Teaching platform management system and method based on multi-source data analysis

Abstract

The invention discloses a teaching platform management system and method based on multi-source data analysis, and relates to the technical field of teaching platform management. Learning features of a target user are analyzed based on historical online learning information of the target user; constructing an attention assessment model based on the learning features of the target user, extracting historical offline cache learning course data of the target user, splitting and analyzing the historical offline cache learning course data of the target user, and constructing a historical offline cache learning analysis pair set; performing data screening and abnormal point elimination on the historical offline cache learning analysis pair set to generate a second historical offline cache learning analysis pair set, and constructing an evaluation duration association model based on the second historical offline cache learning analysis pair set; and splitting and putting management is carried out on the real-time offline cache learning course based on the evaluation duration association model. The accuracy and robustness of the system are improved; and learning efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of teaching platform management, and specifically to a teaching platform management system and method based on multi-source data analysis. Background Art

[0002] With the rapid development of Internet technology, teaching platform education, as a form of information-based education, has gradually become a major highlight in the education field with its characteristics of convenience, high efficiency, and personalization. However, behind the rapid development, the development of educational platforms still faces many challenges and problems.

[0003] The teaching platform has online course learning and offline course learning modes. Existing teaching platform management often only targets online course learning and ignores the management of the offline course learning mode; the existing teaching platform management adopts a one-time delivery mode for the learning resources of offline cached learning courses, but users have a low attention level during the learning of offline cached learning courses, and the one-time delivery mode cannot intervene in the situation of low learning efficiency.

[0004] Therefore, the present invention discloses a teaching platform management system and method based on multi-source data analysis to solve the above problems. Summary of the Invention

[0005] The purpose of the present invention is to provide a teaching platform management system and method based on multi-source data analysis to solve the problems raised in the prior art.

[0006] To achieve the above purpose, the present invention provides the following technical solution: A teaching platform management method based on multi-source data analysis, the method includes the following steps: S1: Obtain the authorization information of the target user, extract the historical online learning information of the target user based on the authorization information of the target user, and analyze the learning characteristics of the target user based on the historical online learning information of the target user; S2: Construct an attention evaluation model based on the learning characteristics of the target user, extract the historical offline cached learning course data of the target user based on the authorization information of the target user, split and analyze the historical offline cached learning course data of the target user, and construct a set of historical offline cached learning analysis pairs; S3: Perform data screening and outlier elimination on the set of historical offline cached learning analysis pairs to generate a second set of historical offline cached learning analysis pairs, and construct an evaluation duration association model based on the second set of historical offline cached learning analysis pairs; S4: When it is detected that the target user starts to learn a real-time offline cached learning course, perform split delivery management on the real-time offline cached learning course based on the evaluation duration association model.

[0007] According to the above solution, in S1, it includes the following content: S101: Obtain authorization information of a target user, and extract facial biometric data and post-class answer data of the target user during historical online learning based on the target user's authorization information; the facial biometric data and post-class answer data are in one-to-one correspondence; obtain the historical exam scores of the target user for the courses corresponding to the post-class answer data, and record the ratio of the post-class answer scores to the historical exam average score as the historical online learning evaluation value; the total score of each post-class answer score is the same as the total score of the historical exam; S102: Extract the coordinates of facial key points for each frame of image based on facial biometric data; analyze the eye aspect ratio and the variance of the eye aspect ratio for each frame of image based on the coordinates of the eye key points in the facial key points, and mark the eye as closed if the eye aspect ratio is less than the corresponding threshold; analyze the blinking frequency according to the eye closing situation; analyze the intensity of the frowning action based on the FACS system, and analyze the mean and variance of the intensity of the frowning action; obtain the head posture angle vector based on the facial biometric data, and analyze the average absolute angle change rate, which is equal to the ratio of the total head yaw angle offset to the total number of image frames; analyze the intensity of the frowning action and the covariance of the head yaw angle.

[0008] This application simultaneously collects facial biometric data (physiological reactions) and after-class question-answering data (cognitive performance) to establish the correlation between physiological behavior and learning outcomes; it breaks through the limitations of traditional evaluation that relies solely on question-answering data, provides a more comprehensive analysis of attention states, captures subtle changes in attention through multi-dimensional biometric features, avoids misjudgment of a single indicator, and improves the accuracy and robustness of the system.

[0009] According to the above scheme, S2 includes the following contents: S201: Recording the mean and variance of the intensity of the frowning movement, the blinking frequency, the variance of the eye aspect ratio, the average absolute angular change rate, and the covariance between the intensity of the frowning movement and the head yaw angle as attention features; constructing an attention evaluation model based on the attention features and the corresponding historical online learning evaluation values, wherein the mean and variance of the intensity of the frowning movement, the blinking frequency, the variance of the eye aspect ratio, the average absolute angular change rate, and the covariance between the intensity of the frowning movement and the head yaw angle are independent variables, and the historical online learning evaluation values are dependent variables, and fitting the weight coefficients of each feature in the attention evaluation model based on the least squares method; S202: extracting the target user's historical offline cached learning course data based on the target user's authorization information; Mark the time points at which target user interaction data appears in historical offline cached learning courses; the target user interaction data includes active user interaction and passive user interaction; active user interaction includes active course acceleration, active course deceleration, active course pause, active course start, active course fast forward, and active course rewind; passive user interaction includes automatic course pause caused by offline cached learning evaluation value being less than the corresponding threshold; Among them, active course fast forward means that the target user actively adjusts the time progress of the historical offline cache learning course forward; active course rewind means that the target user actively adjusts the time progress of the historical offline cache learning course backward; S203: Segment the historical offline cached learning courses based on each marked time point to generate historical offline cached learning course segments; extract facial biometric data corresponding to the historical offline cached learning course segments, analyze the attention features of the historical offline cached learning course segments based on the facial biometric data corresponding to the historical offline cached learning course segments, substitute the attention features of the historical offline cached learning course segments into the fitted attention evaluation model to generate historical offline cached learning evaluation values for the historical offline cached learning course segments; generate historical offline cached learning analysis pairs based on the historical offline cached learning evaluation values and the duration of the corresponding historical offline cached learning course segments, and construct a historical offline cached learning analysis pair set based on the historical offline cached learning analysis pairs.

[0010] An attention evaluation model for online learning is constructed based on attention features. The time points of user control behaviors are marked, and learning courses are segmented. A set of historical offline cached learning analysis pairs with duration is analyzed and generated to provide a data basis for subsequent personalized course tailoring.

[0011] According to the above solution, S3 includes the following: S301: Extract the time point at which each active course in each historical offline cached learning course is retracted, and the system sets a first preset time window; use the time point at which the active course in the historical offline cached learning course is retracted as the end time point of the trimmed course segment, and cut off the historical offline cached learning course of the first preset time window length; analyze the historical offline cached learning evaluation value of each trimmed course segment, and record the average historical offline cached learning evaluation value of the trimmed course segment as the trimming threshold; S302: Extract a set of historical offline cache learning analysis pairs, delete the historical offline cache learning analysis pairs whose historical offline cache learning evaluation values are less than a deletion threshold; arrange the remaining historical offline cache learning analysis pairs in ascending order of historical offline cache learning evaluation values; if any two historical offline cache learning analysis pairs have the same historical offline cache learning evaluation values but different durations of the corresponding historical offline cache learning course segments, only retain the historical offline cache learning analysis pair with the shortest duration of the historical offline cache learning course segment to generate a first set of historical offline cache learning analysis pairs; perform outlier detection on the first set of historical offline cache learning analysis pairs and delete the outliers to generate a second set of historical offline cache learning analysis pairs; S303: Based on the second historical offline cache learning analysis, an evaluation duration association model is constructed for the set, y=α1×x+α2; wherein α1 and α2 represent fitting coefficients, x represents the independent variable of the historical offline cache learning evaluation value, and y represents the dependent variable of the brightness value of the non-lighting equipment. The least squares method is used to calculate and solve α1 and α2 in the evaluation duration association model.

[0012] Active course rewind behavior is used to locate inefficient segments, dynamically calculate deletion thresholds, avoid subjective threshold settings, and reflect real learning pain points; improve data set quality, prevent overfitting, associate evaluation values with learning time, reduce real-time computing load, and meet the real-time response requirements of the teaching platform.

[0013] According to the above scheme, S4 includes the following contents: S401: When it is detected that the target user starts to study the real-time offline cache learning course, a real-time offline cache learning course segment of a first duration is intercepted from the real-time offline cache learning course and delivered to the target user for learning; the first duration is equal to the product of the system average analysis duration and a preset constant, and the system average analysis duration is the average analysis time from obtaining facial biometric data to generating the offline cache learning course duration; while the target user is studying the real-time offline cache learning course segment of the first duration, the facial biometric data of the target user is obtained in real time and the corresponding real-time offline cache learning evaluation value is analyzed, and the real-time offline cache learning evaluation value is substituted into the evaluation duration association model to generate the corresponding real-time offline cache learning course duration; after the first duration, a real-time offline cache learning course segment of the real-time offline cache learning course duration is intercepted and delivered to the target user for learning; S402: During the learning of the real-time offline cached learning course segment with the duration of the real-time offline cached learning course, the system splits the duration of the real-time offline cached learning course based on the first duration to generate real-time offline cached learning sub-course segments, analyzes the real-time offline cached learning evaluation values for the real-time offline cached learning sub-course segments. If the real-time offline cached learning evaluation value is less than the corresponding threshold, the system automatically pauses the real-time offline cached learning course segment; reserves the first duration at the end of the real-time offline cached learning course segment with the duration of the real-time offline cached learning course, intercepts the real-time offline cached learning course segment with the second duration before reserving the first duration, where the second duration is the product of the first duration and a preset constant; analyzes the corresponding real-time offline cached learning evaluation value for the real-time offline cached learning course segment with the second duration, substitutes the real-time offline cached learning evaluation value corresponding to the real-time offline cached learning course segment with the second duration into the evaluation duration correlation model to generate the duration of the corresponding real-time offline cached learning course, and delivers the duration of the real-time offline cached learning course corresponding to the real-time offline cached learning evaluation value corresponding to the real-time offline cached learning course segment with the second duration to the target user for learning until the target user finishes learning.

[0014] Based on the initial analysis result of the first duration, dynamically adjust the length of subsequent course segments; achieve a closed loop of analysis - delivery - re-optimization to adapt to attention fluctuations; real-time detect the evaluation value of sub-course segments and automatically pause the content below the threshold; timely prevent ineffective learning (such as forcing continuation when attention is distracted) and improve learning efficiency.

[0015] Another aspect of the present application provides a teaching platform management system based on multi-source data analysis. The system is implemented by applying the above-mentioned teaching platform management method based on multi-source data analysis. The system includes a learning feature analysis module, an attention evaluation analysis module, an evaluation duration correlation module, and a course delivery management module; The learning feature analysis module is used to obtain the target user authorization information, extract the historical online learning information of the target user based on the target user authorization information, and analyze the learning features of the target user based on the historical online learning information of the target user; The attention evaluation analysis module is used to construct an attention evaluation model based on the learning features of the target user, extract the historical offline cached learning course data of the target user based on the target user authorization information, split and analyze the historical offline cached learning course data of the target user, and construct a historical offline cached learning analysis pair set; The evaluation duration correlation module is used to screen data and eliminate outliers from the historical offline cached learning analysis pair set to generate a second historical offline cached learning analysis pair set, and construct an evaluation duration correlation model based on the second historical offline cached learning analysis pair set; The course delivery management module is used to split and deliver the real-time offline cache learning course based on the evaluation duration association model when it is detected that the target user starts to study the real-time offline cache learning course.

[0016] According to the above solution, the learning feature analysis module includes a question answering data analysis unit and a facial biometric data analysis unit; The answer data analysis unit is used to obtain the target user's authorization information, and extract the target user's post-class answer data in the history online learning based on the target user's authorization information; obtain the history test scores of the course corresponding to the target user and the post-class answer data, and record the ratio of the post-class answer scores to the history test average score as the history online learning evaluation value; the total score of each post-class answer score is the same as the total score of the history test score; The facial biometric data analysis unit is used to extract the coordinates of facial key points for each frame of image based on the facial biometric data; analyze the eye aspect ratio and the variance of the eye aspect ratio for each frame of image based on the coordinates of the eye key points in the facial key points, and mark the eye as closed if the eye aspect ratio is less than the corresponding threshold; analyze the blinking frequency according to the eye closing situation; analyze the intensity of the frowning action based on the FACS system, and analyze the mean and variance of the intensity of the frowning action; obtain the head posture angle vector based on the facial biometric data, and analyze the average absolute angle change rate, which is equal to the ratio of the total head yaw angle offset to the total number of image frames; analyze the intensity of the frowning action and the covariance of the head yaw angle.

[0017] According to the above solution, the attention assessment analysis module includes an attention assessment model construction unit and a learning analysis unit; The attention assessment model construction unit is used to record the mean and variance of the intensity of the frowning action, the blinking frequency, the variance of the eye aspect ratio, the average absolute angle change rate, and the covariance of the intensity of the frowning action and the head yaw angle as attention features; construct an attention assessment model based on the attention features and the corresponding historical online learning evaluation values, in which the mean and variance of the intensity of the frowning action, the blinking frequency, the variance of the eye aspect ratio, the average absolute angle change rate, and the covariance of the intensity of the frowning action and the head yaw angle are independent variables, the historical online learning evaluation values are dependent variables, and the weight coefficients of each feature in the attention assessment model are fitted based on the least squares method; The learning analysis pair analysis unit extracts the target user's historical offline cached learning course data based on the target user's authorization information; marks the time points at which the target user's interaction data appears in the historical offline cached learning course; segments the historical offline cached learning course based on each marked time point to generate historical offline cached learning course segments; extracts facial biometric data corresponding to the historical offline cached learning course segments, analyzes the attention features of the historical offline cached learning course segments based on the facial biometric data corresponding to the historical offline cached learning course segments, substitutes the attention features of the historical offline cached learning course segments into the fitted attention evaluation model to generate a historical offline cached learning evaluation value of the historical offline cached learning course segments; generates a historical offline cached learning analysis pair based on the historical offline cached learning evaluation value and the duration of the corresponding historical offline cached learning course segment, and constructs a historical offline cached learning analysis pair set based on the historical offline cached learning analysis pairs.

[0018] According to the above solution, the evaluation duration association module includes a learning analysis pair screening unit and an evaluation duration association model construction unit; The learning analysis and screening unit extracts the time point at which each active course in each historical offline cache learning course is regressed, and the system sets a first preset time window; uses the time point at which the active course in the historical offline cache learning course is regressed as the end time point of the clipped course segment, and intercepts the historical offline cache learning course of the first preset time window length; analyzes the historical offline cache learning evaluation value of each clipped course segment, and records the average historical offline cache learning evaluation value of the clipped course segment as the deletion threshold. Extract a set of historical offline cache learning analysis pairs, delete the historical offline cache learning analysis pairs whose historical offline cache learning evaluation values are less than a deletion threshold; arrange the remaining historical offline cache learning analysis pairs in ascending order of historical offline cache learning evaluation values; if any two historical offline cache learning analysis pairs have the same historical offline cache learning evaluation values but different durations of the corresponding historical offline cache learning course segments, only retain the historical offline cache learning analysis pair with the shortest duration of the historical offline cache learning course segment to generate a first set of historical offline cache learning analysis pairs; perform outlier detection on the first set of historical offline cache learning analysis pairs and delete the outliers to generate a second set of historical offline cache learning analysis pairs; The evaluation duration association model construction unit is used to construct an evaluation duration association model for the set based on the second historical offline cache learning analysis.

[0019] According to the above solution, the course delivery management module includes a first delivery management unit and a second delivery management unit; The first delivery management unit is used to, when detecting that the target user starts to learn the real-time offline cached learning course, intercept a real-time offline cached learning course segment of the first duration from the real-time offline cached learning course and deliver it to the target user for learning; the first duration is equal to the product of the system average analysis duration and a preset constant, and the system average analysis duration is the average analysis time from obtaining the facial biometric data to generating the offline cached learning course duration; during the learning of the real-time offline cached learning course segment of the first duration, the facial biometric data of the target user is obtained in real time and the corresponding real-time offline cached learning evaluation value is analyzed, and the real-time offline cached learning evaluation value is substituted into the evaluation duration association model to generate the duration of the corresponding real-time offline cached learning course; after the first duration, a real-time offline cached learning course segment of the duration of the real-time offline cached learning course is intercepted and delivered to the target user for learning; The second delivery management unit is used to, during the learning of the real-time offline cached learning course segment of the duration of the real-time offline cached learning course, the system splits the duration of the real-time offline cached learning course into real-time offline cached learning sub-course segments based on the first duration, analyzes the real-time offline cached learning evaluation value of the real-time offline cached learning sub-course segments, and if the real-time offline cached learning evaluation value is less than the corresponding threshold, automatically pauses the course for the real-time offline cached learning course segment; reserve the first duration at the end of the real-time offline cached learning course segment of the duration of the real-time offline cached learning course, intercept a real-time offline cached learning course segment of the second duration before the reserved first duration, and the second duration is the product of the first duration and a preset constant; analyze the corresponding real-time offline cached learning evaluation value based on the real-time offline cached learning course segment of the second duration, substitute the real-time offline cached learning evaluation value corresponding to the real-time offline cached learning course segment of the second duration into the evaluation duration association model to generate the duration of the corresponding real-time offline cached learning course, and intercept and deliver the real-time offline cached learning course segment corresponding to the real-time offline cached learning evaluation value corresponding to the real-time offline cached learning course segment of the second duration to the target user for learning until the target user finishes learning.

[0020] Compared with the prior art, the beneficial effects of the present invention are as follows: This application simultaneously collects facial biometric data and after-class answering data to establish the correlation between physiological behavior and learning effect; it breaks through the evaluation limitation that only relies on answering data traditionally, provides a more comprehensive analysis of the attention state, captures subtle attention changes through multi-dimensional biometric features, avoids misjudgment by a single indicator, and improves the accuracy and robustness of the system; constructs an attention evaluation model for online learning based on attention features, marks the time points of user control behaviors, segments the learning courses, analyzes and generates a set of historical offline cache learning analyses with durations, providing a data basis for subsequent personalized course tailoring; locates inefficient segments using the active course backward behavior, dynamically calculates the deletion threshold, avoids subjectively setting the threshold, and reflects the real learning pain points; improves the quality of the data set, prevents overfitting, correlates the evaluation value with the learning duration, reduces the real-time calculation load, and meets the real-time response requirements of the teaching platform; dynamically adjusts the length of subsequent course segments based on the initial analysis result of the first duration; realizes a closed loop of analysis-delivery-reoptimization to adapt to attention fluctuations; the sub-course segments detect the evaluation value in real time and automatically pause the content below the threshold; timely prevent ineffective learning and improve learning efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The accompanying drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention, and do not constitute a limitation to the present invention. In the drawings: Figure 1 is a schematic flow chart of the teaching platform management method based on multi-source data analysis of the present invention; Figure 2 is a schematic structural diagram of the teaching platform management system based on multi-source data analysis of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0022] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to 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.

[0023] Please refer to Figure 1 , the present invention provides a technical solution: A teaching platform management method based on multi-source data analysis, the method includes the following steps: S1: Obtain the authorization information of the target user, extract the historical online learning information of the target user based on the authorization information of the target user, and analyze the learning characteristics of the target user based on the historical online learning information of the target user; In S1, the following contents are included: S101: Obtain the authorization information of the target user, and extract the facial biometric data and after-class answering data of the target user during historical online learning based on the authorization information of the target user; the facial biometric data and the after-class answering data correspond one by one; obtain the historical exam scores of the target user for the courses corresponding to the after-class answering data, and record the ratio of the after-class answering scores to the average historical exam scores as the historical online learning evaluation value; the total scores of each after-class answering score and the historical exam scores are the same; S102: Extract the facial key point coordinates for each frame of image based on the facial biometric data; analyze the aspect ratio of the eyes and the variance of the aspect ratio of the eyes based on the eye key point coordinates among the facial key points for each frame of image. If the aspect ratio of the eyes is less than the corresponding threshold, it is marked as closed eyes; analyze the blink frequency according to the closed eye situation; analyze the intensity of the frowning action based on the FACS system, and analyze the mean and variance of the intensity of the frowning action; obtain the head pose angle vector based on the facial biometric data, and analyze the mean absolute angle change rate, where the mean absolute angle change rate is equal to the total offset of the head yaw angle divided by the total number of frames of the image; analyze the covariance between the intensity of the frowning action and the head yaw angle.

[0024] S2: Construct an attention evaluation model based on the learning characteristics of the target user, extract the historical offline cached learning course data of the target user based on the authorization information of the target user, split and analyze the historical offline cached learning course data of the target user, and construct a set of historical offline cached learning analysis pairs; In S2, the following contents are included: S201: Record the mean and variance of the intensity of the frowning action, the blink frequency, the variance of the aspect ratio of the eyes, the mean absolute angle change rate, and the covariance between the intensity of the frowning action and the head yaw angle as attention features; Example 1, in this example, record the mean of the intensity of the frowning action as μ Z , record the variance of the intensity of the frowning action as σ Z , record the blink frequency as f, and record the variance of the aspect ratio of the eyes as σ EAR , record the mean absolute angle change rate as △ θy ; record the covariance between the intensity of the frowning action and the head yaw angle as Cov(Z, θy); Construct an attention evaluation model based on the attention features and the corresponding historical online learning evaluation values: P = λ1×μ Z +λ2×σ Z +λ3×f + λ4×σ EAR +λ5×△ θy +λ6×Cov(Z, θy); Where P represents the historical online learning evaluation value, λ1, λ2, λ3, λ4, λ5, and λ6 represent the weight coefficients of each attention feature; the weight coefficients of each feature in the attention evaluation model are fitted based on the least squares method; S202: extracting the target user's historical offline cached learning course data based on the target user's authorization information; Mark the time points at which target user interaction data appears in historical offline cached learning courses; target user interaction data includes active user interaction and passive user interaction; active user interaction includes active course acceleration, active course deceleration, active course pause, active course start, active course fast forward, and active course rewind; passive user interaction includes automatic course pause caused by offline cached learning evaluation value being less than the corresponding threshold; Example 2: In this example, the time point of the current historical offline cached learning course is 20 minutes and 16 seconds. The target user takes the interactive action of actively rewinding the course, and the time point of the historical offline cached learning course is adjusted to 19 minutes and 16 seconds; 20 minutes and 16 seconds is recorded as the time point of the active course rewind; S203: Segment the historical offline cached learning courses based on each marked time point to generate historical offline cached learning course segments; extract facial biometric data corresponding to the historical offline cached learning course segments, analyze the attention features of the historical offline cached learning course segments based on the facial biometric data corresponding to the historical offline cached learning course segments, substitute the attention features of the historical offline cached learning course segments into the fitted attention evaluation model to generate historical offline cached learning evaluation values for the historical offline cached learning course segments; generate historical offline cached learning analysis pairs based on the historical offline cached learning evaluation values and the duration of the corresponding historical offline cached learning course segments, and construct a historical offline cached learning analysis pair set based on the historical offline cached learning analysis pairs.

[0025] S3: Filter data and remove outliers from the historical offline cache learning analysis pair set to generate a second historical offline cache learning analysis pair set, and construct an evaluation duration association model based on the second historical offline cache learning analysis pair set; In S3, include the following: S301: Extract the time point at which each active course in each historical offline cached learning course is retracted, and the system sets a first preset time window; use the time point at which the active course in the historical offline cached learning course is retracted as the end time point of the trimmed course segment, and cut off the historical offline cached learning course of the first preset time window length; analyze the historical offline cached learning evaluation value of each trimmed course segment, and record the average historical offline cached learning evaluation value of the trimmed course segment as the trimming threshold; Embodiment 3: S302: Extract a set of historical offline cache learning analysis pairs, delete the historical offline cache learning analysis pairs whose historical offline cache learning evaluation values are less than a deletion threshold; arrange the remaining historical offline cache learning analysis pairs in ascending order of historical offline cache learning evaluation values; if any two historical offline cache learning analysis pairs have the same historical offline cache learning evaluation values but different durations of the corresponding historical offline cache learning course segments, retain only the historical offline cache learning analysis pair with the shortest duration of the historical offline cache learning course segment to generate a first set of historical offline cache learning analysis pairs; In this embodiment, the historical offline cache learning analysis pairs include: (LAAV=0.9, Dur=2), (LAAV=3, Dur=2.5), (LAAV=3, Dur=3), (LAAV=4, Dur=4), (LAAV=5, Dur=5), (LAAV=5, Dur=6); Where LAAV is the historical offline cache learning assessment value, and Dur represents the duration of the corresponding historical offline cache learning course segment. If the historical offline cache learning analysis pairs (LAAV=3, Dur=2.5) and (LAAV=3, Dur=3) have the same historical offline cache learning assessment values, then (LAAV=3, Dur=2.5) is retained. If the historical offline cache learning analysis pairs (LAAV=5, Dur=5) and (LAAV=5, Dur=6) have the same historical offline cache learning assessment values, then (LAAV=5, Dur=5) is retained. In this embodiment, the deletion threshold is 1, so the historical offline cache learning analysis pair (LAAV=0.9, Dur=2) does not meet the requirements. After screening, the first historical offline cache learning analysis pair set is {(LAAV=3, Dur=2.5), (LAAV=4, Dur=4), (LAAV=5, Dur=5)}; Performing outlier detection on the first historical offline cache learning analysis pair set and deleting the outliers to generate a second historical offline cache learning analysis pair set; Embodiment 4: S302: Extract a set of historical offline cache learning analysis pairs, delete the historical offline cache learning analysis pairs whose historical offline cache learning evaluation values are less than a deletion threshold; arrange the remaining historical offline cache learning analysis pairs in ascending order of historical offline cache learning evaluation values; if any two historical offline cache learning analysis pairs have the same historical offline cache learning evaluation values but different durations of the corresponding historical offline cache learning course segments, only retain the historical offline cache learning analysis pair with the shortest duration of the historical offline cache learning course segment to generate a first set of historical offline cache learning analysis pairs; perform outlier detection on the first set of historical offline cache learning analysis pairs and delete the outliers to generate a second set of historical offline cache learning analysis pairs; In this embodiment, the first historical offline cache learning analysis pair set is {(LAAV=3, Dur=3), (LAAV=4, Dur=4), (LAAV=5, Dur=11), (LAAV=6, Dur=6)}; Extract all Dur values from the data set and calculate the median to be 5; the absolute differences of the Dur values are 2, 1, 6, and 1 respectively; the median of the absolute deviations is 1.5; Calculate the modified Z-score for each Dur value; the constant is 0.6745; Then for Dur=3: |3-5 / (1.5×0.6745)=2 / 1.01175≈1.976; For Dur=4: |4-5| / (1.5×0.6745)=1 / 1.01175≈0.988; For Dur=6: |6-5| / (1.5×0.6745)=1 / 1.01175≈0.988; For Dur=11: |11-5| / (1.5×0.6745)=6 / 1.01175≈5.930; In this example, the threshold is 3.5; therefore, (LAAV=5, Dur=11) is an abnormal point; The second historical offline cache learning analysis pair set is {(LAAV=3, Dur=3), (LAAV=4, Dur=4), (LAAV=6, Dur=6)}; S303: Based on the second historical offline cache learning analysis, an evaluation duration association model is constructed for the set, y=α1×x+α2; wherein α1 and α2 represent fitting coefficients, x represents the independent variable of the historical offline cache learning evaluation value, and y represents the dependent variable of the brightness value of the non-lighting equipment. The least squares method is used to calculate and solve α1 and α2 in the evaluation duration association model.

[0026] S4: When it is detected that the target user starts to study the real-time offline cache learning course, the real-time offline cache learning course is split and delivered based on the evaluation duration association model.

[0027] In S4, include the following: S401: When it is detected that the target user starts to learn the real-time offline cached learning course, intercept a real-time offline cached learning course segment of the first duration from the real-time offline cached learning course and deliver it to the target user for learning; the first duration is equal to the product of the system average analysis duration and a preset constant, and the system average analysis duration is the average analysis time from obtaining the facial biometric data to generating the offline cached learning course duration; during the learning of the real-time offline cached learning course segment of the first duration, the facial biometric data of the target user is obtained in real time and the corresponding real-time offline cached learning evaluation value is analyzed, and the real-time offline cached learning evaluation value is substituted into the evaluation duration association model to generate the duration of the corresponding real-time offline cached learning course; after the first duration, intercept a real-time offline cached learning course segment of the duration of the real-time offline cached learning course and deliver it to the target user for learning; S402: During the learning of the real-time offline cached learning course segment of the duration of the real-time offline cached learning course, the system splits the duration of the real-time offline cached learning course into real-time offline cached learning sub-course segments based on the first duration, analyzes the real-time offline cached learning evaluation value for the real-time offline cached learning sub-course segments. If the real-time offline cached learning evaluation value is less than the corresponding threshold, the course of the real-time offline cached learning course segment is automatically paused; reserve the first duration at the end of the real-time offline cached learning course segment of the duration of the real-time offline cached learning course, intercept a real-time offline cached learning course segment of the second duration before reserving the first duration, and the second duration is the product of the first duration and a preset constant; analyze the corresponding real-time offline cached learning evaluation value based on the real-time offline cached learning course segment of the second duration, substitute the real-time offline cached learning evaluation value corresponding to the real-time offline cached learning course segment of the second duration into the evaluation duration association model to generate the duration of the corresponding real-time offline cached learning course, and intercept and deliver the real-time offline cached learning course segment corresponding to the real-time offline cached learning evaluation value corresponding to the real-time offline cached learning course segment of the second duration to the target user for learning until the target user finishes learning.

[0028] Please refer to Figure 2 For the technical solution provided by the present invention: In another aspect of the present application, a teaching platform management system based on multi-source data analysis is provided. The system is implemented by applying the above-mentioned teaching platform management method based on multi-source data analysis. The system includes a learning feature analysis module, an attention evaluation analysis module, an evaluation duration association module, and a course delivery management module; The learning feature analysis module is used to obtain the target user authorization information, extract the historical online learning information of the target user based on the target user authorization information, and analyze the learning features of the target user based on the historical online learning information of the target user; The attention evaluation and analysis module is used to construct an attention evaluation model for the learning characteristics of the target user, extract the historical offline cached learning course data of the target user based on the target user's authorization information, split and analyze the historical offline cached learning course data of the target user, and construct a set of historical offline cached learning analysis pairs; The evaluation duration association module is used to filter data and eliminate outliers from the set of historical offline cached learning analysis pairs, generate a second set of historical offline cached learning analysis pairs, and construct an evaluation duration association model based on the second set of historical offline cached learning analysis pairs; The course delivery management module is used to perform split delivery management on the real-time offline cached learning course based on the evaluation duration association model when it is detected that the target user starts to learn the real-time offline cached learning course.

[0029] The learning feature analysis module includes a question answering data analysis unit and a facial biometric data analysis unit; The question answering data analysis unit is used to obtain the target user's authorization information, and extract the after-class question answering data of the target user in the historical online learning based on the target user's authorization information; obtain the historical exam scores of the target user corresponding to the courses of the after-class question answering data, and record the ratio of the after-class question answering scores to the average historical exam scores as the historical online learning evaluation value; the total scores of each after-class question answering score and the historical exam scores are the same; The facial biometric data analysis unit is used to extract the facial key point coordinates for each frame of image based on the facial biometric data; analyze the eye aspect ratio and the variance of the eye aspect ratio for each frame of image based on the eye key point coordinates among the facial key points. If the eye aspect ratio is less than the corresponding threshold, it is marked as closed eyes; analyze the blink frequency according to the closed eye situation; analyze the intensity of the frowning action based on the FACS system, and analyze the mean and variance of the intensity of the frowning action; obtain the head pose angle vector based on the facial biometric data, and analyze the mean absolute angular change rate, where the mean absolute angular change rate is equal to the total offset of the head yaw angle divided by the total number of image frames; analyze the covariance between the intensity of the frowning action and the head yaw angle.

[0030] The attention evaluation and analysis module includes an attention evaluation model construction unit and a learning analysis pair analysis unit; The attention evaluation model construction unit is used to record the mean and variance of the intensity of the frowning action, the blink frequency, the variance of the eye aspect ratio, the average absolute angular change rate, and the covariance of the intensity of the frowning action and the head yaw angle as attention features; construct an attention evaluation model based on the attention features and the corresponding historical online learning evaluation values. In the attention evaluation model, the mean and variance of the intensity of the frowning action, the blink frequency, the variance of the eye aspect ratio, the average absolute angular change rate, and the covariance of the intensity of the frowning action and the head yaw angle are independent variables, and the historical online learning evaluation value is the dependent variable. The weight coefficients of each feature in the attention evaluation model are fitted based on the least squares method; The learning analysis pair analysis unit extracts the historical offline cached learning course data of the target user based on the target user authorization information; marks the time points where the target user interaction data appears in the historical offline cached learning course; segments the historical offline cached learning course based on each marked time point to generate historical offline cached learning course segments; extracts the facial biometric data corresponding to the historical offline cached learning course segments, analyzes the attention features of the historical offline cached learning course segments based on the facial biometric data corresponding to the historical offline cached learning course segments, substitutes the attention features of the historical offline cached learning course segments into the fitted attention evaluation model to generate the historical offline cached learning evaluation values of the historical offline cached learning course segments; generates historical offline cached learning analysis pairs from the historical offline cached learning evaluation values and the corresponding durations of the historical offline cached learning course segments, and constructs a set of historical offline cached learning analysis pairs based on the historical offline cached learning analysis pairs.

[0031] The evaluation duration association module includes a learning analysis pair screening unit and an evaluation duration association model construction unit; The learning analysis pair screening unit extracts the time points of each active course backward in each historical offline cached learning course, and the system's first preset time window; uses the time points of active course backward in the historical offline cached learning course as the end time points of the clipped course segments, and intercepts the historical offline cached learning course with the length of the first preset time window; analyzes the historical offline cached learning evaluation values of each clipped course segment, and records the average historical offline cached learning evaluation value of the clipped course segment as the deletion threshold, Extract the set of historical offline cache learning analysis pairs, and delete the historical offline cache learning analysis pairs in which the historical offline cache learning evaluation value is less than the deletion threshold; arrange the remaining historical offline cache learning analysis pairs in ascending order of the historical offline cache learning evaluation value; if the historical offline cache learning evaluation values of any two historical offline cache learning analysis pairs are the same, but the durations of the corresponding historical offline cache learning course segments are different, only retain the historical offline cache learning analysis pair with the smallest duration of the historical offline cache learning course segment to generate the first set of historical offline cache learning analysis pairs; perform outlier detection on the first set of historical offline cache learning analysis pairs and delete the outliers to generate the second set of historical offline cache learning analysis pairs; The evaluation duration correlation model construction unit is used to construct an evaluation duration correlation model based on the second set of historical offline cache learning analysis pairs.

[0032] The course delivery management module includes a first delivery management unit and a second delivery management unit; The first delivery management unit is used to, when it is detected that the target user starts to learn the real-time offline cache learning course, intercept a real-time offline cache learning course segment of the first duration from the real-time offline cache learning course and deliver it to the target user for learning; the first duration is equal to the product of the system average analysis duration and a preset constant, and the system average analysis duration is the average analysis time from obtaining the facial biometric data to generating the offline cache learning course duration; during the learning of the real-time offline cache learning course segment of the first duration by the target user, the facial biometric data of the target user is obtained in real time and the corresponding real-time offline cache learning evaluation value is analyzed, and the real-time offline cache learning evaluation value is substituted into the evaluation duration correlation model to generate the duration of the corresponding real-time offline cache learning course; after the first duration, intercept a real-time offline cache learning course segment with the duration of the real-time offline cache learning course and deliver it to the target user for learning; The second delivery management unit is used in the real-time and offline caching learning of the duration of the learning course. In the learning of the real-time and offline caching learning course segment, the system splits the duration of the real-time and offline caching learning course based on the first duration to generate real-time and offline caching learning sub-course segments, analyzes the real-time and offline caching learning evaluation value for the real-time and offline caching learning sub-course segments. If the real-time and offline caching learning evaluation value is less than the corresponding threshold, the system automatically pauses the real-time and offline caching learning course segment; reserves the first duration at the end of the real-time and offline caching learning course segment of the duration of the real-time and offline caching learning course, intercepts the real-time and offline caching learning course segment of the second duration before reserving the first duration, and the second duration is the product of the first duration and a preset constant; analyzes the corresponding real-time and offline caching learning evaluation value based on the real-time and offline caching learning course segment of the second duration, substitutes the real-time and offline caching learning evaluation value corresponding to the real-time and offline caching learning course segment of the second duration into the evaluation duration association model to generate the duration of the corresponding real-time and offline caching learning course, and intercepts the duration of the real-time and offline caching learning course corresponding to the real-time and offline caching learning evaluation value corresponding to the real-time and offline caching learning course segment of the second duration and delivers it to the target user for learning until the target user finishes learning.

[0033] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device.

[0034] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and can be implemented in other specific forms without departing from the spirit or basic characteristics of the present invention. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, it is intended to include all changes falling within the meaning and scope of the equivalent elements of the claims in the present invention. Any reference signs in the claims should not be regarded as limiting the claimed rights.

Claims

1. A teaching platform management method based on multi-source data analysis, characterized in that The method includes the following steps: S1: Obtain the authorization information of the target user, extract the historical online learning information of the target user based on the authorization information of the target user, and analyze the learning characteristics of the target user based on the historical online learning information of the target user; S2: Construct an attention evaluation model based on the learning characteristics of the target user, extract the historical offline cached learning course data of the target user based on the authorization information of the target user, split and analyze the historical offline cached learning course data of the target user, and construct a set of historical offline cached learning analysis pairs; S3: Perform data screening and outlier removal on the set of historical offline cached learning analysis pairs to generate a second set of historical offline cached learning analysis pairs, and construct an evaluation duration association model based on the second set of historical offline cached learning analysis pairs; S4: When it is detected that the target user starts to learn the real-time offline cached learning course, perform split placement management on the real-time offline cached learning course based on the evaluation duration association model.

2. The teaching platform management method based on multi-source data analysis according to claim 1, wherein: In S1, it includes the following content: S101: Obtain the authorization information of the target user, and extract the facial biometric data and after-class answering data of the target user during historical online learning based on the authorization information of the target user; the facial biometric data and the after-class answering data correspond one by one; obtain the historical exam scores of the target user for the courses corresponding to the after-class answering data, and record the ratio of the after-class answering scores to the average historical exam scores as the historical online learning evaluation value; S102: Extract the facial key point coordinates for each frame of the image based on the facial biometric data; analyze the aspect ratio of the eyes and the variance of the aspect ratio of the eyes based on the eye key point coordinates in the facial key points for each frame of the image. If the aspect ratio of the eyes is less than the corresponding threshold, it is marked as closed eyes; analyze the blink frequency according to the closed-eye situation; analyze the intensity of the frowning action based on the FACS system, and analyze the mean and variance of the intensity of the frowning action; obtain the head pose angle vector based on the facial biometric data, and analyze the average absolute angle change rate, where the average absolute angle change rate is equal to the total offset of the head yaw angle divided by the total number of image frames; analyze the covariance between the intensity of the frowning action and the head yaw angle.

3. The teaching platform management method based on multi-source data analysis according to claim 2, wherein: In S2, it includes the following content: S201: Record the mean and variance of the intensity of the frowning action, the blink frequency, the variance of the aspect ratio of the eyes, the average absolute angle change rate, and the covariance between the intensity of the frowning action and the head yaw angle as attention features; Construct an attention evaluation model based on the attention features and the corresponding historical online learning evaluation value. In the attention evaluation model, the mean and variance of the intensity of the frowning action, the blink frequency, the variance of the aspect ratio of the eyes, the average absolute angle change rate, and the covariance between the intensity of the frowning action and the head yaw angle are independent variables, and the historical online learning evaluation value is the dependent variable. Fit the weight coefficients of each feature in the attention evaluation model based on the least squares method; S202: Extract the historical offline cached learning course data of the target user based on the authorization information of the target user; Mark the time points when target user interaction data appears in the historical offline cached learning courses; The target user interaction data includes user active interaction and user passive interaction; S203: Segment the historical offline cached learning courses based on the time points of each marker to generate historical offline cached learning course segments; extract the facial biometric data corresponding to the historical offline cached learning course segments, analyze the attention features of the historical offline cached learning course segments based on the facial biometric data corresponding to the historical offline cached learning course segments, substitute the attention features of the historical offline cached learning course segments into the fitted attention evaluation model to generate the historical offline cached learning evaluation values of the historical offline cached learning course segments; generate historical offline cached learning analysis pairs by combining the historical offline cached learning evaluation values and the durations of the corresponding historical offline cached learning course segments, and construct a set of historical offline cached learning analysis pairs based on the historical offline cached learning analysis pairs.

4. The teaching platform management method based on multi-source data analysis according to claim 3, characterized in that: In S3, the following contents are included: S301: Extract the time points of each active course backward in each historical offline cached learning course, and the first preset time window of the system; Use the time points of the active course backward in the historical offline cached learning course as the end time points for cropping the course segments, and intercept the historical offline cached learning courses with the length of the first preset time window; analyze the historical offline cached learning evaluation values for each cropped course segment, and record the average historical offline cached learning evaluation value of the cropped course segment as the deletion threshold; S302: Extract the set of historical offline cached learning analysis pairs, and delete the historical offline cached learning analysis pairs with historical offline cached learning evaluation values less than the deletion threshold; sort the remaining historical offline cached learning analysis pairs in ascending order of the historical offline cached learning evaluation values; if the historical offline cached learning evaluation values of any two historical offline cached learning analysis pairs are the same but the durations of the corresponding historical offline cached learning course segments are different, only retain the historical offline cached learning analysis pair with the minimum duration of the historical offline cached learning course segment to generate the first set of historical offline cached learning analysis pairs; perform outlier detection on the first set of historical offline cached learning analysis pairs and delete the outliers to generate the second set of historical offline cached learning analysis pairs; S303: Construct an evaluation duration correlation model based on the second set of historical offline cached learning analysis pairs, y = α1×x + α2; where α1 and α2 represent the fitting coefficients, x represents the independent variable of the historical offline cached learning evaluation value, y represents the dependent variable of the non-illumination device light value, and use the least squares method to calculate and solve α1 and α2 in the evaluation duration correlation model.

5. The teaching platform management method based on multi-source data analysis according to claim 4, characterized in that: In S4, the following contents are included: S401: When it is detected that the target user starts to learn the real-time offline cached learning course, intercept the real-time offline cached learning course segment with the first duration from the real-time offline cached learning course and send it to the target user for learning; During the learning of the real-time offline cached learning course segment within the first duration, the facial biometric data of the target user is obtained in real time and the corresponding real-time offline cached learning evaluation value is analyzed. The real-time offline cached learning evaluation value is substituted into the evaluation duration association model to generate the duration of the corresponding real-time offline cached learning course. After the first duration, the real-time offline cached learning course segment with the duration of the real-time offline cached learning course is intercepted and sent to the target user for learning. S402: During the learning of the real-time offline cached learning course segment with the duration of the real-time offline cached learning course, the system splits the duration of the real-time offline cached learning course based on the first duration to generate real-time offline cached learning sub-course segments, analyzes the real-time offline cached learning evaluation value for the real-time offline cached learning sub-course segments. If the real-time offline cached learning evaluation value is less than the corresponding threshold, the real-time offline cached learning course segment is automatically paused. A first duration is reserved at the end of the real-time offline cached learning course segment with the duration of the real-time offline cached learning course. Before reserving the first duration, a real-time offline cached learning course segment with a second duration is intercepted. The corresponding real-time offline cached learning evaluation value is analyzed based on the real-time offline cached learning course segment with the second duration. The real-time offline cached learning evaluation value corresponding to the real-time offline cached learning course segment with the second duration is substituted into the evaluation duration association model to generate the duration of the corresponding real-time offline cached learning course. The real-time offline cached learning course segment corresponding to the real-time offline cached learning evaluation value corresponding to the real-time offline cached learning course segment with the second duration is intercepted and sent to the target user for learning until the target user finishes learning.

6. A teaching platform management system based on multi-source data analysis, the system is applied to implement the teaching platform management method based on multi-source data analysis described in any one of claims 1-5, characterized in that, The system includes a learning feature analysis module, an attention evaluation analysis module, an evaluation duration association module, and a course delivery management module. The learning feature analysis module is used to obtain the target user authorization information, extract the historical online learning information of the target user based on the target user authorization information, and analyze the learning features of the target user based on the historical online learning information of the target user. The attention evaluation analysis module is used to construct an attention evaluation model based on the learning features of the target user, extract the historical offline cached learning course data of the target user based on the target user authorization information, split and analyze the historical offline cached learning course data of the target user, and construct a historical offline cached learning analysis pair set. The evaluation duration association module is used to perform data screening and outlier elimination on the historical offline cached learning analysis pair set to generate a second historical offline cached learning analysis pair set, and construct an evaluation duration association model based on the second historical offline cached learning analysis pair set. The course delivery management module is used to, when detecting that the target user starts to learn the real-time offline cached learning course, perform split delivery management on the real-time offline cached learning course based on the evaluation duration association model.

7. The teaching platform management system based on multi-source data analysis according to claim 6, characterized in that: The learning feature analysis module includes a question answering data analysis unit and a facial biometric data analysis unit. The question answering data analysis unit is used to obtain the target user authorization information and extract the after-class question answering data of the target user in historical online learning based on the target user authorization information. Obtain the historical exam scores of the target user for the courses corresponding to the after-class question answering data, and record the ratio of the after-class question answering scores to the average historical exam scores as the historical online learning evaluation value; the total scores of each after-class question answering score and the historical exam scores are the same; The facial biometric data analysis unit is used to extract the facial key point coordinates for each frame of image based on the facial biometric data; analyze the eye aspect ratio and the variance of the eye aspect ratio based on the eye key point coordinates in the facial key points for each frame of image. If the eye aspect ratio is less than the corresponding threshold, it is marked as closed eyes; analyze the blink frequency according to the closed eye situation; analyze the intensity of the frowning action based on the FACS system, and analyze the mean and variance of the intensity of the frowning action; obtain the head pose angle vector based on the facial biometric data, and analyze the average absolute angle change rate, where the average absolute angle change rate is equal to the total yaw angle offset of the head divided by the total number of frames of the image; analyze the covariance between the intensity of the frowning action and the head yaw angle.

8. The teaching platform management system based on multi-source data analysis according to claim 6, characterized in that: The attention evaluation and analysis module includes an attention evaluation model construction unit and a learning analysis pair analysis unit; The attention evaluation model construction unit is used to record the mean and variance of the intensity of the frowning action, the blink frequency, the variance of the eye aspect ratio, the average absolute angle change rate, and the covariance between the intensity of the frowning action and the head yaw angle as attention features; Construct an attention evaluation model based on the attention features and the corresponding historical online learning evaluation value. In the attention evaluation model, the mean and variance of the intensity of the frowning action, the blink frequency, the variance of the eye aspect ratio, the average absolute angle change rate, and the covariance between the intensity of the frowning action and the head yaw angle are independent variables, and the historical online learning evaluation value is the dependent variable. Fit the weight coefficients of each feature in the attention evaluation model based on the least squares method; The learning analysis pair analysis unit extracts the historical offline cached learning course data of the target user based on the target user authorization information; marks the time points where the target user interaction data appears in the historical offline cached learning courses; segments the historical offline cached learning courses based on each marked time point to generate historical offline cached learning course segments; extracts the facial biometric data corresponding to the historical offline cached learning course segments, analyzes the attention features of the historical offline cached learning course segments based on the facial biometric data corresponding to the historical offline cached learning course segments, substitutes the attention features of the historical offline cached learning course segments into the fitted attention evaluation model to generate the historical offline cached learning evaluation values of the historical offline cached learning course segments; generates historical offline cached learning analysis pairs from the historical offline cached learning evaluation values and the durations of the corresponding historical offline cached learning course segments, and constructs a set of historical offline cached learning analysis pairs based on the historical offline cached learning analysis pairs.

9. The teaching platform management system based on multi-source data analysis according to claim 6, characterized in that: The evaluation duration association module includes a learning analysis pair screening unit and an evaluation duration association model construction unit; The learning analysis extracts the time points of each active course backward in each historical offline cached learning course in the screening unit, and the first preset time window of the system; uses the time points of the active course backward in the historical offline cached learning course as the end time points of the cropped course segments, and intercepts the historical offline cached learning courses with the length of the first preset time window; analyzes the historical offline cached learning evaluation values for each cropped course segment, and records the average historical offline cached learning evaluation value of the cropped course segment as the deletion threshold. Extracts the set of historical offline cached learning analysis pairs, and deletes the historical offline cached learning analysis pairs with historical offline cached learning evaluation values less than the deletion threshold in it; arranges the remaining historical offline cached learning analysis pairs in ascending order of the historical offline cached learning evaluation values; if the historical offline cached learning evaluation values of any two historical offline cached learning analysis pairs are the same, but the durations of the corresponding historical offline cached learning course segments are different, only retains the historical offline cached learning analysis pair with the smallest duration of the historical offline cached learning course segment to generate the first set of historical offline cached learning analysis pairs; performs outlier detection on the first set of historical offline cached learning analysis pairs and deletes the outliers to generate the second set of historical offline cached learning analysis pairs. The evaluation duration association model construction unit is used to construct an evaluation duration association model based on the second set of historical offline cached learning analysis pairs.

10. The teaching platform management system based on multi-source data analysis according to claim 6, characterized in that: The course delivery management module includes a first delivery management unit and a second delivery management unit. The first delivery management unit is used to, when detecting that a target user starts to learn a real-time offline cached learning course, intercept a real-time offline cached learning course segment with a first duration from the real-time offline cached learning course and deliver it to the target user for learning. The first duration is equal to the product of the system average analysis duration and a preset constant, and the system average analysis duration is the average analysis time from obtaining the facial biometric data to generating the offline cached learning course duration; during the learning of the real-time offline cached learning course segment with the first duration by the target user, the facial biometric data of the target user is obtained in real time and the corresponding real-time offline cached learning evaluation value is analyzed, and the real-time offline cached learning evaluation value is substituted into the evaluation duration association model to generate the duration of the corresponding real-time offline cached learning course; after the first duration, intercept a real-time offline cached learning course segment with the duration of the real-time offline cached learning course and deliver it to the target user for learning. The second delivery management unit is used for the real-time offline cached learning course segment learning of the duration of the real-time offline cached learning course. The system splits the duration of the real-time offline cached learning course based on the first duration to generate real-time offline cached learning sub-course segments, analyzes the real-time offline cached learning evaluation value for the real-time offline cached learning sub-course segments. If the real-time offline cached learning evaluation value is less than the corresponding threshold, the real-time offline cached learning course segment is automatically paused; a first duration is reserved at the end of the real-time offline cached learning course segment of the duration of the real-time offline cached learning course. Before reserving the first duration, a real-time offline cached learning course segment of a second duration is intercepted, and the second duration is the product of the first duration and a preset constant; the corresponding real-time offline cached learning evaluation value is analyzed based on the real-time offline cached learning course segment of the second duration, and the real-time offline cached learning evaluation value corresponding to the real-time offline cached learning course segment of the second duration is substituted into the evaluation duration association model to generate the duration of the corresponding real-time offline cached learning course. The duration of the real-time offline cached learning course corresponding to the real-time offline cached learning evaluation value corresponding to the real-time offline cached learning course segment of the second duration is delivered to the target user for learning until the target user finishes learning.

Images