conducting a risk assessment based on the risk threshold and risk weight to obtain the target risk assessment result for the project to be assessed includes: If the risk weight is greater than or equal to the risk threshold, the preliminary risk assessment result is determined to be an abnormal state, and the number of consecutive abnormalities N is incremented by 1, with N initially set to 0. Determine whether N is greater than or equal to a preset number of times threshold; If N is greater than or equal to a preset number of times threshold, then the target risk assessment result is determined to be that the operating status of the project to be assessed is in a continuous abnormal state; If N is less than a preset number of times threshold, then the target risk assessment result is determined to be that the operating status of the project to be assessed is in a discontinuous abnormal state. If the risk weight is less than the risk threshold, the preliminary risk assessment result is determined to be in a normal state, and N = 0 is set.

8. A processing device for dynamic threshold adjustment, characterized in that, The device includes: Data acquisition unit: used to acquire the risk weights, risk thresholds, and preset risk weight distribution thresholds of the project under evaluation in its current operating status; Feature extraction unit: used to extract features using the risk weights and a preset distribution feature extraction algorithm to obtain the distribution features of the risk weights, which are used to reflect the fluctuation of the risk weights of the project to be evaluated; First assessment unit: used to conduct risk assessment based on the risk threshold and risk weight, and obtain the target risk assessment result of the project to be assessed; The second evaluation unit is used to perform fluctuation evaluation based on the distribution characteristics and distribution threshold, and to determine the target fluctuation evaluation result of the distribution characteristics. Semantic analysis unit: If the target risk assessment result indicates that the operating status of the project to be assessed is in a continuous abnormal state, and the target fluctuation assessment result indicates that the risk weight is in an abnormal fluctuation state, then based on the target risk assessment result and the target fluctuation assessment result, it outputs target early warning information to a preset terminal, and uses the target early warning information and a preset semantic analysis model to perform semantic analysis processing to obtain a threshold correction value. Threshold correction unit: Used to correct the threshold using the threshold correction value and the risk threshold to obtain the corrected risk threshold, and return to execute the step of obtaining the risk weight, risk threshold and preset risk weight distribution threshold of the current operating status of the project to be evaluated.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it causes the processor to perform the steps of the method as described in any one of claims 1 to 7.

10. A computer device, comprising a memory and a processor, characterized in that, The memory stores a computer program that, when executed by the processor, causes the processor to perform the steps of the method as described in any one of claims 1 to 7.

Images