Data management system and method for digital education platform

Abstract

The invention discloses a data management system and method for a digital education platform, and relates to the technical field of education platform data management, and the method comprises the following steps: collecting the related access data of a network course of the education platform, and carrying out the data cleaning processing; a first index of the network course is calculated, and a school year course access prediction model is established; predicting an access index of each school year of the network course by using a school year course access prediction model to obtain first access index data, and performing first index processing to obtain to-be-compressed course data; carrying out dynamic decompression management on the network course; the method is used for solving the problem that when an existing education platform data management technology stores network courses of an education platform, the network courses cannot be precisely and dynamically decompressed according to access conditions of historical years of the network courses under the condition that normal use of the network courses is not affected, and the network courses cannot be accurately and dynamically decompressed according to the access conditions of the historical years of the network courses. And the storage space of an education platform database is saved.

Description

Technical Field

[0001] The present invention relates to the technical field of educational platform data management, and specifically to a data management system and method for a digital educational platform. Background Art

[0002] Educational platform data management technology refers to a series of technical means and methods used to collect, store, organize, process, analyze, and protect education-related data in the educational platform environment. Among them, the data storage of the educational platform refers to the use of various technical means and devices to efficiently, securely, and reliably store and manage the massive data generated by the educational platform.

[0003] Most of the existing educational platform data management technologies adopt a relatively fixed storage method when storing the online courses of the educational platform, that is, the course resources are stored in the original format all the time after being uploaded, without dynamic adjustment according to the actual usage of the courses; this results in some unpopular courses or low-activity courses occupying a large amount of storage space for a long time, and these spaces are not fully utilized most of the time; for example, some highly specialized niche courses may only be accessed by a few users during a specific period, but they occupy the same storage resources as popular courses, causing idle waste of storage resources; for instance, in the patent application with the publication number CN118760686A, a data storage method and system for a smart educational platform are disclosed. Although this solution can save storage space by removing redundant learning materials, it cannot handle those courses that occupy a large amount of storage space for a long time but are non-redundant and unpopular; moreover, in order to save storage space, some educational platforms may over-compress the courses; although this reduces the storage cost to a certain extent, it will have a negative impact on the user experience; the over-compressed course videos may have problems such as degraded image quality and audio distortion, and during user access, the decompression process may be relatively complex, resulting in a long course loading time; for example, for some high-definition experimental demonstration courses, if over-compressed, students may not be able to clearly observe the experimental details, affecting the learning effect; therefore, when the existing educational platform data management technologies store the online courses of the educational platform, they cannot accurately perform dynamic decompression on the online courses without affecting the normal use of the online courses according to the access situation of the online courses in the historical school years, so as to save the storage space of the educational platform database. Summary of the Invention

[0004] The present invention aims to solve at least one of the technical problems in the prior art to some extent. By collecting relevant access data of online courses on an educational platform, data cleaning and processing are carried out; then the first index of the online courses is calculated, and an annual course access prediction model is established; the annual course access prediction model is used to predict the access index of the online courses each academic year, and the first index processing is carried out to obtain the course data to be compressed; then dynamic decompression management is carried out on the online courses; so as to solve the problem that when the existing educational platform data management technology stores the online courses on the educational platform, it is impossible to accurately perform dynamic decompression on the online courses without affecting the normal use of the online courses according to the access situation in the historical academic years of the online courses, so as to save the storage space of the educational platform database.

[0005] To achieve the above object, in the first aspect, the present application provides a data management system for a digital educational platform, including a data acquisition module, a model construction module, a prediction processing module, and a compression management module;

[0006] The data acquisition module includes a data collection unit and a data cleaning unit. The data collection unit is used to collect relevant access data of the online courses on the educational platform, marked as the first access data, and the data cleaning unit is used to perform cleaning and screening processing to obtain the second access data;

[0007] The model construction module calculates the first index of the online courses based on the second access data and establishes an annual course access prediction model;

[0008] The prediction processing module includes an index prediction unit and an index processing unit. The index prediction unit uses the annual course access prediction model to predict the access index of the online courses each academic year to obtain the first access index data, and the index processing unit is used to perform the first index processing to obtain the course data to be compressed;

[0009] The compression management module performs dynamic decompression management on the online courses based on the course data to be compressed and the first access data.

[0010] Further, the data collection unit is configured with a data collection strategy, and the data collection strategy includes:

[0011] Obtain the total number of daily accesses and the total access duration of all online courses on the educational platform, and calculate the average access duration per access of all online courses per day, denoted as the first average duration;

[0012] For any online course on the educational platform, obtain the relevant access data of the corresponding online course per day, including: obtain the total number of people accessing the corresponding online course per day, the total number of accesses to the corresponding online course per day, and the average access duration per access, and obtain the size of the storage space occupied by the corresponding online course per day;

[0013] Repeatedly obtain the daily relevant access data and the first average duration of all online courses on the education platform, record the acquisition date, classify and store them according to the corresponding online courses, and mark them as the first access data.

[0014] Furthermore, the data cleaning unit is configured with a data cleaning strategy, and the data cleaning strategy includes:

[0015] Based on the first access data, sort the first average duration in the first access data according to the acquisition date, denoted as the first duration sequence, and sort the total number of people accessing the online course, the total number of times of accessing the online course, the average access duration each time, and the size of the storage space occupied by the relevant access data of any online course in the first access data according to the acquisition date, and denote them as the number sequence, the times sequence, the second duration sequence, and the course size sequence in order;

[0016] And perform data cleaning processing on the first duration sequence, the number sequence, the times sequence, the second duration sequence, and the course size sequence respectively, and obtain the second access data after completion;

[0017] The data cleaning processing includes: Sort the corresponding data sorted according to the acquisition date, denoted as the time data sequence. For any data in the time data sequence, denoted as A i , where i represents the serial number in the time data sequence. If A i+n exists; then based on the time data sequence, extract A i to A i+n , denoted as the first sequence, and extract A i to A i-n , denoted as the second sequence, calculate the average value and standard deviation of the first sequence, denoted as B0 and B1 in order, and calculate the average value and standard deviation of the second sequence, denoted as C0 and C1 in order, and judge whether A i simultaneously satisfies |A i -B0|>3*B1 and |A i -C0|>3*C1. If both are satisfied, mark A i as abnormal data and perform replacement processing, otherwise mark A i as normal data; if A i+n does not exist, mark A i as normal data; repeat marking all data in the time data sequence;

[0018] The replacement processing includes: For the abnormal data A i , in the corresponding time data sequence, obtain two normal data closest to A i , denoted as RA i and LAi ; Calculate A through the first replacement formula i 's replacement value TA i , and the first replacement formula is as follows: TA i = q1 * RA i + q2 * LA i , where q1 and q2 are weight coefficients, q1 + q2 = 1, and the value ranges of q1 and q2 are [0, 1]; and use TA i to replace A i .

[0019] Furthermore, the model construction module is configured with a model construction strategy, and the model construction strategy includes:

[0020] For any online course, calculate the first indicator of the corresponding online course daily based on the second access data, and the first indicator includes a course activity indicator and a course consumption ratio indicator;

[0021] Calculating the course activity indicator includes: calculating the course activity indicator of the online course on the current day through the course activity formula, and the course activity formula is as follows: where DC represents the course activity indicator of the online course on the current day, E represents the total number of accesses to the online course on the current day, T represents the average access duration of each access to the online course on the current day, V represents the first average duration on the current day, and F represents the total number of people accessing the online course on the current day;

[0022] Calculating the course consumption ratio indicator includes: calculating the course consumption ratio indicator of the online course on the current day through the resource consumption calculation formula, and the course consumption ratio indicator formula is as follows: where REC represents the course consumption ratio indicator of the online course on the current day, and H represents the size of the storage space occupied by the online course on the current day;

[0023] Repeat calculating the first indicators of all online courses daily, and mark them as the third access data.

[0024] Furthermore, the model construction strategy also includes:

[0025] Perform normalization processing on the third access data according to the data type respectively, scale the sizes of all data in the third access data to [0, 1], and after completion, obtain the fourth access data;

[0026] Construct an initial access prediction model based on the long short-term memory network. The initial access prediction model includes an input layer, an LSTM processing layer, a Dropout layer, and a fully connected layer. The LSTM processing layer includes x1 LSTM layers;

[0027] Set the learning rate for model training to x2, the batch size to x3, and the number of training epochs to x4; use the fourth access data to train the initial access prediction model, and after completion, obtain the academic year course access prediction model.

[0028] Further, the index prediction unit is configured with an index prediction strategy, and the index prediction strategy includes:

[0029] For any online course on the education platform, obtain the course activity index and course consumption ratio index for each day of k1 academic years corresponding to the online course; record it as the first sample data, normalize the first sample data, and input it into the academic year course access prediction model to obtain the course activity index and course consumption ratio index for each day of the next academic year corresponding to the online course, which is recorded as the first academic year prediction data.

[0030] Further, the index processing unit is configured with an index processing strategy, and the index processing strategy includes:

[0031] For the first academic year prediction data of any online course on the education platform, obtain the course activity index and course consumption ratio index for each day in the corresponding first academic year prediction data, and calculate the decompression index for each day through the decompression index formula. The decompression index formula is as follows: DR = q3 * DC + q4 * RC, where DR is the decompression index, q3 and q4 are weight coefficients, q3 + q4 = 1, and the value ranges of q3 and q4 are [0, 1]; repeatedly calculate the decompression index for each day in the corresponding first academic year prediction data, and after completion, obtain the second prediction data;

[0032] Set the decompression threshold to DR0, judge the decompression index for each day in the second prediction data. If the decompression index for the current day is less than or equal to DR0, mark the current day as a compression day, otherwise mark the current day as a decompression day. After completion, obtain the third prediction data;

[0033] Repeatedly obtain the third prediction data of all online courses and mark it as the course data to be compressed.

[0034] Further, the compression management module is configured with a compression management strategy, and the compression management strategy includes:

[0035] Based on the course data to be compressed, divide the third prediction data of any online course according to the month time, and judge and mark each month, including: if the number of days of the compression day in the current month is greater than or equal to k2, mark the current month as a compression month; if the number of days of the compression day in the current month is less than or equal to k3, mark the current month as a decompression month; if the number of days of the compression day in the current month is less than k2 and greater than k3, mark the current month as a fluctuating month. After completion, obtain the compression management data of the corresponding online course.

[0036] Further, the compression management policy further includes:

[0037] For any online course on the education platform, perform decompression management according to the corresponding compression management data, including:

[0038] If the online course is in the compression month, perform compression processing on the online course;

[0039] If the online course is in the decompression month, perform decompression processing on the online course;

[0040] If the online course is in the fluctuation month, if the decompression indicators of the online course for k4 consecutive days are less than or equal to the decompression threshold, perform compression processing on the online course. If the average value of the decompression indicators of the online course for k4 consecutive days is greater than or equal to the decompression threshold, perform decompression processing on the online course.

[0041] In a second aspect, the present application provides a data management method for a digital education platform, including the following steps:

[0042] Collect relevant access data of the online courses on the education platform, mark it as the first access data, and perform data cleaning processing to obtain the second access data;

[0043] Calculate the first indicator of the online course based on the second access data, and establish an academic year course access prediction model;

[0044] Use the academic year course access prediction model to predict the access indicators of the online course for each academic year, obtain the first access indicator data, and perform first indicator processing to obtain the course data to be compressed;

[0045] Perform dynamic decompression management on the online course based on the course data to be compressed and the first access data.

[0046] Advantages of the present invention: By collecting relevant access data of the online courses on the education platform, marking it as the first access data, and performing data cleaning processing to obtain the second access data; calculating the first indicator of the online course based on the second access data, and establishing an academic year course access prediction model; using the academic year course access prediction model to predict the access indicators of the online course for each academic year, obtaining the first access indicator data, and performing first indicator processing to obtain the course data to be compressed; performing dynamic decompression management on the online course based on the course data to be compressed and the first access data; it is possible to accurately perform dynamic decompression on the online course without affecting the normal use of the online course, so as to save the storage space of the education platform database;

[0047] By dynamically decompressing the online courses on the education platform, the present invention has the advantages that it can compress the semesters and holidays when the online courses are not used, which can release a large amount of storage space and reduce the storage cost of the education platform; by calculating the course activity index and course consumption ratio index of the online courses, it can provide accurate quantitative information about the course usage and resource utilization of the online courses for the education platform, improving the accuracy of decompression management of the online courses; the academic year course access prediction model predicts the access index of the online courses, and the advantage is that it can understand the usage and resource demand of the online courses in advance, enabling the education platform to plan in advance and achieve more accurate data management. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 is the principle block diagram of the system of the present invention;

[0049] Figure 2 is the step flow chart of the method of the present invention;

[0050] Figure 3 is the index processing strategy flow chart of the present invention;

[0051] Figure 4 is the decompression management flow chart of the present invention;

[0052] Figure 5 is the structural schematic diagram of the electronic device of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0053] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0054] Embodiment 1. Please refer to Figure 1 As shown, the present application provides a data management system for a digital education platform, including a data acquisition module, a model construction module, a prediction processing module, and a compression management module;

[0055] The data acquisition module includes a data collection unit and a data cleaning unit. The data collection unit is used to collect relevant access data of the online courses on the education platform, marked as the first access data, and the data cleaning unit is used to perform cleaning and screening processing to obtain the second access data;

[0056] The data acquisition unit is configured with a data acquisition strategy, and the data acquisition strategy includes: obtaining the total daily access times and total access durations of all online courses on the education platform, and calculating the average access duration per access for all online courses on a daily basis, that is, the total daily access duration / the total daily access times, denoted as the first average duration;

[0057] For any online course on the education platform, obtain the relevant access data for the corresponding online course on a daily basis, including: obtaining the total number of people accessing the corresponding online course on a daily basis, the total number of times accessing the corresponding online course on a daily basis, and the average access duration per access, and obtaining the size of the storage space occupied by the corresponding online course on a daily basis. The size of the storage space occupied by the online course will change with the change of course content and the accumulation of user interaction data;

[0058] Repeatedly obtain the relevant access data and the first average duration for all online courses on the education platform on a daily basis, record the acquisition date, and store them classified according to the corresponding online courses, marked as the first access data;

[0059] The data cleaning unit is configured with a data cleaning strategy, and the data cleaning strategy includes: based on the first access data, sort the first average duration in the first access data according to the acquisition date, denoted as the first duration sequence, and sort the total number of people accessing the online course, the total number of times accessing the online course, the average access duration per access, and the size of the storage space occupied by any online course in the relevant access data of the first access data according to the acquisition date, denoted as the number sequence, the times sequence, the second duration sequence, and the course size sequence in order;

[0060] And perform data cleaning processing on the first duration sequence, the number sequence, the times sequence, the second duration sequence, and the course size sequence respectively. After completion, obtain the second access data;

[0061] The data cleaning processing includes: sorting the corresponding data sorted according to the acquisition date, including the first duration sequence, the number sequence, the times sequence, the second duration sequence, and the course size sequence; denoted as the time data sequence. For any data in the time data sequence, denoted as A i , where i represents the serial number in the time data sequence. If A i+n exists; then based on the time data sequence, extract A i to A i+n , denoted as the first sequence, and extract A i to A i-n , denoted as the second sequence, calculate the average value and standard deviation of the first sequence, denoted as B0 and B1 in order, and calculate the average value and standard deviation of the second sequence, denoted as C0 and C1 in order;

[0062] Judge Ai Whether both |A i - B0| > 3 * B1 and |A i - C0| > 3 * C1 are satisfied. If both are satisfied, mark A i as abnormal data and perform replacement processing. Otherwise, mark A i as normal data; if A i+n does not exist, mark A i as normal data; repeat the marking for all data in the time data sequence; in this embodiment, n = 6, that is, obtain the sequences of the first 7 days and the last 7 days of A i and judge them separately; the reason for judging as abnormal data only when both |A i - B0| > 3 * B1 and |A i - C0| > 3 * C1 are satisfied is that the data of online courses is affected by factors such as holidays, course start times, and exams, which may cause sudden changes in data. For example, when a holiday starts on a certain day, the number of visitors suddenly decreases. If the data of that day and the previous 6 days without holidays are used to form a sequence for judgment, the data on the day when the holiday starts will be judged as abnormal; by comparing the deviation degree of the data point from the average value and standard deviation of the front and back sequences, abnormal data in the time data sequence can be accurately identified without misjudging normal data that has mutated due to factors such as holidays, course start times, and exams as abnormal data;

[0063] The replacement processing includes: for abnormal data A i , in the corresponding time data sequence, obtain two normal data closest to A i , denoted as RA i and LA i respectively; calculate the replacement value TA i of A i through the first replacement formula. The first replacement formula is as follows: TA i = q1 * RA i + q2 * LA i , where q1 and q2 are weight coefficients, q1 + q2 = 1, and the value ranges of q1 and q2 are [0, 1]; and use TA i to replace A i ; RA i and LA i are preferably the data on both sides of A i . If the distances of RA i and LA i from A i are the same, q1 and q2 can both be set to 0.5. If they are different, the weight coefficient of the data farther away can be appropriately reduced, and the weight coefficient of the data closer can be increased;

[0064] In the specific implementation process, most of the existing education platforms adopt a relatively fixed storage method, that is, the course resources are stored in the original format after uploading. This leads to some unpopular courses or low-activity courses occupying a large amount of storage space for a long time, and these spaces are not fully utilized most of the time; for example, some highly professional niche courses may only be visited by a few users in a specific period, but occupy the same storage resources as popular courses, resulting in a waste of storage resources of the education platform.

[0065] The model building module calculates the first indicator of the online course based on the second access data and establishes a course access prediction model for the academic year;

[0066] The model building module is configured with a model building strategy, which includes: for any online course, calculating the first indicator of the corresponding online course every day based on the second access data, the first indicator including the course activity indicator and the course consumption ratio indicator;

[0067] Calculating the course activity index includes: calculating the course activity index of the online course on that day through the course activity formula. The course activity formula is as follows: DC represents the course activity index of the online course on that day, E represents the total number of visits to the online course on that day, T represents the average visit duration of each visit to the online course on that day, V represents the first average duration of the day, and F represents the total number of people who visit the online course on that day; the course activity index is used to measure the activity level of various interactive behaviors of the online course in one day. E*T reflects the total interaction level of all users on the online course on that day, taking into account the visit frequency and the visit duration of each person. Dividing by V is to eliminate the impact of the difference in visit time caused by the different content nature of different online courses, so that the activity of different online courses is comparable. Multiplying by A is because the number of visitors is the key factor in measuring the activity of online courses. Courses with more visitors are usually more active. By comparing the online course activity indicators of different courses, we can intuitively understand which courses are more active on that day; and online courses with high activity are not suitable for compression;

[0068] Calculating the course consumption ratio index includes: calculating the course consumption ratio index of the online course on that day through the resource consumption calculation formula. The course consumption ratio index formula is as follows: REC represents the course consumption ratio of the online course on that day, and H represents the size of the storage space occupied by the online course on that day. The course consumption ratio reflects the relationship between the daily course resource occupation and actual usage. A higher course consumption ratio means that the course has achieved higher utilization efficiency with relatively less storage resource investment, that is, the course resources are fully utilized. A lower course consumption ratio indicates that the course occupies more storage resources, but the actual usage is relatively low, which may be a waste of storage resources.

[0069] Recalculate the first metric of all online courses daily and mark it as the third access data;

[0070] Normalize the third access data according to the data type, that is, normalize the course activity metric and the course consumption ratio metric in the third access data respectively, and scale all data sizes in the third access data to [0, 1]. After completion, the fourth access data is obtained;

[0071] Construct an initial access prediction model based on the long short-term memory network. The initial access prediction model includes an input layer, an LSTM processing layer, a Dropout layer, and a fully connected layer. The LSTM processing layer includes x1 LSTM layers; in this embodiment, x1 = 2, that is, 2 LSTM layers. x1 can be set according to the actual application scenario, generally 1 - 3 layers, but not too many, as too many LSTM layers will bring an increase in computational cost and the risk of overfitting; the Dropout layer is used to prevent overfitting; during training, it randomly sets the outputs of some neurons to 0, which can force the model to learn features on different subsets of neurons and increase the generalization ability of the model; the Dropout rate is usually set between 0.1 - 0.5. For example, setting it to 0.3 means that 30% of the neuron outputs are randomly discarded in each training iteration;

[0072] Set the learning rate of model training to x2, the batch size to x3, and the number of training epochs to x4; in this embodiment, x2 = 0.001, x3 = 256, x4 = 200. Use the fourth access data to train the initial access prediction model, and after completion, obtain the academic year course access prediction model;

[0073] In the specific implementation process, by calculating the course activity metric and the course consumption ratio metric of the online courses, precise quantitative information about the course usage and resource utilization of the online courses can be provided for the education platform, and it is possible to clearly understand the activity level of each course and the consumption of storage resources, which is convenient for subsequent precise decompression management based on these data.

[0074] The prediction processing module includes a metric prediction unit and a metric processing unit. The metric prediction unit uses the academic year course access prediction model to predict the access metrics of the online courses for each academic year to obtain the first access metric data, and the metric processing unit is used to perform the first metric processing to obtain the course data to be compressed;

[0075] The index prediction unit is configured with an index prediction strategy, and the index prediction strategy includes: for any online course on the education platform, obtain the daily course activity index and course consumption ratio index of the corresponding online course for k1 school years; record it as the first sample data, normalize the first sample data, and input it into the school year course access prediction model to obtain the daily course activity index and course consumption ratio index of the corresponding online course for the next school year, which is recorded as the first prediction data for the school year; in this embodiment, k1 = 3, that is, use the data of 3 school years to predict the data of the next school year;

[0076] The index processing unit is configured with an index processing strategy, and the index processing strategy includes: Please refer to Figure 4 As shown, for the first prediction data of the school year of any online course on the education platform, obtain the daily course activity index and course consumption ratio index in the corresponding first prediction data of the school year, and calculate the daily decompression index through the decompression index formula. The decompression index formula is as follows: DR = q3 * DC + q4 * RC, where DR is the decompression index, q3 and q4 are weight coefficients, q3 + q4 = 1, and the value ranges of q3 and q4 are [0, 1]; repeatedly calculate the daily decompression index in the corresponding first prediction data of the school year to complete and obtain the second prediction data; in this embodiment, q3 = q4 = 0.5, and q3 and q4 can be adjusted according to the actual application scenario. The decompression index after weighted average can be regarded as a quantification of the comprehensive importance of the course; the decompression index obtained by combining these two indexes can more comprehensively evaluate the status of the course in platform resource management and user services, and provide a more accurate basis for the decompression decision of the course;

[0077] Set the decompression threshold as DR0, judge the daily decompression index in the second prediction data. If the decompression index of the current day is less than or equal to DR0, mark the current day as the compression day, otherwise mark the current day as the decompression day, and after completion, obtain the third prediction data;

[0078] In the specific implementation process, the decompression threshold can be set based on the balance of storage cost and experience. If the platform hopes to significantly reduce the storage cost by compressing courses, the decompression threshold DR0 can be set relatively high. For example, through analysis, it is found that when 30% of the course resources are deeply compressed, the storage cost can be reduced by 20%. Then, DR0 can be determined according to this goal, so that a sufficient number of courses are in the compressed state most of the time to achieve the purpose of cost control; however, in order to ensure the learning experience of students, the possibility of compressing popular courses should also be minimized. For example, for courses in the final exam review stage, they should be set to not be compressed or less compressed to ensure that students can review smoothly.

[0079] The compression management module performs dynamic decompression management on the online courses based on the course data to be compressed and the first access data;

[0080] The compression management module is configured with a compression management strategy, which includes: based on the course data to be compressed, dividing the third prediction data of any online course according to the month time, that is, from January to December; and judging and marking any month, including: if the number of days of the compression date in the current month is greater than or equal to k2, then mark the current month as the compression month, if the number of days of the compression date in the current month is less than or equal to k3, then mark the current month as the decompression month, if the number of days of the compression date in the current month is less than k2 and greater than k3, then mark the current month as the fluctuating month. After completion, the compression management data of the corresponding online course is obtained; in this embodiment, k2 = 27, k3 = 15; dividing by month allows the education platform to plan resources from a macro perspective; the usage of courses in different months may vary due to factors such as semester arrangements and holidays; for example, during the school's summer and winter vacations, the activity of most regular courses may decrease, and the resource requirements also decrease accordingly; by dividing by month, the platform can predict and arrange large-scale resource compression in these months in advance, avoiding frequent compression and decompression of the same online course;

[0081] Please refer to Figure 5 As shown, for any online course on the education platform, decompression management is performed according to the corresponding compression management data, including:

[0082] If the online course is in the compression month, compress the online course;

[0083] If the online course is in the decompression month, decompress the online course;

[0084] If the online course is in the fluctuating month, if the decompression indicators of the online course for k4 consecutive days are less than or equal to the decompression threshold, compress the online course to avoid miscompressing an online course that is still in use and causing a decline in the learning experience; if the average value of the decompression indicators of the online course for k4 consecutive days is greater than or equal to the decompression threshold, decompress the online course; decompression processing can be performed in a timely manner when the online course starts; in this embodiment, k4 is 5 days;

[0085] In the specific implementation process, when compressing or decompressing an online course, if the corresponding online course has already been compressed or decompressed, there is no need to perform secondary compression or decompression processing. For example, if an online course has already been compressed, when it is in the compression month, there is no need for secondary compression; after compressing an online course, it does not mean that the online course cannot be watched and learned. It's just that when students watch and learn this online course, the education platform needs to decompress it first, which requires a long loading time.

[0086] Example 2, please refer to Figure 2 As shown in the figure, the present application provides a data management method for a digital education platform, including the following steps:

[0087] Step S1, collect the relevant access data of the online courses on the education platform, mark it as the first access data, and perform data cleaning processing to obtain the second access data; Step S1 includes the following sub-steps:

[0088] Step S101, obtain the total number of daily accesses and the total access duration of all online courses on the education platform, and calculate the average access duration per access of all online courses per day, denoted as the first average duration;

[0089] Step S102, for any online course on the education platform, obtain the relevant access data of the corresponding online course per day, including: obtain the total number of people accessing the corresponding online course per day, the total number of times accessing the corresponding online course per day, and the average access duration per access, and obtain the size of the storage space occupied by the corresponding online course per day;

[0090] Step S103, repeatedly obtain the relevant access data and the first average duration of all online courses on the education platform per day, and record the acquisition date, and store them classified according to the corresponding online courses, marked as the first access data;

[0091] Step S104, based on the first access data, sort the first average duration in the first access data according to the acquisition date, denoted as the first duration sequence, and sort the total number of people accessing the online course, the total number of times accessing the online course, the average access duration per access, and the size of the storage space occupied by any online course in the first access data according to the acquisition date, and denote them as the number sequence, the number of times sequence, the second duration sequence, and the course size sequence in order;

[0092] Step S105, and perform data cleaning processing on the first duration sequence, the number sequence, the number of times sequence, the second duration sequence, and the course size sequence respectively, and after completion, obtain the second access data;

[0093] Step S106, data cleaning processing, Step S106 includes the following sub-steps:

[0094] Step S1061, denote the data sorted according to the acquisition date as the time data sequence, and for any data in the time data sequence, denoted as A i , where i represents the serial number in the time data sequence. If A i+n exists; then based on the time data sequence, extract A i to A i+n , denoted as the first sequence, and extract Ai to A i-n is denoted as the second sequence;

[0095] Step S1062, calculate the average value and standard deviation of the first sequence, denoted as B0 and B1 in sequence, and calculate the average value and standard deviation of the second sequence, denoted as C0 and C1 in sequence;

[0096] Step S1063, determine whether A i simultaneously satisfies |A i - B0| > 3 * B1 and |A i - C0| > 3 * C1. If both are satisfied, mark A i as abnormal data and perform replacement processing. Otherwise, mark A i as normal data; if A i+n does not exist, mark A i as normal data; repeat marking all data in the time data sequence;

[0097] Step S1064, the replacement processing includes: for the abnormal data A i , in the corresponding time data sequence, obtain two normal data closest to A i , denoted as RA i and LA i respectively; calculate the replacement value TA i of A through the first replacement formula. The first replacement formula is as follows: TA i = q1 * RA i + q2 * LA i , where q1 and q2 are weight coefficients, q1 + q2 = 1, and the value ranges of q1 and q2 are [0, 1]; and use TA i to replace A i . i .

[0098] Step S2, calculate the first index of the online course based on the second access data and establish an academic year course access prediction model; Step S2 includes the following sub-steps:

[0099] Step S201, for any online course, calculate the corresponding first index of the online course daily based on the second access data. The first index includes a course activity index and a course consumption ratio index;

[0100] Step S202, calculating the course activity index includes: calculating the course activity index of the online course on the current day through the course activity formula. The course activity formula is as follows: Among them, DC represents the course activity index of the online course on the current day, E represents the total number of visits to the online course on the current day, T represents the average visit duration per visit to the online course on the current day, V represents the first average duration on the current day, and F represents the total number of people accessing the online course on the current day;

[0101] Step S203, calculating the course consumption ratio index includes: calculating the course consumption ratio index of the online course on the current day through the resource consumption calculation formula. The formula for the course consumption ratio index is as follows: Among them, REC represents the course consumption ratio index of the online course on the current day, and H represents the size of the storage space occupied by the online course on the current day;

[0102] Step S204, repeatedly calculate the first index of each online course daily and mark it as the third access data;

[0103] Step S205, perform normalization processing on the third access data according to the data type respectively, scale the size of all data in the third access data to [0, 1], and after completion, obtain the fourth access data;

[0104] Step S206, construct an initial access prediction model based on the long short-term memory network. The initial access prediction model includes an input layer, an LSTM processing layer, a Dropout layer, and a fully connected layer. The LSTM processing layer includes x1 LSTM layers;

[0105] Step S207, set the learning rate size of model training to x2, the batch size to x3, and the number of training epochs to x4; use the fourth access data to train the initial access prediction model, and after completion, obtain the academic year course access prediction model.

[0106] Step S3, use the academic year course access prediction model to predict the access indicators of each online course per academic year, obtain the first access indicator data, and perform the first indicator processing to obtain the course data to be compressed; Step S3 includes the following sub-steps:

[0107] Step S301, for any online course on the education platform, obtain the course activity index and the course consumption ratio index of the corresponding online course for k1 academic years daily; denoted as the first sample data;

[0108] Step S302, perform normalization processing on the first sample data and input it into the academic year course access prediction model to obtain the course activity index and the course consumption ratio index of the corresponding online course for the next academic year daily, denoted as the academic year first prediction data;

[0109] Step S303: For the first-year prediction data of any online course on the education platform, obtain the daily course activity index and course consumption ratio index in the corresponding first-year prediction data, and calculate the daily decompression index through the decompression index formula. The decompression index formula is as follows: DR = q3 * DC + q4 * RC, where DR is the decompression index, q3 and q4 are weight coefficients, q3 + q4 = 1, and the value ranges of q3 and q4 are [0, 1].

[0110] Step S304: Repeatedly calculate the daily decompression index in the corresponding first-year prediction data to obtain the second prediction data.

[0111] Step S305: Set the decompression threshold as DR0, and judge the daily decompression index in the second prediction data. If the daily decompression index is less than or equal to DR0, mark the current day as a compression day; otherwise, mark the current day as a decompression day. After completion, obtain the third prediction data.

[0112] Step S306: Repeatedly obtain the third prediction data of all online courses and mark them as the course data to be compressed.

[0113] Step S4: Perform dynamic decompression management on the online courses based on the course data to be compressed and the first access data. Step S4 includes the following sub-steps:

[0114] Step S401: Based on the course data to be compressed, divide the third prediction data of any online course according to the monthly time, and judge and mark any month, including: if the number of days of compression days in the current month is greater than or equal to k2, mark the current month as a compression month; if the number of days of compression days in the current month is less than or equal to k3, mark the current month as a decompression month; if the number of days of compression days in the current month is less than k2 and greater than k3, mark the current month as a fluctuating month. After completion, obtain the compression management data of the course.

[0115] Step S402: For any online course on the education platform corresponding to the network, perform decompression management according to the corresponding compression management data. Step S402 includes the following sub-steps:

[0116] Step S4021: If the online course is in a compression month, perform compression processing on the online course.

[0117] Step S4022: If the online course is in a decompression month, perform decompression processing on the online course.

[0118] Step S4023: If the online course is in a fluctuating month, if the decompression index of the online course for k4 consecutive days is less than or equal to the decompression threshold, perform compression processing on the online course; if the average value of the decompression index of the online course for k4 consecutive days is greater than or equal to the decompression threshold, perform decompression processing on the online course.

[0119] Example 3. Refer to Figure 5 as shown Figure 5 which illustrates a schematic structural diagram of an electronic device. The electronic device may include: a processor, a communication interface, a memory, and a communication bus. Among them, the processor, the communication interface, and the memory complete mutual communication through the communication bus. The memory stores computer-readable instructions. The processor can call the instructions in the memory. When the computer-readable instructions are executed by the processor, the steps in a data management method of a digital education platform are run to achieve the following functions: collecting relevant access data of online courses on the education platform, marking it as first access data, and performing data cleaning processing to obtain second access data; calculating a first index of the online courses based on the second access data, and establishing an academic year course access prediction model; using the academic year course access prediction model to predict the access index of the online courses for each academic year, obtaining first access index data, and performing first index processing to obtain course data to be compressed; performing dynamic decompression management on the online courses based on the course data to be compressed and the first access data.

[0120] In addition, when the logical instructions in the above-mentioned memory are implemented in the form of software function units and sold or used as an independent product, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present application. The foregoing storage medium includes: USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs, etc., which can store program codes.

[0121] Embodiment 4. The present application further provides a computer-readable storage medium. The present application provides a storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps in the above data management method of a digital education platform are run to achieve the following functions: collecting relevant access data of the online courses on the education platform, marking them as first access data, and performing data cleaning processing to obtain second access data; calculating the first metrics of the online courses based on the second access data, and establishing an academic year course access prediction model; using the academic year course access prediction model to predict the access metrics of the online courses for each academic year, obtaining first access metric data, and performing first metric processing to obtain data of courses to be compressed; performing dynamic decompression management on the online courses based on the data of courses to be compressed and the first access data.

[0122] Through the description of the above embodiments, the embodiments of the present invention can be provided as a method, a system or a computer program product. Based on such an understanding, the above technical solutions, in essence, or the part that contributes to the prior art can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disc, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0123] In the embodiments provided by the present application, it should be understood that the disclosed system or method can be implemented in other ways. The above-described embodiments are merely illustrative. For example, the division of modules or units is only a logical function division, and there can be other division methods in actual implementation. Also, for example, multiple modules or units can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some communication interfaces. The indirect couplings or communication connections of systems, modules, and units can be electrical, mechanical, or other forms.

[0124] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of each embodiment of the present application.

Claims

1. A data management system for a digital education platform, characterized in that: It includes a data acquisition module, a model building module, a prediction processing module and a compression management module; The data acquisition module includes a data acquisition unit and a data cleaning unit. The data acquisition unit is used to collect relevant access data of online courses on the education platform, marked as first access data, and the data cleaning unit is used to perform cleaning and screening processing to obtain second access data. The model building module calculates the first indicator of the online course based on the second access data and establishes a course access prediction model for the academic year; The prediction processing module includes an indicator prediction unit and an indicator processing unit. The indicator prediction unit predicts the access indicator of the online course in each school year by using the school year course access prediction model to obtain the first access indicator data. The indicator processing unit is used to perform the first indicator processing to obtain the course data to be compressed. The compression management module performs dynamic decompression management on the network course based on the course data to be compressed and the first access data.

2. The data management system of a digital education platform according to claim 1, characterized in that: The data collection unit is configured with a data collection strategy, which includes: Obtain the total number of visits and total visit duration of all online courses on the education platform every day, and calculate the average visit duration of all online courses every day, which is recorded as the first average duration; For any online course on the education platform, obtain the relevant daily access data of the corresponding online course, including: obtain the total number of people who visit the corresponding online course every day, the total number of visits to the corresponding online course every day and the average duration of each visit, and obtain the size of the storage space occupied by the corresponding online course every day; Repeatedly obtain the daily relevant access data and the first average duration of all online courses on the education platform, record the acquisition date, classify and store them according to the corresponding online courses, and mark them as the first access data.

3. The data management system of a digital education platform according to claim 2, characterized in that: The data cleaning unit is configured with a data cleaning strategy, which includes: Based on the first access data, the first average duration in the first access data is sorted according to the acquisition date and recorded as a first duration sequence, and the total number of people who visited the online course, the total number of times the online course was visited, the average duration of each visit and the size of the storage space occupied in the access data related to any online course in the first access data are sorted according to the acquisition date and recorded in order as a number sequence, a number sequence, a second duration sequence and a course size sequence; The first duration sequence, the number of people sequence, the number of times sequence, the second duration sequence, and the course size sequence are cleaned and processed respectively, and the second access data is obtained after completion; The data cleaning process includes: recording the corresponding data sorted by the acquisition date as a time data sequence, and recording any data in the time data sequence as A i , where i represents the sequence number in the time data series. If A i+n If exists, then based on the time data series, extract A i To A i+n , recorded as the first sequence, and extract A i To A i-n , recorded as the second sequence, calculate the mean and standard deviation of the first sequence, recorded as B0 and B1 respectively, and calculate the mean and standard deviation of the second sequence, recorded as C0 and C1 respectively, determine A i Does it satisfy |A i -B0|>3*B1 and |A i -C0|>3*C1, if both are satisfied, mark A i is abnormal data and is replaced, otherwise it is marked as A i is normal data; if A i+n If it does not exist, mark A i is normal data; repeat marking all data in the time data series; The replacement process includes: for abnormal data A i , in the corresponding time data sequence, obtain two distances A i The most recent normal data are recorded as RA i and LA i ; Calculate A by the first substitution formula i The replacement value of TA i , the first replacement formula is as follows: TA i =q1*RA i +q2*LA i , where q1 and q2 are weight coefficients, q1+q2=1, and the value range of q1 and q2 is [0, 1]; and using TA i Replace A i .

4. The data management system of a digital education platform according to claim 3, characterized in that: The model building module is configured with a model building strategy, which includes: For any online course, the first indicator of the corresponding online course is calculated based on the second access data, and the first indicator includes a course activity indicator and a course consumption ratio indicator; Calculating the course activity index includes: calculating the course activity index of the online course on that day through the course activity formula. The course activity formula is as follows: Among them, DC represents the course activity index of the online course on that day, E represents the total number of visits to the online course on that day, T represents the average visit duration of each visit to the online course on that day, V represents the first average duration of the day, and F represents the total number of people who visited the online course on that day; Calculating the course consumption ratio index includes: calculating the course consumption ratio index of the online course on that day through the resource consumption calculation formula. The course consumption ratio index formula is as follows: Among them, REC represents the course consumption ratio index of the online course on that day, and H represents the size of the storage space occupied by the online course on that day; Repeat the calculation of the first indicator for all online courses every day, marked as the third visit data.

5. The data management system of a digital education platform according to claim 4, characterized in that: Model building strategies also include: Normalizing the third access data according to the data type, scaling all data sizes in the third access data to [0, 1], and obtaining fourth access data after completion; An initial access prediction model is constructed based on a long short-term memory network. The initial access prediction model includes an input layer, an LSTM processing layer, a Dropout layer, and a fully connected layer. The LSTM processing layer includes x1 LSTM layers. Set the learning rate size of the model training to x2, the batch size to x3, and the number of training rounds to x4; use the fourth visit data to train the initial visit prediction model, and after completion, obtain the academic year course visit prediction model.

6. The data management system of a digital education platform according to claim 5, characterized in that: The indicator prediction unit is configured with an indicator prediction strategy, which includes: For any online course on the education platform, obtain the daily course activity index and course consumption ratio index of the corresponding online course for k1 academic years; record them as the first sample data, normalize the first sample data, and input them into the academic year course access prediction model to obtain the daily course activity index and course consumption ratio index of the corresponding online course in the next academic year, which are recorded as the first predicted data of the academic year.

7. The data management system of a digital education platform according to claim 6, characterized in that: The indicator processing unit is configured with an indicator processing strategy, which includes: For the first forecast data of the school year for any online course of the education platform, obtain the daily course activity index and course consumption ratio index in the corresponding first forecast data of the school year, and calculate the daily decompression index through the decompression index formula. The decompression index formula is as follows: DR = q3*DC+q4*RC, where DR is the decompression index, q3 and q4 are weight coefficients, q3+q4=1, and the value range of q3 and q4 is [0, 1]; repeatedly calculate the daily decompression index in the corresponding first forecast data of the school year to obtain the second forecast data; Set the decompression threshold to DR0, and judge the daily decompression index in the second prediction data. If the decompression index of the day is less than or equal to DR0, mark the day as a compression day, otherwise mark the day as a decompression day. After completion, obtain the third prediction data; Repeat the process of obtaining the third prediction data of all online courses and marking them as course data to be compressed.

8. The data management system of a digital education platform according to claim 7, characterized in that: The compression management module is configured with compression management strategies, which include: Based on the course data to be compressed, the third predicted data of any online course is divided according to the month time, and any month is judged and marked, including: if the number of compressed days in the month is greater than or equal to k2, then the month is marked as a compressed month; if the number of compressed days in the month is less than or equal to k3, then the month is marked as a decompression month; if the number of compressed days in the month is less than k2 and greater than k3, then the month is marked as a fluctuating month. After completion, the compression management data of the corresponding online course is obtained.

9. The data management system of a digital education platform according to claim 8, characterized in that: Compression management strategies also include: For any online course on the education platform, decompression management is performed according to the corresponding compression management data, including: If the online course is in a compressed month, the online course will be compressed; If the online course is in the decompression month, the online course will be decompressed; If the online course is in a fluctuating month, if the decompression index of the online course for k4 consecutive days is less than or equal to the decompression threshold, the online course will be compressed; if the average value of the decompression index of the online course for k4 consecutive days is greater than or equal to the decompression threshold, the online course will be decompressed.

10. A data management method for a digital education platform, used to implement the data management system of any digital education platform of claims 1-9, characterized in that: The steps include: Collect relevant access data of online courses on the education platform, mark it as first access data, and perform data cleaning to obtain second access data; Calculate the first index of the online course based on the second access data and establish a course access prediction model for the academic year; Using the academic year course access prediction model to predict the access index of the online course in each academic year, obtaining the first access index data, and performing the first index processing to obtain the course data to be compressed; Dynamically decompress and manage the online course based on the course data to be compressed and the first access data.

Images