Online learning platform operation supervision system based on applet

Abstract

The invention relates to the technical field of educational informatization, in particular to an online learning platform operation supervision system based on an applet, which comprises a resource allocation module, a performance evaluation module, an anomaly prediction module, an intervention scheduling module and a load balancing module. According to the method, the course access frequency, the user behavior mode and the server load change rate are dynamically monitored, the platform operation state is mastered in time, the matching analysis of the access fluency and the response time is realized, the performance fluctuation interval is accurately screened, and the abnormal behavior is identified in combination with the time correlation degree of the behavior track and the response time. An intervention time period is dynamically selected based on communication delay and instruction intervals, a processing path and execution time are optimized in combination with task priorities and available load ratios, the exception handling response efficiency and the platform operation stability are effectively improved, execution bottlenecks caused by lagging and unbalanced resource allocation during manual monitoring and handling of sudden problems are relieved, and the system performance is improved. And the data coping capability in a complex environment is enhanced.

Description

Technical Field

[0001] The present invention relates to the field of educational informatization technology, and in particular to an online learning platform operation supervision system based on mini-programs. Background Art

[0002] The field of educational informatization technology refers to the use of information technology to digitize, network, and intelligently process educational resources, promoting reform and innovation in educational models, teaching methods, and management approaches. This technology involves core content such as educational resource sharing, online learning platforms, intelligent education systems, and data analysis and processing. With the development of information technology, educational informatization has gradually covered multiple aspects, from educational content and teaching management to learner assessment. Information technology has become a key driving force for educational modernization. Traditional online learning platform operation supervision systems refer to technical means for real-time monitoring and management of online learning platform operations. Traditional online learning platform operation supervision relies on manual methods or rudimentary data monitoring tools, which manage and adjust the learning platform's various operating indicators, such as user login status, course access status, and interaction frequency. This is not only inefficient but also difficult to meet the various needs of the complex and ever-changing online learning environment.

[0003] Existing technologies rely on manual inspections and primary data tools to monitor and manage platform operating indicators, making it difficult to finely identify and respond to information such as course access, user behavior, and server load. In scenarios with intensive user access, complex and changeable behavior, or drastic load fluctuations, there are problems with identification delays and scheduling lags, resulting in the inability to capture and handle platform operating status anomalies in a timely manner. For example, when course access suddenly surges, performance bottlenecks are not discovered in time, causing the learning process to freeze and user experience to decline. The lack of automatic correlation analysis means also makes it difficult to form an effective feedback path for potential anomalies, thereby restricting the overall stable operation of the platform and resource scheduling efficiency. Summary of the Invention

[0004] The purpose of the present invention is to solve the shortcomings of the prior art and to propose an online learning platform operation supervision system based on mini-programs.

[0005] In order to achieve the above objectives, the present invention adopts the following technical solutions: A mini-program-based online learning platform operation supervision system includes: The resource allocation module monitors the course access frequency, user behavior patterns, and server load change rate based on data from the course access component, user behavior monitoring component, and server load detection component in the online learning platform, and generates a platform operation status set. The performance evaluation module collects course access fluency and user behavior response time data based on the platform operation status set, identifies the matching between the fluency interval and the response time interval, filters the data segments that deviate from the stable range, and generates the performance fluctuation monitoring interval; The anomaly prediction module calls the performance fluctuation monitoring interval, extracts the behavior trajectory of the corresponding time period in the user behavior monitoring component, analyzes the temporal correlation between the change amplitude of the behavior trajectory and the response time fluctuation amplitude, filters the abnormal behavior fragments, and obtains the platform abnormal behavior identification set; The intervention scheduling module detects the communication delay index and instruction execution interval index between the mini program and the server based on the platform abnormal behavior identification set, selects the low-latency time period as the intervention window, and synchronously maps the time period corresponding to the abnormal behavior with the intervention window to generate an intervention scheduling mapping table.

[0006] As a further solution of the present invention, the platform operation status set includes course access level, user activity level, and service load abnormality indicators; the performance fluctuation monitoring interval includes fluency fluctuation range, response time variation range, and data deviation segment; the platform abnormal behavior identification set includes abnormal behavior trajectory, abnormal response mode, and abnormal behavior type; the intervention scheduling mapping table includes abnormal time node, intervention window, and synchronization association information.

[0007] As a further solution of the present invention, the resource allocation module includes: The rate monitoring submodule is based on the data of the course access component, user behavior monitoring component, and server load detection component in the online learning platform. It uses the course access component to detect access frequency data and the user behavior monitoring component to detect behavior change data. It calculates the access frequency, behavior change rate, and load change rate per unit time, and compares them with the rate baseline value to generate the component rate offset. The state identification submodule calls the component rate offset, classifies the data according to the set course access, user activity and service load state rate intervals based on the access frequency, behavior change rate and load change rate, and marks the corresponding state of the data to obtain the real-time operation state set of the platform; The control instruction generation submodule calls the real-time operation status set of the platform, and screens the platform execution resource allocation, task adjustment and record control actions under the corresponding status according to the course access, user activity and service load status types, identifies the corresponding control instructions based on the action threshold associated with the differentiated status, and generates the platform operation status set.

[0008] As a further solution of the present invention, the performance evaluation module includes: The fluency identification submodule analyzes the change rate and fluctuation range of course access fluency based on the platform operation status set, selects data segments with low change rate and fluctuation range less than the fluency stability threshold, and obtains the fluency stability interval; The response time analysis submodule calls the fluency stability interval, identifies the response time change rate through the synchronously collected user behavior response time data, calculates the difference in response time changes, filters the data segments below the response time consistency threshold, and generates the response time consistency interval; The performance fluctuation screening submodule calls the response time consistency interval, and screens data segments that deviate from the stability threshold and the consistency threshold according to the fluency fluctuation amplitude and the response time consistency, to generate a performance fluctuation monitoring interval.

[0009] As a further solution of the present invention, the abnormality prediction module includes: The trajectory synchronization extraction submodule calls the performance fluctuation monitoring interval, extracts the user behavior monitoring component data, extracts the behavior trajectory points and operation change values ​​according to the timestamp, and organizes them into a synchronous data structure to generate a synchronous behavior trajectory set; The timing-related offset calculation submodule calls the synchronized behavior trajectory set, identifies the normalized change rate of the behavior trajectory change amplitude and the response time fluctuation amplitude of the response time change sequence, superimposes the offset term, compensates for the impact of the operation change, and filters the time segments above the abnormal offset threshold to generate the platform-related offset set; The abnormal behavior collection submodule calls the platform-related offset set, combines the trajectory change frequency and the operation mutation amplitude, filters the behavior segments that exceed the abnormal threshold, and obtains the platform abnormal behavior identification set.

[0010] As a further solution of the present invention, the intervention scheduling module includes: The communication delay detection submodule detects the communication delay index and instruction execution interval index between the mini-program end and the server end based on the platform abnormal behavior identification set, records the communication transmission delay and instruction execution response time respectively, and generates a communication delay data set; The low-latency screening submodule calls the communication delay data set, extracts communication delay and instruction execution interval indicators, screens time periods that meet low-latency requirements, identifies signal sending and receiving delays and instruction execution sending and receiving intervals within each period, counts the total number of detections, calculates the intervention time period score, screens low-scoring time intervals based on the score value, and generates a screening time interval list; The time period mapping submodule matches the time period corresponding to the abnormal behavior based on the screening time interval list, analyzes the time synchronization mapping relationship, and generates an intervention scheduling mapping table.

[0011] As a further solution of the present invention, the system further includes: The load balancing module calls the intervention scheduling mapping table, collects the current server load occupancy and abnormal behavior processing requirements, analyzes the ratio between processing requirements and available load capacity, adjusts task priority sorting rules, selects the optimal processing channel and execution time period, and generates a platform load balancing control table; The platform load balancing control table includes task priority setting, processing channel selection, and execution time arrangement.

[0012] As a further solution of the present invention, the load balancing module includes: The load demand collection submodule calls the intervention scheduling mapping table to collect the platform's real-time task types, processing requirements, and server load occupancy, detects the load proportion of the task type, and generates a platform load demand offset rate based on the load benchmark rate; The dynamic priority adjustment submodule detects tasks whose demand deviation rate exceeds a threshold based on the platform load demand deviation rate, processing urgency, and load demand intensity, and reorders tasks based on urgency and deviation amplitude to determine load distribution priorities and obtain a multi-functional platform task priority sequence; The task control allocation submodule selects channels whose load capacity and time period resources meet the requirements according to the multi-functional platform task priority sequence and the channel parameters and time period parameters in the intervention scheduling mapping table, allocates tasks to the optimal processing channel and execution time period, and generates a platform load balancing control table.

[0013] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, by dynamically monitoring the frequency of course access, user behavior patterns and server load change rate, it is possible to timely grasp the platform operation status, realize the matching analysis of access fluency and behavior response time, accurately screen the performance fluctuation range, and further combine the time correlation between behavior trajectory and response time to complete abnormal behavior identification. At the same time, based on the communication delay and instruction execution interval, the intervention time period is dynamically selected, and the processing path and execution time are optimized and scheduled with the help of task priority and available load matching strategy, which effectively improves the response efficiency of exception processing and the stability of platform operation, alleviates the execution bottleneck caused by the lag of manual monitoring and the imbalance of resource allocation when dealing with sudden performance problems, and improves the system's ability to cope with multi-dimensional data fluctuations in complex online learning environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 is a system flow chart of the present invention; Figure 2 This is a flow chart of the resource allocation module in the present invention; Figure 3 This is a flow chart of the performance evaluation module in the present invention; Figure 4 This is a flow chart of the abnormality prediction module in the present invention; Figure 5 This is a flow chart of the intervention scheduling module in the present invention; Figure 6 This is a flow chart of the load balancing module in the present invention. DETAILED DESCRIPTION

[0015] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0016] In the description of the present invention, it should be understood that the terms "length," "width," "up," "down," "front," "back," "left," "right," "vertical," "horizontal," "top," "bottom," "inside," "outside," and the like, indicating positions or relationships, are based on the positions or relationships shown in the accompanying drawings and are intended only to facilitate the description of the present invention and simplify the description. They do not indicate or imply that the devices or elements referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limiting the present invention. Furthermore, in the description of the present invention, "plurality" means two or more, unless otherwise expressly and specifically defined.

[0017] See also Figure 1 ,A mini-program-based online learning platform operation supervision system includes: The resource allocation module monitors the course access frequency, user behavior patterns, and server load change rate based on data from the course access component, user behavior monitoring component, and server load detection component in the online learning platform, and generates a platform operation status set. The performance evaluation module collects data on course access fluency and user behavior response time based on the platform operation status set, identifies the matching between fluency intervals and response time intervals, filters data segments that deviate from the stable range, and generates performance fluctuation monitoring intervals; The anomaly prediction module calls the performance fluctuation monitoring interval, extracts the behavior trajectory of the corresponding time period in the user behavior monitoring component, analyzes the temporal correlation between the change amplitude of the behavior trajectory and the fluctuation amplitude of the response time, filters the abnormal behavior fragments, and obtains the platform abnormal behavior identification set; The intervention scheduling module detects the communication delay and instruction execution interval between the mini-program and the server based on the platform's abnormal behavior identification set. It selects low-latency time periods as intervention windows, synchronously maps the time periods corresponding to abnormal behaviors with the intervention windows, and generates an intervention scheduling mapping table. The load balancing module calls the intervention scheduling mapping table to collect the current server load occupancy and abnormal behavior processing requirements, analyze the ratio between processing requirements and available load capacity, adjust the task priority sorting rules, screen the optimal processing channel and execution period, and generate the platform load balancing control table.

[0018] The platform operation status set includes course access level, user activity level, and service load abnormality indicators. The performance fluctuation monitoring interval includes fluency fluctuation range, response time variation range, and data deviation segment. The platform abnormal behavior identification set includes abnormal behavior trajectory, abnormal response mode, and abnormal behavior type. The intervention scheduling mapping table includes abnormal time node, intervention window, and synchronization correlation information. The platform load balancing control table includes task priority setting, processing channel selection, and execution time scheduling.

[0019] See also Figure 2 , the resource allocation module includes: The rate monitoring submodule is based on the data of the course access component, user behavior monitoring component, and server load detection component in the online learning platform. It uses the course access component to detect access frequency data and the user behavior monitoring component to detect behavior change data. It calculates the access frequency, behavior change rate, and load change rate per unit time, and compares them with the rate baseline value to generate the component rate offset. Based on the data of the course access component, user behavior monitoring component and server load detection component in the online learning platform, the course access component detects the access frequency data, for example, the online learning platform has 5,000 access requests to the "Advanced Mathematics" course within one minute, the user behavior monitoring component detects the user's behavior change data on the platform, for example, the number of times a user jumps from course A to course B within five minutes is 100 times, and the server load detection component detects the server CPU usage, memory occupancy and other load data, for example, the server CPU usage is 80%, and then calculates the access frequency per unit time, for example, the course access frequency per second on the online learning platform is 83.33 times (5,000 times / 60 seconds), the behavior change rate, for example, the user behavior change rate per minute is 20 times / minute (100 times / 5 minutes), and the load change rate, for example, the server The server CPU load change rate is 0.5% / second (the current CPU usage is 80% minus 79.5% in the previous second) and is compared with the pre-set rate baseline value. The rate baseline value is set based on the historical operation data of the online learning platform and system design requirements, through statistical analysis of the average access frequency, user behavior changes, and server load in different time periods (such as peak and off-peak periods), and combined with expert experience. For example, the benchmark value of course access frequency is set at 100 times / second, the benchmark value of user behavior change rate is set at 5 times / minute, and the benchmark value of server load change rate is set at 0.1% / second. Experimental verification shows that when the course access frequency is higher than 100 times / second, the user behavior change rate is higher than 5 times / minute, or the server load change rate is higher than 0.1% / second, the platform performance will decline significantly. The experimental data are shown in Table 1.

[0020] Table 1: Experimental data table for setting rate reference values As shown in Table 1, when the course access frequency reaches 100 times / second, the platform performance is normal. If it exceeds this value, lag will occur. Therefore, it is reasonable to set the rate baseline value of the course access frequency to 100 times / second. Similarly, the baseline value of the user behavior change rate is set to 5 times / minute, and the baseline value of the server load change rate is set to 0.1% / second. The calculated current rate is subtracted from the baseline value. For example, the component rate offset is -16.67 times / second by subtracting the baseline value of 100 times / second from the current course access frequency of 83.33 times / second, the user behavior change rate is 20 times / minute minus the baseline value of 5 times / minute to get 15 times / minute, and the load change rate is 0.5% / second minus the baseline value of 0.1% / second to get 0.4% / second, which generates the component rate offset.

[0021] The state identification submodule calls the component rate offset and classifies the data into the set course access, user activity and service load state rate intervals based on the access frequency, behavior change rate and load change rate, and marks the corresponding state of the data to obtain the real-time operation state set of the platform; Call component rate offset, such as calling course access frequency offset -16.67 times / second, user behavior change rate offset 15 times / minute, server load change rate offset 0.4% / second, according to access frequency, behavior change rate and load change rate, according to the set course access, user activity and service load status rate range classification, rate range is based on the statistical analysis of the historical operation data of the online learning platform, combined with the platform's tolerance for different load conditions and user experience requirements. The rationality is verified by a large number of simulation experiments. The experimental data show that when the course access rate is [0, 80] times / second, the user activity rate is [0, 3] times / minute, and the service load rate is [0, 0.05]% / second, the platform is in a "normal" state; when the course access rate is [80, 120] times / second, the user activity rate is [0, 120] times / second, and the service load rate is [0, 0.05]% / second, the platform is in a "normal" state; when the course access rate is [80, 120] times / second, the user activity rate is [0, 120] times / second, and the service load rate is [0, 0.05]% / second, the platform is in a "normal" state. When the access rate is [3, 8] times / minute and the service load rate is [0.05, 0.15]% / second, the platform is in a "busy" state; when the course access rate is higher than 120 times / second, the user activity rate is higher than 8 times / minute, and the service load rate is higher than 0.15% / second, the platform is in an "overloaded" state. For example, the current course access frequency is 83.33 times / second, which belongs to the rate interval of [80, 120] times / second in the "busy" state. The user behavior change rate is 20 times / minute, which belongs to the rate interval of the "overloaded" state higher than 8 times / minute. The server load change rate is 0.4% / second, which belongs to the rate interval of the "overloaded" state higher than 0.15% / second. The corresponding data status is marked, for example, the course access status is marked as "busy", the user activity status is marked as "overloaded", and the service load status is marked as "overloaded", and the real-time operation status set of the platform is obtained.

[0022] The control instruction generation submodule calls the platform's real-time operating status set. Based on the course access, user activity, and service load status types, it selects the platform's resource allocation, task adjustment, and record control actions under the corresponding status. It identifies the corresponding control instructions based on the action thresholds associated with the differentiated status and generates the platform's operating status set. Call the real-time operation status set of the platform, for example, call the course access status "busy", user activity status "overload", and service load status "overload" in the real-time operation status set of the platform. According to the course access, user activity and service load status types, for example, the current status type is "busy", "overload", and "overload", filter the platform to perform resource allocation, task adjustment, and record control actions in the corresponding state. For example, when the course access status is "busy", filter out resource allocation actions such as increasing server bandwidth and optimizing database queries. When the user activity status is "overloaded", filter out task adjustment actions such as limiting the number of concurrent users and enabling content distribution. When the service load status is "overloaded", filter out record control actions such as starting backup servers and reducing the priority of non-core services. Identify corresponding control instructions based on the action threshold associated with differentiated states. The action threshold is set based on the platform's historical operation data and system fault tolerance, by analyzing the system response and resource consumption in different states, and combining operation and maintenance experience, and is verified through simulation tests. For example, for the "increase server bandwidth" action, its trigger threshold is set to when the course When the access frequency offset exceeds 20 times / second, that is, the actual access frequency is higher than 120 times / second, it is triggered. For the "limit the number of concurrent users" action, its trigger threshold is set to when the user behavior change rate offset exceeds 10 times / minute, that is, when the actual behavior change rate is higher than 15 times / minute, it is triggered. For the "start backup server" action, its trigger threshold is set to when the server load change rate offset exceeds 0.2% / second, that is, when the actual load change rate is higher than 0.3% / second, it is triggered. Experimental data shows that when the threshold is reached, if no corresponding action is taken, The platform performance will drop sharply. Based on the current course access frequency offset of -16.67 times / second, it has not reached the threshold for triggering "increasing server bandwidth". Based on the current user behavior change rate offset of 15 times / minute, it has exceeded the threshold for triggering "limiting the number of concurrent users" of 10 times / minute. Based on the current server load change rate offset of 0.4% / second, it has exceeded the threshold for triggering "starting the backup server" of 0.2% / second. Identify the corresponding control instructions, such as identifying the "limiting the number of concurrent users" and "starting the backup server" instructions, and generate a platform operation status set.

[0023] See also Figure 3 , the performance evaluation module includes: The fluency identification submodule analyzes the change rate and fluctuation range of course access fluency based on the platform operation status set, selects data segments with low change rate and fluctuation range less than the fluency stability threshold, and obtains the fluency stability interval; Based on the platform operation status set, such as the data on course access status, user activity status and service load status in the platform operation status set, the course access fluency change rate and fluency fluctuation range are analyzed. For example, the fluency change rate is calculated by monitoring the number of video freezes every 10 seconds when the user is watching the course video. If the number of freezes changes from 0 times to 2 times within 10 seconds, the fluency change rate is 0.2 times / second. The fluency fluctuation range is calculated by recording the range of the number of video freezes within one minute. For example, the maximum number of freezes within one minute is 5 times, and the minimum is 1 time, then the fluctuation range is 4 times. The video with low change rate and fluctuation range less than the fluency stability threshold is screened. The fluency stability threshold is set based on users' general expectations of the fluency of the online learning platform and the system performance goals, and is calibrated through user questionnaires and actual test data. For example, the fluency stability threshold is set to a change rate lower than 0.1 times / second and a fluctuation amplitude less than 2 times. When the jamming change rate is lower than 0.1 times / second, users will hardly notice the jamming. When the jamming fluctuation amplitude is less than 2 times, the user experience is good. Experimental data shows that when this threshold is exceeded, user satisfaction drops significantly. Therefore, the data segment with a change rate lower than 0.1 times / second and a fluctuation amplitude less than 2 times is identified as the fluency stability interval, and the fluency stability interval is obtained.

[0024] The response time analysis submodule calls the fluency stability interval and identifies the response time change rate through the synchronously collected user behavior response time data. The formula is: ; Obtain the differences in response time changes, filter out data segments below the response time consistency threshold, and generate response time consistency intervals; in, represents the response time change rate, Representative The response time at a time point, represents the average response time at a given point in time, represents the total number of time points; Call the fluency stable interval, for example, call the data within the fluency stable interval, through the synchronous collection of user behavior response time data, such as collecting the response time data from the user clicking the course play button to the start of the video playback, identify the response time change rate, for example, the response time collected at 5 consecutive time points is = 0.5 seconds, = 0.6 seconds, = 0.55 seconds, = 0.7 seconds, = 0.65 seconds, where Represents the response time change rate, which is used to measure the fluctuation of the response time series, that is, the dispersion of the response time relative to its average value. Represents the total number of time points, that is, the total number of response time data involved in the calculation. represents the response time at the jth time point, that is, the system's feedback time to the user's operation at a specific moment, represent The average response time at each time point is the arithmetic mean of all response times in the selected time period. The calculation logic of the formula is to first calculate the average response time at each time point. Average response time at all time points The absolute difference between the two values ​​represents the degree to which each response time data point deviates from the average level. Then all the absolute differences are added together to get the total deviation, and finally the total deviation is divided by the total number of time points. , thus obtaining the mean absolute deviation, which quantifies the overall fluctuation of the response time and reflects the difference in its changes. The formula is: To find the difference in response time changes, first calculate the average response time , =(0.5+0.6+0.55+0.7+0.65) / 5=3.0 / 5=0.6 seconds, then calculate the absolute difference between the response time at each time point and the average response time and sum them; |0.5-0.6|=0.1, |0.6-0.6|=0.0, |0.55-0.6|=0.05, |0.7-0.6|=0.1, |0.65-0.6|=0.05; =0.1+0.0+0.05+0.1+0.05=0.3, and finally calculate the response time change rate , Q=0.3 / 5=0.06 seconds, filter out data segments below the response time consistency threshold. The response time consistency threshold is set based on the online learning platform's requirements for user operation response speed and the user's psychological expectations of response time, and is optimized through large-scale user testing and A / B testing. For example, the response time consistency threshold is set to 0.1 seconds. When the response time change rate is less than 0.1 seconds, users believe that the system responds quickly and stably, and the user experience is good. Experimental data show that when the response time change rate exceeds 0.1 seconds, user complaints increase significantly. Therefore, when the response time change rate Q=0.06 seconds, which is lower than the response time consistency threshold of 0.1 seconds, the data segment is filtered out to generate a response time consistency interval. By calculating the mean absolute deviation between the response time and the average response time, we can accurately quantify the overall fluctuation of the response time and reveal the differences in its changes, thus providing an objective and quantifiable indicator for performance evaluation. The result shows that the response time change rate is 0.06 seconds, which is lower than the response time consistency threshold of 0.1 seconds. This indicates that the response time fluctuation of this data segment is small and the response is relatively stable. Therefore, this data segment is classified as the response time consistency range.

[0025] The performance fluctuation screening submodule calls the response time consistency interval. Based on the fluency fluctuation amplitude and response time consistency, it screens the data segments that deviate from the stability threshold and consistency threshold to generate the performance fluctuation monitoring interval. Call the response time consistency interval, for example, call the data of the response time consistency interval, based on the fluency fluctuation amplitude and response time consistency, for example, assuming that in a certain response time, within the consistency interval, the fluency fluctuation amplitude is 1.5 times, and the response time consistency is 0.06 seconds, filter out data segments that deviate from the stability threshold and consistency threshold, where the stability threshold is the fluency stability threshold 2 times, and the consistency threshold is the response time consistency threshold 0.1 seconds. Compare whether the fluency fluctuation amplitude of 1.5 times deviates from the fluency stability threshold 2 times, that is, 1.5 times is less than 2 times, and there is no deviation. Compare whether the response time consistency of 0.06 seconds deviates from the response time consistency threshold 0.1 seconds, that is, 0.06 seconds is less than 0.1 seconds, and there is no deviation. If any of them deviates, it is considered that the data segment has performance fluctuations, otherwise it is considered that the performance is stable, and a performance fluctuation monitoring interval is generated.

[0026] See also Figure 4 , the anomaly prediction module includes: The trajectory synchronization extraction submodule calls the performance fluctuation monitoring interval to extract the user behavior monitoring component data, extracts the behavior trajectory points and operation change values ​​according to the timestamp, and organizes them into a synchronous data structure to generate a synchronous behavior trajectory set; Call the performance fluctuation monitoring interval, for example, call the data within the performance fluctuation monitoring interval, extract user behavior monitoring component data, for example, extract the user's click, slide, input and other operation data on the online learning platform, extract the behavior trajectory points and operation change values ​​according to the timestamp, for example, the user clicked "Course A" at the timestamp 1718000000, the operation change value is "Enter Course A", slid the page at the timestamp 1718000010, the operation change value is "Page Scroll", and entered "Advanced Mathematics" at the timestamp 1718000020, the operation change value is "Search Keywords", and organize them into a synchronous data structure, for example, organize the timestamps, behavior trajectory points, and operation change values ​​into a list or array in chronological order to generate a synchronous behavior trajectory set.

[0027] The timing-related offset calculation submodule calls the synchronized behavior trajectory set. It identifies the normalized rate of change between the behavior trajectory change amplitude and the response time fluctuation amplitude of the response time change sequence, superimposes the offset term, compensates for the impact of the operation change, and filters the time segments above the abnormal offset threshold to generate the platform-related offset set. By analyzing user behavior trajectories and response time changes, the normalized rate of change of the number of behavior trajectory points and the standard deviation of the response time is calculated to eliminate the dimensional effect. An offset term based on empirical data is then superimposed, such as adding 0.01 seconds to each click operation and 0.03 seconds to each input operation, to compensate for the impact of user operations on system performance. By analyzing historical data, an abnormal offset threshold (such as 0.15 seconds) is set to filter out time segments where the sum of the normalized rate of change and the superimposed offset is higher than this threshold. This method ensures that an early warning is issued only when there is a significant abnormality in the system, avoiding false alarms. For example, if the normalized rate of change is 0.1 seconds, the offset is 0.06 seconds, and the total is 0.16 seconds, which exceeds the threshold, then the time segment is considered to have an abnormal offset, and a platform-associated offset set is generated.

[0028] The abnormal behavior collection submodule calls the platform-related offset set, combines the trajectory change frequency and the operation mutation amplitude, and filters the behavior segments that exceed the abnormal threshold to obtain the platform abnormal behavior identification set; Call the platform-associated offset set, such as the abnormal offset time segment in the platform-associated offset set, and combine the trajectory change frequency and the operation mutation amplitude. For example, count the number of user behavior trajectory points in the abnormal offset time segment as the trajectory change frequency, calculate the number of categories of operation change values ​​or the proportion of important operations as the operation mutation amplitude, and filter out behavior segments that exceed the abnormal threshold. The abnormal threshold is set according to the online learning platform's recognition accuracy requirements and false alarm rate tolerance for abnormal behaviors, and is optimized through manual annotation of historical abnormal behaviors and machine learning model training. For example, the abnormal threshold is set to when the trajectory change frequency is greater than 5 times / second, or the operation mutation amplitude is greater than 2, that is, within an abnormal offset time segment, if the trajectory points of user behavior change more than 5 times per second, or the user operation type (such as a sudden jump from browsing to payment) changes significantly, it is determined to be abnormal behavior. This threshold ensures that only when there is a significant difference between the behavior pattern and the normal pattern is it identified as an abnormality. Experimental data show that under this threshold, the recall rate of abnormal behavior reaches more than 95% and the false alarm rate is less than 5%. For example, the trajectory change frequency of a certain abnormal offset time segment is 6 times / second, and the operation mutation amplitude is 1. At this time, the trajectory change frequency of 6 times / second exceeds the abnormal threshold of 5 times / second. Therefore, this behavior segment is screened out to obtain the platform abnormal behavior identification set.

[0029] See also Figure 5 , the intervention scheduling module includes: The communication delay detection submodule detects the communication delay index and instruction execution interval index between the mini-program end and the server end based on the platform abnormal behavior identification set, records the communication transmission delay and instruction execution response time respectively, and generates a communication delay data set; Based on the platform abnormal behavior identification set, for example, based on the abnormal behavior fragments in the platform abnormal behavior identification set, the communication delay indicators and instruction execution interval indicators between the mini program and the server are detected. For example, by sending a heartbeat packet from the mini program to the server, and recording the time from sending to receiving the server response as the communication transmission delay, the time from the mini program sending an instruction (such as clicking to buy) to the server executing the instruction and returning the result is recorded as the instruction execution response time. The communication transmission delay and the instruction execution response time are recorded separately to generate a communication delay dataset.

[0030] The low-latency screening submodule calls the communication delay dataset, extracts the communication delay and instruction execution interval indicators, screens the time periods that meet the low-latency requirements, identifies the signal sending and receiving delays and instruction execution sending and receiving intervals within each period, and counts the total number of detections using the formula: ; Calculate the intervention time period score, filter out low-scoring time intervals based on the score value, and generate a screening time interval list; in, represents the score of the intervention period, represents the sending delay of the i-th segment signal, represents the reception delay of the i-th segment signal, Represents the instruction execution interval of the i-th signal, Represents the time interval between sending and receiving the i-th signal, Indicates the total number of signals; Call the communication delay dataset, for example, call the communication transmission delay and instruction execution response time data in the communication delay dataset, extract the communication delay and instruction execution interval indicators, for example, extract the first Theoretical sending delay of segment signal , actual sending delay , Theoretical receiving delay and the actual reception delay , filter the time period that meets the low latency requirement. The low latency requirement is set according to the real-time interaction requirements of the online learning platform and the user's expectation of response speed, and is calibrated by comparing industry standards and user experience research. For example, the low latency requirement is set to a difference between the theoretical and actual delays of less than 50 milliseconds, and the square root of the difference between the theoretical and actual receiving delays is greater than 0. This requirement ensures that users feel a smooth and seamless interactive experience when using the mini program. Experimental data shows that when the delay exceeds this threshold, the user experience satisfaction will drop significantly. Identify the signal sending and receiving delays and the instruction execution sending and receiving intervals in each segment, and count the total number of detections. For example, for a certain signal, the theoretical sending delay milliseconds, actual sending delay milliseconds, theoretical receiving delay milliseconds, actual receiving delay Milliseconds, total number of signals ,in, Represents the intervention time period score, which is used to comprehensively evaluate the delay performance of all signals in a time period. Represents the total number of signals, that is, the number of signals detected in the current time period, Representative The theoretical sending delay of the segment signal is the time required for the system to send the expected signal from the applet to the server. Representative The actual sending delay of the segment signal is the time required for the signal to be actually sent from the app end to the server end. Representative The theoretical reception delay of the segment signal is the time required for the system to expect the signal to return from the server to the mini program. Representative The actual reception delay of the segment signal is the time required for the signal to actually return from the server to the mini program. The calculation logic of the formula is that for each signal, the absolute difference between the theoretical sending delay and the actual sending delay is first calculated. , which represents the deviation of the sending link, and then calculate the difference between the theoretical receiving delay and the actual receiving delay , and take the square root of the result, This represents the volatility of the receiving link. The absolute difference in the sending delay is multiplied by one fraction of the receiving delay volatility. That is, this multiplication relationship causes the deviation of the sending delay to be amplified when the receiving delay fluctuation is small (the denominator is small), and to be suppressed when the fluctuation is large. The calculation results of all signals are accumulated to obtain the total intervention time period score, thereby comprehensively evaluating the delay performance of this time period. The formula is used to calculate the intervention time period score. Assume that the data of two signals are detected as shown in Table 3.

[0031] Table 2: Theoretical and actual delay data There are 3 signals in total, for signal 1: , , , For signal 2: , , , For signal 3: , , , Total number of signals , substitute the value into the formula to calculate the intervention period score : ; Low-scoring time intervals are filtered based on the score. The low-scoring time intervals are set based on the online learning platform's tolerance for communication delays and the priority of the system's real-time response, and are calibrated by evaluating system performance under different score values ​​and user experience feedback. For example, the low-scoring time interval is set to when the score is lower than 20, that is, T<20, then the time period is considered to be a low-latency time interval. This score range ensures that only when the communication delay and instruction execution interval are at a good level will it be identified as a low-latency time period. Experimental data shows that when the score exceeds 20, users will perceive obvious lag and delay. Therefore, time intervals with scores lower than 20 are filtered out. For example, the currently calculated intervention time period score is 18.61, which is lower than the low-scoring time interval threshold of 20. Therefore, this time period is filtered as a low-scoring time interval, and a filtered time interval list is generated. The benefit of the formula is that by combining the absolute difference in send delay and the inverse of receive delay fluctuation, it can more sensitively capture delay fluctuations. In particular, when receive delay fluctuations are small, it can amplify the impact of send delay deviation, thereby more accurately identifying time periods requiring intervention. The results show that the intervention time period score is 18.61, which is lower than the low-score time interval threshold of 20. This means that the communication delay and instruction execution interval within this time period both meet the low latency requirements, so this time period is included in the screening time interval list.

[0032] The time period mapping submodule filters the time interval list, matches the time period corresponding to the abnormal behavior, analyzes the time synchronization mapping relationship, and generates an intervention scheduling mapping table; Based on the screening time interval list, for example, based on the list of low-latency time periods in the screening time interval list, match the time periods corresponding to the abnormal behavior. For example, match a low-latency time period in the screening time interval list with the time period of abnormal behavior occurrence recorded in the platform abnormal behavior identification center, analyze the time synchronization mapping relationship, for example, determine whether the time period of abnormal behavior occurrence is completely contained in a low-latency time period, or there is an overlap between the two, and generate an intervention scheduling mapping table.

[0033] See also Figure 6 , the load balancing module includes: The load demand collection submodule calls the intervention scheduling mapping table to collect the platform's real-time task types, processing requirements, and server load occupancy, detects the load proportion of task types, and generates the platform load demand offset rate based on the load baseline rate; The intervention scheduling mapping table is called, for example, the abnormal behavior time period and the corresponding low-latency time period that are successfully matched in the intervention scheduling mapping table are called, and the platform real-time task type, processing requirements, and server load occupancy are collected. For example, it is collected that during a certain abnormal behavior time period, the real-time task type being processed by the platform is "video transcoding", the processing requirements are high computing resources and high I / O, and the server load occupancy is 90% CPU usage and 95% memory usage. The load proportion of the task type is detected. For example, if the current total task load of the platform is 1000 units, of which the video transcoding task occupies 600 units, the load proportion of the video transcoding task is 60%. The load benchmark rate is set based on the average load and system carrying capacity during normal operation of the online learning platform and is verified through statistical analysis and stress testing of historical load data. For example, the load benchmark rate is set to 70% CPU usage and 80% memory usage. When the actual load exceeds the benchmark rate, it indicates that the system load is high. Experimental data shows that under this benchmark rate, the platform performance remains stable, and the platform load demand offset rate is generated.

[0034] The dynamic priority adjustment submodule detects tasks whose demand deviation rate exceeds the threshold based on the platform load demand deviation rate, processing urgency and load demand intensity, and reorders them according to the urgency and deviation amplitude to determine the load distribution priority and obtain the multi-functional platform task priority sequence; Based on the platform load demand offset rate, for example, the CPU load offset rate of 20% and the memory load offset rate of 15% in the platform load demand offset rate, according to the processing urgency and load demand intensity, the processing urgency is set according to the impact of the task on the user experience and the task completion time limit, for example, the video streaming task of the live course has high urgency, while the background data backup task has low urgency, the load demand intensity is set according to the consumption of system resources by the task, for example, the load demand intensity of the video transcoding task is high, and the load demand intensity of the text message sending task is low, the task whose demand offset rate exceeds the threshold is detected, and the threshold is set according to the platform's load tolerance for different task types and the system priority division strategy, and is calibrated by evaluating the task completion time and system performance under different load offset rates, For example, the threshold is set to when the CPU load offset rate exceeds 15% and the memory load offset rate exceeds 10%, it is identified as a high-demand offset. This threshold ensures that priority adjustment is only made when the task's demand for system resources is significant and beyond the norm. Experimental data shows that tasks below this threshold can be completed under normal scheduling, while tasks above this threshold need to be prioritized to avoid performance bottlenecks. For example, the current CPU load offset rate of 20% exceeds the threshold of 15%, and the memory load offset rate of 15% exceeds the threshold of 10%. Therefore, the current task is judged to be a high-demand offset task and is re-sorted according to the urgency and offset amplitude. For example, tasks with high urgency and high demand offset are placed at the front to determine the load distribution priority. For example, the video transcoding task is determined to be the highest priority, and the multi-functional platform task priority sequence is obtained.

[0035] The task control allocation submodule selects channels whose load capacity and time period resources meet the requirements according to the multi-functional platform task priority sequence and the channel parameters and time period parameters in the intervention scheduling mapping table, allocates tasks to the optimal processing channel and execution time period, and generates a platform load balancing control table; Based on the multi-functional platform task priority sequence, for example, the video transcoding task with the highest priority in the multi-functional platform task priority sequence, and based on the channel parameters and time period parameters in the intervention scheduling mapping table, the channel parameters are set based on the server's hardware configuration, network bandwidth, concurrent processing capability, etc. For example, the channel parameters of server A are "high bandwidth, 8-core CPU, 64GB memory." The time period parameters are set based on the platform's real-time load, historical task processing volume, and expected traffic. For example, if a time period parameter is "off-peak period, many idle channels," channels with load capacity and time period resources that meet the requirements are selected. For example, server B (with channel parameters of "high bandwidth, 16-core CPU, 128GB memory") is selected to meet the high computing resource and high I / O requirements of the video transcoding task. Server B also has sufficient idle resources in the current time period. The task is then assigned to the optimal processing channel and execution time period. For example, the video transcoding task is assigned to server B for processing and scheduled for execution during the current off-peak period. This generates a platform load balancing control table.

[0036] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. An online learning platform operation supervision system based on mini-programs, characterized by: The system comprises: The resource allocation module monitors the course access frequency, user behavior patterns, and server load change rate based on data from the course access component, user behavior monitoring component, and server load detection component in the online learning platform, and generates a platform operation status set. The performance evaluation module collects course access fluency and user behavior response time data based on the platform operation status set, identifies the matching between the fluency interval and the response time interval, filters the data segments that deviate from the stable range, and generates the performance fluctuation monitoring interval; The anomaly prediction module calls the performance fluctuation monitoring interval, extracts the behavior trajectory of the corresponding time period in the user behavior monitoring component, analyzes the temporal correlation between the change amplitude of the behavior trajectory and the response time fluctuation amplitude, filters the abnormal behavior fragments, and obtains the platform abnormal behavior identification set; The intervention scheduling module detects the communication delay index and instruction execution interval index between the mini program and the server based on the platform abnormal behavior identification set, selects the low-latency time period as the intervention window, and synchronously maps the time period corresponding to the abnormal behavior with the intervention window to generate an intervention scheduling mapping table.

2. The mini-program-based online learning platform operation supervision system according to claim 1 is characterized in that: The platform operation status set includes course access level, user activity level, and service load abnormality indicators; the performance fluctuation monitoring interval includes fluency fluctuation range, response time variation range, and data deviation segment; the platform abnormal behavior identification set includes abnormal behavior trajectory, abnormal response mode, and abnormal behavior type; the intervention scheduling mapping table includes abnormal time node, intervention window, and synchronization association information.

3. The mini-program-based online learning platform operation supervision system according to claim 1 is characterized in that: The resource allocation module includes: The rate monitoring submodule is based on the data of the course access component, user behavior monitoring component, and server load detection component in the online learning platform. It uses the course access component to detect access frequency data and the user behavior monitoring component to detect behavior change data. It calculates the access frequency, behavior change rate, and load change rate per unit time, and compares them with the rate baseline value to generate the component rate offset. The state identification submodule calls the component rate offset, classifies the data according to the set course access, user activity and service load state rate intervals based on the access frequency, behavior change rate and load change rate, and marks the corresponding state of the data to obtain the real-time operation state set of the platform; The control instruction generation submodule calls the real-time operation status set of the platform, and screens the platform execution resource allocation, task adjustment and record control actions under the corresponding status according to the course access, user activity and service load status types, identifies the corresponding control instructions based on the action threshold associated with the differentiated status, and generates the platform operation status set.

4. The mini-program-based online learning platform operation supervision system according to claim 3 is characterized in that: The performance evaluation module includes: The fluency identification submodule analyzes the change rate and fluctuation range of course access fluency based on the platform operation status set, selects data segments with low change rate and fluctuation range less than the fluency stability threshold, and obtains the fluency stability interval; The response time analysis submodule calls the fluency stability interval, identifies the response time change rate through the synchronously collected user behavior response time data, calculates the difference in response time changes, filters the data segments below the response time consistency threshold, and generates the response time consistency interval; The performance fluctuation screening submodule calls the response time consistency interval, and screens data segments that deviate from the stability threshold and the consistency threshold according to the fluency fluctuation amplitude and the response time consistency, to generate a performance fluctuation monitoring interval.

5. The mini-program-based online learning platform operation supervision system according to claim 4 is characterized in that: The abnormality prediction module includes: The trajectory synchronization extraction submodule calls the performance fluctuation monitoring interval, extracts the user behavior monitoring component data, extracts the behavior trajectory points and operation change values ​​according to the timestamp, and organizes them into a synchronous data structure to generate a synchronous behavior trajectory set; The timing-related offset calculation submodule calls the synchronized behavior trajectory set, identifies the normalized change rate of the behavior trajectory change amplitude and the response time fluctuation amplitude of the response time change sequence, superimposes the offset term, compensates for the impact of the operation change, and filters the time segments above the abnormal offset threshold to generate the platform-related offset set; The abnormal behavior collection submodule calls the platform-related offset set, combines the trajectory change frequency and the operation mutation amplitude, filters the behavior segments that exceed the abnormal threshold, and obtains the platform abnormal behavior identification set.

6. The mini-program-based online learning platform operation supervision system according to claim 5 is characterized in that: The intervention scheduling module includes: The communication delay detection submodule detects the communication delay index and instruction execution interval index between the mini-program end and the server end based on the platform abnormal behavior identification set, records the communication transmission delay and instruction execution response time respectively, and generates a communication delay data set; The low-latency screening submodule calls the communication delay data set, extracts communication delay and instruction execution interval indicators, screens time periods that meet low-latency requirements, identifies signal sending and receiving delays and instruction execution sending and receiving intervals within each period, counts the total number of detections, calculates the intervention time period score, screens low-scoring time intervals based on the score value, and generates a screening time interval list; The time period mapping submodule matches the time period corresponding to the abnormal behavior based on the screening time interval list, analyzes the time synchronization mapping relationship, and generates an intervention scheduling mapping table.

7. The mini-program-based online learning platform operation supervision system according to claim 1 is characterized in that: The system also includes: The load balancing module calls the intervention scheduling mapping table, collects the current server load occupancy and abnormal behavior processing requirements, analyzes the ratio between processing requirements and available load capacity, adjusts task priority sorting rules, selects the optimal processing channel and execution time period, and generates a platform load balancing control table; The platform load balancing control table includes task priority setting, processing channel selection, and execution time arrangement.

8. The mini-program-based online learning platform operation supervision system according to claim 7 is characterized in that: The load balancing module includes: The load demand collection submodule calls the intervention scheduling mapping table to collect the platform's real-time task types, processing requirements, and server load occupancy, detects the load proportion of the task type, and generates a platform load demand offset rate based on the load benchmark rate; The dynamic priority adjustment submodule detects tasks whose demand deviation rate exceeds a threshold based on the platform load demand deviation rate, processing urgency, and load demand intensity, and reorders tasks based on urgency and deviation amplitude to determine load distribution priorities and obtain a multi-functional platform task priority sequence; The task control allocation submodule selects channels whose load capacity and time period resources meet the requirements according to the multi-functional platform task priority sequence and the channel parameters and time period parameters in the intervention scheduling mapping table, allocates tasks to the optimal processing channel and execution time period, and generates a platform load balancing control table.

Images