Outpatient service queue AI decision bus method and system based on dynamic fusing and self-healing

Abstract

The invention discloses an outpatient queue AI decision bus method and system based on dynamic fusing and self-healing. The outpatient queue AI decision bus method comprises the following steps: S1, data acquisition and preprocessing: acquiring state, sorting, priority, load and sign-in rule data from a patient table and an association table; according to the method, automatic degradation and service continuity in an abnormal scene are realized through health degree monitoring and a fuse state machine, self-adaptive coordination of a multi-source rule is achieved by relying on a unified health degree evaluation and weight scheduling mechanism, factors such as fairness punishment and priority crowds are quantified by means of an AI multi-factor scoring model so as to balance fairness and efficiency, and the safety of the system is improved. According to the method, strategy recovery compliance and controllability are ensured through shadow parallel evaluation and full-process log auditing, meanwhile, an existing system data table and an SQL are reused, an existing interface is compatible, and on the basis that system robustness is enhanced, decision scientificity is improved, service fairness is guaranteed and supervision compliance is met, hospital deployment cost is reduced.

Description

Technical Field

[0001] This invention relates to the field of smart outpatient technology, and in particular to an AI decision bus method and system for outpatient queues based on dynamic circuit breaking and self-healing. Background Technology

[0002] With the popularization of hospital information systems, the "queueing / triage / guide" capabilities in outpatient and medical technology scenarios have become the key to experience and efficiency. Existing systems generally adopt rule-driven queue management models, which combine check-in switches, display number selection, priority groups, manual insertion and locking, and clinic closure / load status for sorting and scheduling. Taking the existing system as an example, at the data level, the typical structure and rules are: multiple serial numbers and displays, coexistence of strategies, check-in constraints, multi-state mechanisms, allocation priority, and manual intervention. However, the existing solutions have the following shortcomings: (1) Vulnerability of rules and data sources: Problems encountered on site, such as abnormal check-in service, incorrect rule templates, clinic closure, data quality fluctuations or equipment failure, the system lacks automatic circuit breaking / detour and rapid recovery capabilities, which can easily lead to long periods of no response or abnormal sorting. (2) Lack of unified dynamic decision-making: There is a lack of unified health assessment and weight scheduling mechanism among the multi-source rules (sign-in, display number strategy, priority group, fairness penalty, clinic load balancing), making it difficult to adapt to abnormal scenarios; (3) Insufficient balance between fairness and experience: Factors such as penalties for missing appointments, manual queue jumping, and priority for follow-up visits are coupled in a complex way, and existing practices often fail to guarantee both fairness and efficiency at the same time; (4) Lack of shadow assessment and retrospective analysis: The strategy switching and ranking results lack shadow comparison assessment and audit, making it difficult to prove that the recovery strategy is effective and compliant and controllable; In summary, this invention proposes an AI decision bus method and system for outpatient queues based on dynamic circuit breaking and self-healing. Summary of the Invention

[0003] Based on the technical problems existing in the background technology, this invention proposes an AI decision bus method and system for outpatient queues based on dynamic circuit breaking and self-healing.

[0004] The outpatient queue AI decision bus method proposed in this invention, based on dynamic circuit breaking and self-healing, includes the following steps: S1: Data Acquisition and Preprocessing: Acquire status, sorting, priority, load, and check-in rule data from the patient table and related tables; S2: Health Monitoring: Construct health probes for check-in switches, display number strategies, load statistics, and automatic triage marking to generate health status H and fault indicators; S3: Circuit Breaker Detection: Maintain a circuit breaker state machine for each rule / data source and trigger a degradation strategy when an exception occurs; S4: Strategy set selection: Select the main strategy set based on the current health status. The main strategy set includes the display number, priority constraint, fairness penalty, and load balancing weight. S5: AI multi-factor scoring: Score and sort candidate patients, output queue order and display number, and retain shadow strategy for parallel scoring; S6: Shadow Evaluation: Statistically compare the differences between the shadow strategy and the main strategy on key indicators to provide evidence for "half-open probing / recovery closure"; S7: Half-open probing and self-healing: When the shadow strategy statistically analyzed in S6 is stable, a small proportion of the flow is switched to the alternative strategy, and the closed loop is restored after the stability is observed. S8: Audit Records and Visualization: Records the factor composition, weights, circuit breaker status, and switching reasons for each decision, and supports traceability and supervision; S9: Online learning and adaptive weights: Automatically adjusts the weights and thresholds of each factor based on historical evaluation data; S10: Connect to existing queue output: Connect to existing interfaces and pages in a compatible manner to ensure backward compatibility.

[0005] Preferably, in S2, the formula used to calculate the health rate H is: H = 1 - α·FailRatei - β·TimeoutRatei - γ·MissingRatei - δ·ConflictRatei; Where FailRatei is the error rate, TimeoutRatei is the timeout rate, MissingRatei is the missing rate of key fields, ConflictRatei is the rule conflict rate, and α, β, γ and δ are configurable weights.

[0006] Preferably, in step S3, the conditions for determining whether the circuit has been broken are as follows: (1) Opening condition: If H < θopen or an abnormality is triggered K times consecutively, the fuse state machine enters OPEN; (2) Half-open condition: If the shadow evaluation stability index Mshadow≥θshadow and H rises more than ΔH in the observation window, then enter HALF-OPEN; (3) Closure condition: When the trial is successful within the semi-open quantity ratio r and the number of abnormalities is ≤ ε, CLOSED is restored.

[0007] Preferably, in step S5, the formula used for multi-factor scoring is as follows: S(p)=w1·Ftime(p)+w2·Fpriority(p)+w3·Ffairness(p)+w4·Fload(p)+w5·Fsign(p)-w6·Flock(p); Where Ftime is the time / sequence factor, Fpriority is the priority factor, Ffairness is the fairness factor, Fload is the load balancing factor, Fsign is the sign-in constraint factor, which adds points to patients who have signed in when sign-in is enabled, and punishes or directly restricts and filters patients who have not signed in, and Flock is the locking factor, which freezes the position or punishes patients with is_lock_type=1.

[0008] Preferably, in S4, the condition for selecting the display number strategy is: when the display number rule is broken, the backup display number is used as the downgrade, and in the broken state, the unavailable main sequence number is automatically rolled back.

[0009] Preferably, in step S6, the formula used for the shadow evaluation stability index is as follows: Shadow policy score: Mshadow=η1·Norm(-AvgWaitTime)+η2·Norm(-OverRate)+η3·Norm(Throughput)+η4·Norm(Fairness)+η5·Norm(Balance), and only after reaching the threshold θshadow will it enter the half-open trial.

[0010] This invention also proposes an outpatient queue AI decision bus system based on dynamic circuit breaking and self-healing, including a data acquisition and feature construction module, a rule and data health monitoring module, a circuit breaking and self-healing management module, an AI decision engine module, a shadow evaluation module, a log auditing and visualization module, and a strategy management and configuration module. The data acquisition and feature construction module is used to interface with the out_sick_info, qcs_visit_info, and qcs_rule_template_info tables; The rules and data health monitoring module is used to establish health probes for check-in constraints, display number rules, load statistics, and clinic start / stop status. The fuse and self-healing management module includes a multi-fuse state machine. When the multi-fuse state machine enters OPEN under abnormal conditions, the shadow evaluation passes the half-open probe and successfully restores the closed state. The AI ​​decision engine module is used for multi-factor scoring and constraint solving, and selects different weights for different health states; The shadow evaluation module calculates the scoring results of the backup strategy in parallel, without affecting the online process, and serves as the basis for restoring closure. The log auditing and visualization module is used to record the details of each decision to assist in quality control and compliance. The strategy management and configuration module is used to visualize configuration weights, thresholds, trial ratios, and recovery criteria.

[0011] Preferably, the logic of the rules and data health monitoring module is to establish a health level H for the sign-in probe isEnableSignIn, the display number selection serial_number_type, the load data selectDoctorQueueCount, and the clinic start / stop getConsultationRoomQueueNumber. The circuit breaker and self-healing management module is used to maintain the state machine field and timestamp of each rule_code, and to perform recovery in conjunction with shadow evaluation; The input to the AI ​​decision engine module is a list of candidate patients. The process is: multi-factor scoring / learning sorting + constraints + display number degradation logic. The output is a sorted list and displayNumber. The shadow evaluation module calculates scores in parallel with the main strategy, without changing the online ranking, and records evaluation indicators. The logging and auditing module interfaces with the controller / service layer.

[0012] Compared with existing technologies, the beneficial effects of this invention are: 1. By using health monitoring and circuit breaker state machine, the system automatically triggers policy degradation when anomalies occur, avoiding prolonged periods of no response or abnormal sorting. It also covers scenarios such as abnormal check-in service, clinic closures, and data quality fluctuations, ensuring uninterrupted queue scheduling and enhancing system robustness and service continuity. 2. By establishing a unified health assessment system, the weights of multi-source rules (check-in, display number, load, etc.) are dynamically adjusted, and the optimal strategy is automatically adapted in abnormal scenarios, solving the problem of lack of coordination among multiple rules in the existing system; 3. By using an AI multi-factor scoring model, factors such as fairness penalties, priority groups, and handling of missed appointments are quantified to avoid the imbalance of rights caused by manual queue jumping and priority for follow-up visits, optimize average waiting time, missed appointment rate and throughput, improve the overall patient experience, and balance fairness and service efficiency. 4. Through parallel evaluation of shadow strategies, data support is provided for strategy recovery, ensuring that the recovery process is compliant and controllable. It can also fully record information such as decision factors, weights, and circuit breaker status, meeting the hospital's quality control and regulatory requirements, and has shadow evaluation and full-process traceability capabilities. 5. Reuse existing data tables and SQL logic, only add health monitoring and log audit tables, without large-scale modification, and be compatible with existing interfaces and pages, reducing hospital deployment and migration costs and achieving compatibility and ease of implementation.

[0013] This invention achieves automatic degradation and service continuity in abnormal scenarios through health monitoring and a circuit breaker state machine. It achieves adaptive collaboration of multi-source rules by relying on a unified health assessment and weight scheduling mechanism. It uses an AI multi-factor scoring model to quantify factors such as fairness penalties and priority groups to balance fairness and efficiency. Shadow parallel evaluation and full-process log auditing ensure that policy recovery is compliant and controllable. At the same time, it reuses existing system data tables and SQL, and is compatible with existing interfaces. While enhancing system robustness, improving the scientific nature of decision-making, ensuring service fairness, and meeting regulatory compliance, it reduces hospital deployment costs and effectively improves the experience and efficiency of outpatient queuing, triage, and guidance. Attached Figure Description

[0014] Figure 1 The flowchart shows the outpatient queue AI decision bus method based on dynamic circuit breaking and self-healing proposed in this invention. Figure 2 This is a block diagram of the outpatient queue AI decision bus system based on dynamic circuit breaking and self-healing proposed in this invention. Detailed Implementation

[0015] The present invention will be further explained below with reference to specific embodiments.

[0016] Example 1 Reference Figure 1 This embodiment proposes an AI decision bus method for outpatient queues based on dynamic circuit breaking and self-healing, including the following steps: S1: Data Acquisition and Preprocessing: Acquire status, sorting, priority, load, and check-in rule data from the patient table and related tables; S2: Health Monitoring: Construct health probes for check-in switches, display number strategies, load statistics, and automatic triage marking to generate health status H and fault indicators; The formula used to calculate the health rating H is: H = 1 - α·FailRatei - β·TimeoutRatei - γ·MissingRatei - δ·ConflictRatei; Where FailRatei is the error rate, TimeoutRatei is the timeout rate, MissingRatei is the key field missing rate, ConflictRatei is the rule conflict rate, and α, β, γ and δ are configurable weights; S3: Circuit Breaker Detection: Maintain a circuit breaker state machine for each rule / data source and trigger a degradation strategy when an exception occurs; The conditions for triggering the circuit breaker are as follows: (1) Opening condition: If H < θopen or an abnormality is triggered K times consecutively, the fuse state machine enters OPEN; (2) Half-open condition: If the shadow evaluation stability index Mshadow≥θshadow and H rises more than ΔH in the observation window, then enter HALF-OPEN; (3) Closure condition: When the trial is successful within the semi-open volume ratio r and the number of anomalies is ≤ ε, CLOSED is restored; S4: Strategy set selection: Select the main strategy set based on the current health status. The main strategy set includes the display number, priority constraint, fairness penalty, and load balancing weight. The conditions for selecting the display number strategy are as follows: when the display number rule is broken, the backup display number is used as the downgrade, and in the broken state, the unavailable main sequence number is automatically rolled back. S5: AI multi-factor scoring: Score and sort candidate patients, output queue order and display number, and retain shadow strategy for parallel scoring; The formula used when performing multi-factor scoring is as follows: S(p)=w1·Ftime(p)+w2·Fpriority(p)+w3·Ffairness(p)+w4·Fload(p)+w5·Fsign(p)-w6·Flock(p); Ftime is the time / sequence factor, Fpriority is the priority factor, Ffairness is the fairness factor, Fload is the load balancing factor, Fsign is the sign-in constraint factor, which adds points to patients who have signed in when sign-in is enabled, and punishes or directly restricts and filters patients who have not signed in. Flock is the locking factor, which freezes the position or punishes patients with is_lock_type=1. S6: Shadow Evaluation: Statistically compare the differences between the shadow strategy and the main strategy on key indicators to provide evidence for "half-open probing / recovery closure"; The formula used for the stability index of the shadow assessment is as follows: Shadow policy score: Mshadow = η1·Norm(-AvgWaitTime) + η2·Norm(-OverRate) + η3·Norm(Throughput) + η4·Norm(Fairness) + η5·Norm(Balance), and only after reaching the threshold θshadow will it enter the half-open trial; S7: Half-open probing and self-healing: When the shadow strategy statistically analyzed in S6 is stable, a small proportion of the flow is switched to the alternative strategy, and the closed loop is restored after the stability is observed. S8: Audit Records and Visualization: Records the factor composition, weights, circuit breaker status, and switching reasons for each decision, and supports traceability and supervision; S9: Online learning and adaptive weights: Automatically adjusts the weights and thresholds of each factor based on historical evaluation data; S10: Connect to existing queue output: Integrate with existing interfaces and pages in a compatible manner to ensure backward compatibility; This embodiment achieves automatic degradation and service continuity in abnormal scenarios through health monitoring and circuit breaker state machines. It achieves adaptive collaboration of multi-source rules by relying on a unified health assessment and weight scheduling mechanism. It uses an AI multi-factor scoring model to quantify factors such as fairness penalties and priority groups to balance fairness and efficiency. Shadow parallel assessment and full-process log auditing ensure that policy recovery is compliant and controllable. At the same time, it reuses existing system data tables and SQL, and is compatible with existing interfaces. While enhancing system robustness, improving the scientific nature of decision-making, ensuring service fairness, and meeting regulatory compliance, it reduces hospital deployment costs and effectively improves the experience and efficiency of outpatient queuing / triage / guidance.

[0017] Example 2 Reference Figure 2 This embodiment proposes an outpatient queue AI decision bus system based on dynamic circuit breaking and self-healing, including a data acquisition and feature construction module, a rule and data health monitoring module, a circuit breaking and self-healing management module, an AI decision engine module, a shadow evaluation module, a log auditing and visualization module, and a strategy management and configuration module. The data acquisition and feature construction module is used to interface with the out_sick_info, qcs_visit_info, and qcs_rule_template_info tables; The rules and data health monitoring module is used to establish health probes for sign-in constraints, display number rules, load statistics, and clinic start / stop status. The logic of the rules and data health monitoring module is to establish the health level H for the sign-in probe isEnableSignIn, the display number selection serial_number_type, the load data selectionDoctorQueueCount, and the clinic start / stop getConsultationRoomQueueNumber. The circuit breaker and self-healing management module includes a multi-fuse state machine. When the multi-fuse state machine enters OPEN under abnormal conditions, the shadow evaluation successfully restores the closed state machine through the half-open probe. The circuit breaker and self-healing management module is used to maintain the state machine field and timestamp of each rule_code and to perform recovery in conjunction with the shadow evaluation. The AI ​​decision engine module is used for multi-factor scoring and constraint solving. Different weights are selected for different health states. The input of the AI ​​decision engine module is a list of candidate patients. The process is: multi-factor scoring / learning sorting + constraints + display number degradation logic. The output is a sorted list and displayNumber. The shadow evaluation module calculates the scoring results of the backup strategy in parallel, which does not affect the online system and serves as the basis for restoring closure. The shadow evaluation module calculates the score in parallel with the main strategy, does not change the online ranking, and records the evaluation indicators. The log auditing and visualization module is used to record the details of each decision to assist in quality control and compliance; The strategy management and configuration module is used to visualize configuration weights, thresholds, trial ratios, and recovery criteria, while the log and audit module interfaces with the controller / service layer.

[0018] Typical implementation case: Case 1: Check-in service exception scenario: In the waiting area where check-in restrictions are in place 1. Environment: isEnableSignIn=1, rule.serial_number_type=1 (prioritize reality_sort); 2. Process: (1) Collect candidate patients (status_valuein(1,2), and reality_sort!=0); (2) Health monitoring detected jitter in the check-in interface response, H dropped to 0.62 < θopen = 0.7, and the fuse state machine entered OPEN; (3) Strategy downgrade: The display number strategy is switched to register_sort; the sign-in constraint bypass is used to give -λsign penalty to patients who have not signed in instead of direct filtering; (4) The AI ​​decision-making process scores and ranks all candidates, constrains patients with is_lock_type=1 to not move, and imposes a fairness penalty on patients with is_manual_insertion=1. (5) Parallel shadow evaluation continues to calculate the scoring indicators of the sign-in constraint strategy and count AvgWaitTime, OverRate, Balance, etc.; (6) After observing the window, if the shadow policy Mshadow≥0.8 and H rises back to 0.85, enter HALF-OPEN; attempt recovery with r=10% flow. If there are no abnormalities within ε=1, restore CLOSED; 3. Results: The system remained stable in the check-in queue despite service fluctuations, automatically returning to the main strategy after service recovery, and the entire process was audited.

[0019] Case 2: Uneven workload and clinic closures across multiple clinics: 1. Environment: is_stop=1 in one consultation room; the other two consultation rooms have high load. 2. Process: The AI ​​decision engine increases the Fload weight and dynamically guides patients to less busy examination rooms based on the statistical data of selectDoctorQueueCount and getConsultationRoomQueueNumber, ensuring throughput and fairness. 3. Results: Load variance decreased and the pass rate decreased.

[0020] Case 3: Manual queue jumping and penalties for exceeding the queue number: 1. Environment: Records with is_insert=1 and is_manual_insertion=1 exist; 2. Process: Add a Fairness penalty to the scoring function to limit the depth and frequency of queue jumping; increase the penalty for patients who miss their appointment multiple times, but combine this with repeat_number to control the experience of repeat patients; 3. Result: Balancing the rights and interests of priority groups with overall fairness.

[0021] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. An AI decision bus method for outpatient queues based on dynamic circuit breaking and self-healing, characterized in that, Includes the following steps: S1: Data Acquisition and Preprocessing: Acquire status, sorting, priority, load, and check-in rule data from the patient table and related tables; S2: Health Monitoring: Construct health probes for check-in switches, display number strategies, load statistics, and automatic triage marking to generate health status H and fault indicators; S3: Circuit Breaker Detection: Maintain a circuit breaker state machine for each rule / data source and trigger a degradation strategy when an exception occurs; S4: Strategy set selection: Select the main strategy set based on the current health status. The main strategy set includes the display number, priority constraint, fairness penalty, and load balancing weight. S5: AI multi-factor scoring: Score and sort candidate patients, output queue order and display number, and retain shadow strategy for parallel scoring; S6: Shadow Evaluation: Statistically compare the differences between the shadow strategy and the main strategy on key indicators to provide evidence for "half-open probing / recovery closure"; S7: Half-open probing and self-healing: When the shadow strategy statistically analyzed in S6 is stable, a small proportion of the flow is switched to the alternative strategy, and the closed loop is restored after the stability is observed. S8: Audit Records and Visualization: Records the factor composition, weights, circuit breaker status, and switching reasons for each decision, and supports traceability and supervision; S9: Online learning and adaptive weights: Automatically adjusts the weights and thresholds of each factor based on historical evaluation data; S10: Connect to existing queue output: Connect to existing interfaces and pages in a compatible manner to ensure backward compatibility.

2. The outpatient queue AI decision bus method based on dynamic circuit breaking and self-healing as described in claim 1, characterized in that, In S2, the formula used to calculate the health H is: H = 1 - α·FailRatei - β·TimeoutRatei - γ·MissingRatei - δ·ConflictRatei; Where FailRatei is the error rate, TimeoutRatei is the timeout rate, MissingRatei is the missing rate of key fields, ConflictRatei is the rule conflict rate, and α, β, γ and δ are configurable weights.

3. The outpatient queue AI decision bus method based on dynamic circuit breaking and self-healing as described in claim 1, characterized in that, In S3, the conditions for determining whether the circuit breaker has tripped are as follows: (1) Opening condition: If H < θopen or an abnormality is triggered K times consecutively, the fuse state machine enters OPEN; (2) Half-open condition: If the shadow evaluation stability index Mshadow≥θshadow and H rises more than ΔH in the observation window, then enter HALF-OPEN; (3) Closure condition: When the trial is successful within the semi-open quantity ratio r and the number of abnormalities is ≤ ε, CLOSED is restored.

4. The outpatient queue AI decision bus method based on dynamic circuit breaking and self-healing as described in claim 1, characterized in that, In S5, the formula used for multi-factor scoring is as follows: S(p)=w1·Ftime(p)+w2·Fpriority(p)+w3·Ffairness(p)+w4·Fload(p)+w5·Fsign(p)-w6·Flock(p); Where Ftime is the time / sequence factor, Fpriority is the priority factor, Ffairness is the fairness factor, Fload is the load balancing factor, Fsign is the sign-in constraint factor, which adds points to patients who have signed in when sign-in is enabled, and punishes or directly restricts and filters patients who have not signed in, and Flock is the locking factor, which freezes the position or punishes patients with is_lock_type=1.

5. The outpatient queue AI decision bus method based on dynamic circuit breaking and self-healing according to claim 1, characterized in that, In S4, the conditions for selecting the display number strategy are as follows: when the display number rule is circuit-broken, the backup display number is used as the downgrade, and in the circuit-broken state, the unavailable main sequence number is automatically rolled back.

6. The outpatient queue AI decision bus method based on dynamic circuit breaking and self-healing according to claim 3, characterized in that, In S6, the formula used for the stability index of the shadow assessment is as follows: Shadow policy score: Mshadow=η1·Norm(-AvgWaitTime)+η2·Norm(-OverRate)+η3·Norm(Throughput)+η4·Norm(Fairness)+η5·Norm(Balance), and only after reaching the threshold θshadow will it enter the half-open trial.

7. An outpatient queue AI decision bus system based on dynamic circuit breaking and self-healing, used to implement the method described in any one of claims 1-6, characterized in that, It includes modules for data collection and feature construction, rules and data health monitoring, circuit breaking and self-healing management, AI decision engine, shadow evaluation, log auditing and visualization, and policy management and configuration. The data acquisition and feature construction module is used to interface with the out_sick_info, qcs_visit_info, and qcs_rule_template_info tables; The rules and data health monitoring module is used to establish health probes for check-in constraints, display number rules, load statistics, and clinic start / stop status. The fuse and self-healing management module includes a multi-fuse state machine. When the multi-fuse state machine enters OPEN under abnormal conditions, the shadow evaluation passes the half-open probe and successfully restores the closed state. The AI ​​decision engine module is used for multi-factor scoring and constraint solving, and selects different weights for different health states; The shadow evaluation module calculates the scoring results of the backup strategy in parallel, without affecting the online process, and serves as the basis for restoring closure. The log auditing and visualization module is used to record the details of each decision to assist in quality control and compliance. The strategy management and configuration module is used to visualize configuration weights, thresholds, trial ratios, and recovery criteria.

8. The outpatient queue AI decision bus system based on dynamic circuit breaking and self-healing according to claim 7, characterized in that, The logic of the rules and data health monitoring module is to establish a health level H for the sign-in probe isEnableSignIn, the display number selection serial_number_type, the load data selectionDoctorQueueCount, and the clinic start / stop getConsultationRoomQueueNumber. The circuit breaker and self-healing management module is used to maintain the state machine field and timestamp of each rule_code, and to perform recovery in conjunction with shadow evaluation; The input to the AI ​​decision engine module is a list of candidate patients. The process is: multi-factor scoring / learning sorting + constraints + display number degradation logic. The output is a sorted list and displayNumber. The shadow evaluation module calculates scores in parallel with the main strategy, without changing the online ranking, and records evaluation indicators. The logging and auditing module interfaces with the controller / service layer.

Images