Learner behavior portrait construction system and method based on data analysis

Abstract

The invention relates to the technical field of behavior portrait construction, and discloses a learner behavior portrait construction system and method based on data analysis, and the system comprises a learning behavior collection module, a learning rhythm mutation detection module, a behavior track stability evaluation module, a knowledge point backtracking index calculation module, and a portrait dynamic adjustment module. According to the method, the learning behavior fluctuation rate is calculated, the knowledge point jumping frequency and the task completion interval fluctuation condition are combined, the sudden change characteristics of the learning behaviors are judged, the recognition capacity for sudden learning behavior changes is enhanced, the behavior path overlapping degree of different time periods is compared, and the page switching frequency and the dwell time change range are calculated; the method comprises the steps of analyzing the stability of learning behaviors, performing hierarchical analysis on a knowledge point access mode of a learner according to a backtracking condition of a learning path, and optimizing the sensitivity of a portrait to learning path adjustment and enhancing the portrait adaptability by dynamically adjusting a portrait updating strategy and self-adapting to real-time behavior changes of the learner.

Description

Technical Field

[0001] The present invention relates to the technical field of behavior portrait construction, and particularly to a system and method for constructing a learner behavior portrait based on data analysis. Background Art

[0002] The technical field of behavior portrait construction includes collecting, analyzing, and modeling the behavior data of individuals or groups to form a visual and structured user feature description. In this technical field, data sources include interaction records, operation habits, behavior trajectories, etc. of users in a specific environment. By using data analysis techniques to extract key features, and combining methods such as pattern recognition, feature matching, and statistical analysis, models are built for users' preferences, habits, and potential needs, which are widely applied in multiple fields such as education, medical care, and e-commerce to support personalized services and intelligent recommendations.

[0003] Among them, a learner behavior portrait construction system based on data analysis refers to collecting, processing, and modeling the data generated by learners during the learning process to form an individualized learner behavior portrait, mainly covering technical matters such as data collection, data preprocessing, behavior feature extraction, and portrait construction. First, data such as learners' learning records, knowledge point mastery, and learning paths are obtained through a learning management system or an online learning platform. Secondly, feature engineering is used to normalize and screen features of learners' operation sequences, stay times, learning frequencies, etc., and core behavior features are extracted. Subsequently, classification models and clustering methods are used to model learners' learning habits, learning styles, knowledge mastery, etc., and a personalized portrait is constructed. Finally, the learner portrait is stored in a database in combination with data storage technology to support subsequent personalized recommendations and teaching interventions.

[0004] In the process of constructing a learner portrait by existing online learning platforms, the continuous changes in the learning path are not effectively captured, and updates rely on fixed-period or single-dimensional statistical data, making it difficult to reflect the changing characteristics of learners' short-term behaviors. At the same time, sudden changes in learning behaviors are not effectively identified, resulting in the portrait still being based on past data when there are short-term fluctuations in the learning rhythm, making it difficult to accurately reflect the changes in the learning state and affecting the real-time nature of the portrait. The stability assessment of behavior trajectories mainly relies on static statistical analysis of page access records and operation habits, and no cross-time behavior consistency analysis is established, making it difficult to accurately distinguish persistent patterns from short-term abnormal behaviors during the learning process. The assessment of knowledge point mastery mainly relies on learning progress or knowledge test results, without considering the retrospective characteristics of the learning path and ignoring learners' repeated access behaviors to some knowledge points, affecting the ability to identify understanding deviations of knowledge points. The portrait update mechanism lacks the ability of dynamic adjustment and does not combine the fluctuations in the learning rhythm, resulting in the portrait adjustment frequency being difficult to adapt to the changes in individual learning patterns and affecting the long-term stability and adaptability of the portrait. Summary of the Invention

[0005] The object of the present invention is to solve the deficiencies existing in the prior art, and to propose a system and method for constructing a learner behavior portrait based on data analysis.

[0006] To achieve the above object, the present invention adopts the following technical solutions: A system for constructing a learner behavior portrait based on data analysis includes:

[0007] The learning behavior acquisition module, based on an online learning platform, combines the course learning log and the interaction behavior log to record the learner's learning path sequence and store it, generating a learning behavior data set;

[0008] The learning rhythm mutation detection module calculates the learning behavior volatility in adjacent time periods according to the learning behavior data set, judges the learning behavior mutation situation, analyzes the change in the jump frequency between knowledge points, detects the time interval fluctuation situation, screens the behavior patterns with rhythm mutations, and obtains the learning rhythm mutation status record;

[0009] The behavior trajectory stability evaluation module calculates the behavior path overlap degree in different time periods according to the learning behavior data set, analyzes the stability of the learner's behavior pattern, and obtains the behavior trajectory stability status record;

[0010] The knowledge point backtracking index calculation module calculates the learner's same knowledge point backtracking frequency and learning path deviation rate according to the learning behavior data set, and obtains the knowledge point backtracking status record;

[0011] The portrait dynamic adjustment module combines the learning rhythm mutation status record, the behavior trajectory stability status record and the knowledge point backtracking status record, adjusts the portrait update trigger condition and judges the portrait update requirement, and generates a learner portrait construction plan.

[0012] As a further solution of the present invention, the learning behavior data set includes a learning behavior sequence, a knowledge point switching record, a task completion timestamp, a mouse trajectory coordinate sequence, a page stay duration, a learning path sequence; the learning rhythm mutation status record includes a learning behavior volatility, a change value of the knowledge point jump frequency, a task completion interval fluctuation value, a rhythm mutation behavior pattern identifier; the behavior trajectory stability status record includes a behavior path overlap degree, a change value of the page switching frequency, a change range of the pause time, a learning behavior pattern stability identifier; the knowledge point backtracking status record includes a knowledge point backtracking frequency, a learning path deviation rate, a knowledge point repeated access times, a jump path backtracking degree; the learner portrait construction plan includes a portrait update trigger condition, a portrait update requirement determination value, a learning path deviation adjustment strategy, a knowledge point backtracking adjustment strategy, a learner portrait adjustment plan.

[0013] As a further solution of the present invention, the learning behavior acquisition module includes a knowledge point switching record sub-module, an interaction behavior data processing sub-module, and a learning path storage sub-module;

[0014] The knowledge point switching record sub-module collects the knowledge point switching information in the course learning log, obtains the knowledge point access records of each learner at different time points, calculates the time interval and switching frequency of adjacent knowledge point switches, aggregates the data of all learners, and generates knowledge point switching feature data;

[0015] Based on the interaction behavior log, the interaction behavior data processing sub-module extracts the mouse trajectory coordinate sequence, calculates the average speed, acceleration, and the number of stop points of the mouse movement, and counts the page stay duration. Using the formula:

[0016]

[0017] Calculate the interaction behavior eigenvalue S to obtain the interaction behavior feature data, where x i and y i represent the coordinate values of the mouse at the i-th time point, t i represents the timestamp of this time point, and N represents the total number of operations;

[0018] The learning path storage sub-module serializes and stores the learning paths of learners according to the knowledge point switching feature data and the interaction behavior feature data according to the set time window, records the knowledge point learning order and interaction behavior patterns of learners in different time periods, and obtains the learning behavior data set.

[0019] As a further solution of the present invention, the learning rhythm mutation detection module includes a learning behavior volatility calculation sub-module, a knowledge point jump frequency analysis sub-module, and a rhythm mutation behavior screening sub-module;

[0020] Based on the learning behavior data set, the learning behavior volatility calculation sub-module extracts the behavior data of learners in adjacent time periods, calculates the learning activity frequency in each time period, using the formula:

[0021]

[0022] Calculate the learning behavior volatility W, analyze it in combination with the time series, and generate the learning behavior volatility data, where B i represents the number of learning behavior activities in the i-th time period, B i+1 represents the number of learning behavior activities in the i + 1-th time period, t i represents the timestamp of the corresponding time point i, t i+1 represents the timestamp of the corresponding time point i + 1, and w is the number of calculation samples;

[0023] The knowledge point jump frequency analysis sub-module analyzes the switching of knowledge points of the learner in each time period according to the learning behavior volatility data, counts the number of knowledge point jumps of the learner in consecutive time periods, calculates the knowledge point switching frequency, judges the change trend, and obtains the knowledge point jump frequency data;

[0024] The rhythm mutation behavior screening sub-module detects the time interval fluctuation of the learner based on the knowledge point jump frequency data, calculates the change rate of the time interval between adjacent time periods, compares it with the set fluctuation threshold, screens the learner behavior patterns with rhythm mutations, and obtains the learning rhythm mutation status record.

[0025] As a further solution of the present invention, the behavior trajectory stability evaluation module includes a mouse trajectory comparison sub-module, a page switching frequency calculation sub-module, and a pause time fluctuation analysis sub-module;

[0026] The mouse trajectory comparison sub-module obtains the mouse movement trajectory data in the learning behavior dataset, and according to the mouse trajectory coordinate point sequence in adjacent time periods, uses the formula:

[0027]

[0028] Calculate the mouse trajectory offset value S t , and obtain the mouse trajectory overlap degree analysis result. Among them, X i,t , Y i,t respectively represent the coordinates of the i-th trajectory point at time t, X i,t+1 , Y i,t+1 represent the corresponding coordinates at time t + 1, and G represents the total number of trajectory points;

[0029] The page switching frequency calculation sub-module collects the page access records in adjacent time periods based on the learning behavior dataset, calculates the number of page switches, and counts the change rate of page switches per unit time to obtain the page switching frequency value;

[0030] The pause time fluctuation analysis sub-module screens the pause time sequence of the same knowledge point according to the page switching frequency value and the mouse trajectory overlap degree analysis result, calculates the difference between the maximum value and the minimum value, determines the pause time change range, judges the pause time fluctuation of the learner at the same knowledge point, and obtains the behavior trajectory stability status record.

[0031] As a further solution of the present invention, the knowledge point backtracking index calculation module includes a learning behavior record sub-module, a knowledge point backtracking calculation sub-module, and a learning path backtracking analysis sub-module;

[0032] The learning behavior recording sub-module obtains the access data of the learner on the same knowledge points according to the learning behavior data set, records the access times, access order, and jump path, detects the behavior data of the learner in each learning stage, counts the repeated access records of the same knowledge points and the jump trajectory of the learning path, and obtains the knowledge point access trajectory data;

[0033] Based on the knowledge point access trajectory data, the knowledge point backtracking calculation sub-module counts the access order in different learning stages, analyzes the repeatability in the learning path, quantifies the knowledge point backtracking degree in combination with the access data, and uses the formula:

[0034]

[0035] Calculate the backtracking frequency R of the learner on the same knowledge point f , and obtain the backtracking frequency data, where q i represents the number of times of accessing the knowledge point for the i-th time, W i represents the importance weight of this access, P i represents the path deviation degree of the i-th time in the learning path, and Z represents the total number of times the learner accesses this knowledge point;

[0036] Based on the backtracking frequency data, the learning path backtracking analysis sub-module analyzes the backtracking degree of the learning path during the knowledge point backtracking process of the learner, calculates the intensity of the learning path backtracking, compares the backtracking behavior patterns of different learners, combines the path backtracking frequency and backtracking level, summarizes the understanding deviation of the learner on different knowledge points, and obtains the knowledge point backtracking status record.

[0037] As a further solution of the present invention, the portrait dynamic adjustment module includes a portrait update trigger sub-module, a portrait adjustment calculation sub-module, and a portrait construction scheme generation sub-module;

[0038] Based on the learning rhythm mutation status record, the behavior trajectory stability status record, and the knowledge point backtracking status record, the portrait update trigger sub-module analyzes the change trend of the learner's behavior, judges whether the learning state has mutated, compares the change rate of the learning rhythm before and after, combines the behavior trajectory stability evaluation value and the knowledge point backtracking frequency, calculates the trigger condition for portrait update, and obtains the portrait update requirement record;

[0039] Based on the portrait update requirement record, the portrait adjustment calculation sub-module combines the knowledge point backtracking frequency and the learning path deviation rate, combines the learning rhythm mutation state and the behavior trajectory stability, quantifies the portrait update amplitude, and uses the formula:

[0040]

[0041] Calculate the portrait dynamic adjustment parameter U p, where K i represents the knowledge point backtracking frequency at the i-th time, M i represents the importance weight of this knowledge point, D i represents the learning path deviation rate at the i-th time, L i represents the learning path recommendation value, I i represents the learning rhythm mutation state value, o i represents the behavioral trajectory stability coefficient, and zn represents the total number of learning stages of the learner;

[0042] The portrait construction scheme generation sub-module is based on the portrait adjustment parameter record, compares the portrait adjustment amplitudes of different learners, calculates the portrait change trend, analyzes the stability of the learner portrait, adjusts the portrait weight distribution, summarizes the portrait update strategy, and obtains the learner portrait construction scheme.

[0043] A method for constructing a learner behavior portrait based on data analysis includes the following steps:

[0044] S1: Collect the course learning logs of the online learning platform, record the knowledge point switching time points and task completion timestamps, obtain the interaction behavior logs, extract the mouse trajectory coordinate sequence and page stay duration, and segment and store the learning path according to the time window to generate a learning behavior data set;

[0045] S2: Call the learning behavior data set, calculate the learning behavior volatility within adjacent time periods, judge the change of the jump frequency between knowledge points, obtain the time interval fluctuation situation, compare it with the set fluctuation threshold, screen the behavior patterns with fluctuating learning rhythm, and obtain the learning rhythm fluctuation state record;

[0046] S3: Call the learning behavior data set, calculate the overlap degree of the mouse movement trajectories, the difference in page switching frequencies, and the change range of the pause time within adjacent time periods, compare the fluctuation of the pause time for the same knowledge point, and obtain the behavioral trajectory stability state record;

[0047] S4: Call the learning behavior data set, count the number of repeated visits of the learner to the same knowledge point, record the knowledge point jump path, calculate the knowledge point backtracking frequency, obtain the deviation degree of the knowledge point backtracking in the learning path, and obtain the knowledge point backtracking state record;

[0048] S5: According to the learning rhythm fluctuation state record, the behavioral trajectory stability state record and the knowledge point backtracking state record, judge the portrait update requirement, adjust the learner portrait update trigger condition, optimize the learner behavior characteristics in combination with the learning path deviation rate, and generate a learner portrait construction scheme.

[0049] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0050] In the present invention, by calculating the learning behavior volatility, combining the knowledge point jump frequency and the task completion interval fluctuation, the mutation characteristics of the learning behavior are judged, the recognition ability of the sudden learning behavior change is enhanced, the overlap degree of the behavior paths in different time periods is compared, the page switching frequency and the change range of the pause time are calculated, the stability of the learning behavior is analyzed, according to the backtracking situation of the learning path, the knowledge point backtracking frequency and the learning path deviation rate are calculated, the knowledge point access mode of the learner is hierarchically analyzed to reflect potential understanding deviations, and by dynamically adjusting the portrait update strategy, the real-time behavior changes of the learner are adapted, the sensitivity of the portrait to the learning path adjustment is optimized, and the portrait adaptability is enhanced. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 is the system flow chart of the present invention;

[0052] Figure 2 is the flow chart of the learning behavior acquisition module of the present invention;

[0053] Figure 3 is the flow chart of the learning rhythm mutation detection module of the present invention;

[0054] Figure 4 is the flow chart of the behavior trajectory stability evaluation module of the present invention;

[0055] Figure 5 is the flow chart of the knowledge point backtracking index calculation module of the present invention;

[0056] Figure 6 is the flow chart of the portrait dynamic adjustment module of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0057] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0058] In the description of the present invention, it should be understood that the orientation or positional relationship indicated by the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. is based on the orientation or positional relationship shown in the accompanying drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention. In addition, in the description of the present invention, "a plurality of" means two or more unless otherwise specifically defined.

[0059] Please refer to Figure 1, a learner behavior portrait construction system based on data analysis includes:

[0060] The learning behavior acquisition module, based on the online learning platform, collects the knowledge point switching records and task completion timestamps according to the course learning logs, collects the mouse trajectory coordinate sequences and page stay durations according to the interaction behavior logs, records the learner learning path sequences and stores them according to time windows, and generates a learning behavior dataset;

[0061] The learning rhythm mutation detection module calculates the learning behavior volatility of adjacent time periods according to the learning behavior dataset, judges the learning behavior mutation situation, analyzes the change of the jump frequency between knowledge points, detects the time interval fluctuation situation, and combines the set volatility threshold to screen the learner behavior patterns with rhythm mutations and obtains the learning rhythm mutation status record;

[0062] The behavior trajectory stability evaluation module compares the mouse movement trajectories, page switching frequencies and pause times within adjacent time periods according to the learning behavior dataset, calculates the behavior path overlap degree of the difference time periods, analyzes the stability of the learner behavior patterns, calculates the change range of the pause time, and judges the pause time fluctuation situation of the learner at the same knowledge point to obtain the behavior trajectory stability status record;

[0063] The knowledge point backtracking index calculation module records the repeated access times and jump path backtracking situations of the learner at the same knowledge point according to the learning behavior dataset, calculates the knowledge point backtracking frequency, analyzes the understanding deviation situation of the learner for the knowledge points, calculates the learning path deviation rate, and compares the backtracking degree of the learning path sequences to obtain the knowledge point backtracking status record;

[0064] The portrait dynamic adjustment module combines the learning rhythm mutation status record, the behavior trajectory stability status record and the knowledge point backtracking status record, adjusts the portrait update trigger conditions, judges the portrait update requirements, and dynamically adjusts the learner portrait according to the knowledge point backtracking frequency and the learning path deviation rate to generate a learner portrait construction plan.

[0065] The learning behavior dataset includes learning behavior sequences, knowledge point switching records, task completion timestamps, mouse trajectory coordinate sequences, page stay durations, and learning path sequences; the learning rhythm mutation status record includes learning behavior volatility, knowledge point jump frequency change values, task completion interval fluctuation values, and rhythm mutation behavior pattern identifiers; the behavior trajectory stability status record includes behavior path overlap degree, page switching frequency change values, pause time change ranges, and learning behavior pattern stability identifiers; the knowledge point backtracking status record includes knowledge point backtracking frequency, learning path deviation rate, knowledge point repeated access times, and jump path backtracking degree; the learner portrait construction plan includes portrait update trigger conditions, portrait update requirement determination values, learning path deviation adjustment strategies, knowledge point backtracking adjustment strategies, and learner portrait adjustment plans.

[0066] Please refer to Figure 2 , the learning behavior acquisition module includes a knowledge point switching record sub-module, an interaction behavior data processing sub-module, and a learning path storage sub-module;

[0067] The knowledge point switching record sub-module collects the knowledge point switching information in the course learning log, obtains the knowledge point access records of each learner at different time points, calculates the time interval and switching frequency of adjacent knowledge point switches, and aggregates the data of all learners to generate knowledge point switching feature data;

[0068] First, extract the knowledge point switching situation of the learner. Extract the timestamps and knowledge point identifiers from the course learning log of the online learning platform and arrange them in chronological order to obtain the knowledge point access records of each learner. If the learner accesses different knowledge points at adjacent time points, it is recorded as a knowledge point switch. Taking the data of a certain learner as an example, his learning record is as follows:

[0069] 2024-02-12 10:05:00 - Knowledge point A

[0070] 2024-02-12 10:15:00 - Knowledge point B

[0071] 2024-02-12 10:25:00 - Knowledge point B

[0072] 2024-02-12 10:40:00 - Knowledge point C

[0073] Based on the above data, it can be obtained that the knowledge point switches occur at 10:05 - 10:15 (A→B) and 10:25 - 10:40 (B→C). Next, calculate the switching time interval, obtain the time interval of adjacent knowledge point switches, and count the knowledge point switching frequency of the learner. Taking the data of 5 learners as an example, their knowledge point switching frequencies are as follows:

[0074] Table 1.1 Knowledge Point Switching Statistical Table

[0075]

[0076] As shown in Table 1.1, the average switching interval calculation method for each learner is as follows:

[0077]

[0078] For example, the calculation process of U1 is as follows:

[0079] Finally, aggregate the data of all learners to form knowledge point switching feature data.

[0080] Based on the interaction behavior log, the interaction behavior data processing sub-module extracts the mouse trajectory coordinate sequence, calculates the average speed, acceleration and the number of stop points of the mouse movement, and statistically analyzes the page stay duration. Using the formula:

[0081]

[0082] Calculate the interaction behavior eigenvalue S to obtain the interaction behavior characteristic data, where x i and y i represent the coordinate values of the mouse at the i-th time point, t i represents the timestamp of this time point, and N represents the total number of operations;

[0083] The collected data is as follows, Table 1.2:

[0084]

[0085] According to Table 1.2, calculate the speed for each time period (as shown in Table 1.3):

[0086]

[0087]

[0088] According to Table 1.3, perform summation:

[0089]

[0090] Calculate the sum of squares of time terms (as shown in Table 1.4):

[0091]

[0092]

[0093]

[0094] According to Table 1.4, perform summation:

[0095] 4 + 4 + 9 + 9 = 26

[0096] Calculate the root mean square:

[0097]

[0098] Calculate the final interaction behavior eigenvalue

[0099] S = 20.00 + 2.55 = 22.55

[0100] The calculated interaction behavior eigenvalue S = 22.55 (unit: pixels / second).

[0101] In addition, the page dwell time is an important parameter for analyzing the learner's concentration. The entry time and departure time of each page are extracted from the interaction log, and the dwell time is calculated. For example:

[0102] Page a: Entry time 10:00:00, departure time 10:05:00, dwell time 5 minutes

[0103] Page b: Entry time 10:05:00, departure time 10:15:00, dwell time 10 minutes

[0104] Finally, data such as mouse trajectory features and page dwell time are comprehensively summarized to form interaction behavior feature data.

[0105] The learning path storage sub-module serializes and stores the learner's learning path according to the knowledge point switching feature data and interaction behavior feature data within the set time window, records the knowledge point learning order and interaction behavior pattern of the learner within different time periods, and obtains the learning behavior data set;

[0106] Based on the knowledge point switching feature data and interaction behavior feature data, the learner's behavior data is integrated according to the time window, and the learning path sequence is recorded. Taking the time window set to 10 minutes as an example, the data of a certain learner may be as follows:

[0107] 2024-02-12 10:00-10:10: Knowledge point A, mouse speed 5 pixels / second, dwell time 6 minutes

[0108] 2024-02-12 10:10-10:20: Knowledge point B, mouse speed 8 pixels / second, dwell time 4 minutes

[0109] The data of all learners is stored in a serialized structure, so that the learning path can be conveniently traced back during subsequent analysis, such as:

[0110] {(T1, K1, S1),......, (T n , K n , S n )}

[0111] Among them, T i represents the time window, K i represents the knowledge point accessed by the learner in this time window, and S i represents the interaction behavior feature value. Integrate the data of all learners to obtain the learning behavior data set.

[0112] Please refer to Figure 3 , the learning rhythm mutation detection module includes a learning behavior volatility calculation sub-module, a knowledge point jump frequency analysis sub-module, and a rhythm mutation behavior screening sub-module;

[0113] The learning behavior volatility calculation sub-module extracts the learner's behavior data in adjacent time periods based on the learning behavior dataset, calculates the learning activity frequency in each time period, and uses the formula:

[0114]

[0115] Calculate the learning behavior volatility W, analyze it in combination with the time series, and generate learning behavior volatility data. Among them, B i represents the number of learning behavior activities in the i-th time period, and B i+1 represents the number of learning behavior activities in the (i + 1)-th time period, t i represents the timestamp corresponding to the time point i, and t i+1 represents the timestamp corresponding to the time point i + 1, and w is the number of calculation samples;

[0116] Based on the learning behavior dataset, extract the learner's learning behavior data in adjacent time periods, obtain the learning activity frequency in each time period, and calculate the learning behavior volatility. Specifically, extract the learning behavior events of each learner in each time period from the dataset, such as course clicks, video plays, quiz submissions, etc., and count the total number of behaviors in each time period. Taking a certain learner as an example, his learning behavior data is as follows:

[0117] Table 2.1 Learning behavior statistical data:

[0118]

[0119] As shown in Table 2.1, the number of learning behavior events in each time period is used to calculate the volatility between adjacent time periods, and the calculation is carried out using the formula. Taking the data of this learner as an example:

[0120]

[0121] Finally, the calculated learning behavior volatility data of this learner is W = 1.25.

[0122] The knowledge point jump frequency analysis sub-module analyzes the knowledge point switching situation of the learner in each time period according to the learning behavior volatility data, counts the number of knowledge point jumps of the learner in consecutive time periods, calculates the knowledge point switching frequency, and judges the change trend to obtain the knowledge point jump frequency data;

[0123] Call the learning behavior volatility data, analyze the knowledge point switching situation of the learner in different time periods, extract the knowledge point access records of each learner in consecutive time periods, and calculate the number of knowledge point jumps. Taking the data of a certain learner as an example, assuming his knowledge point access records are as follows:

[0124] Table 2.2 Knowledge point jump statistics:

[0125]

[0126] Calculate the knowledge point jump frequency:

[0127]

[0128] Finally, the knowledge point jump frequency data f = 0.12 of the learner is calculated.

[0129] Based on the knowledge point jump frequency data, the rhythm mutation behavior screening sub-module detects the time interval fluctuation of the learner, calculates the change rate of the time interval between adjacent time periods, compares it with the set fluctuation threshold, screens the learner behavior patterns with rhythm mutations, and obtains the learning rhythm mutation status record;

[0130] Call the knowledge point jump frequency data, detect the time interval fluctuation of the learner, calculate the change rate of the time interval between adjacent time periods, and perform screening in combination with the set fluctuation threshold. Calculate the change rate of the time interval:

[0131]

[0132] Taking the data of a certain learner as an example, the following time interval data is set:

[0133] Table 2.3 Time interval data:

[0134] Time period (minutes) Time interval (seconds) 0-10 600 10-20 650 20-30 580 30-40 720 40-50 680

[0135] Calculate the change rate of the time interval:

[0136]

[0137] The setting of the fluctuation threshold θ is based on the median of the time interval volatility of all learners And take 1.5 times the median as the threshold, that is:

[0138]

[0139] Suppose in the data of 500 learners, it is calculated that

[0140]

[0141] Then

[0142] θ = 1.5 × 0.08 = 0.12

[0143] Finally, it is detected that the learner's learning rhythm has mutated between 20 - 30 minutes and 30 - 40 minutes, and then the learning rhythm mutation status record is obtained.

[0144] Please refer to Figure 4,The behavior trajectory stability evaluation module includes the mouse trajectory comparison submodule, the page switching frequency calculation submodule, and the pause time fluctuation analysis submodule;

[0145] The mouse trajectory comparison submodule obtains the mouse movement trajectory data in the learning behavior dataset, and uses the formula according to the mouse trajectory coordinate point sequence in adjacent time periods:

[0146]

[0147] Calculate the mouse track offset value S t , get the mouse track overlap analysis results, where X i,t , Y i,t Represent the coordinates of the i-th trajectory point at time t, X i,t+1 , Y i,t+1 represents the corresponding coordinates at time t+1, and G represents the total number of trajectory points;

[0148] Get the mouse movement trajectory data in the learning behavior dataset, select the mouse trajectory data sequence with a time interval of 5 seconds, extract the mouse trajectory coordinate point set in each time period, set the minimum time span of each trajectory point to 0.1 seconds, and the number of trajectory points in each time period is determined by the mouse movement speed and dwell time. Call the mouse trajectory data of adjacent time periods, compare the trajectory shape and trajectory point distribution, and calculate the mouse trajectory similarity index. Assume that in a certain time period, the mouse trajectory contains 5 key coordinate points:

[0149] (2, 3), (4, 6), (8, 9), (12, 15), (15, 20)

[0150] In the next time period t+1, the corresponding coordinate point becomes:

[0151] (2.1, 3.2), (4.2, 6.1), (7.8, 9.3), (12.5, 15.4), (15.3, 19.9)

[0152] The overall deviation of the mouse track is calculated using the formula, and the actual data is used in the calculation:

[0153]

[0154] As shown in Table 3.1, the mouse track deviation of a user in different time periods is listed:

[0155] Table 3.1 Mouse track overlap calculation table

[0156]

[0157] As shown in Table 3.1, the calculated overlap degree of the mouse trajectory indicates the differences in the stability of mouse movement in different time periods. The larger the trajectory deviation value, the more unstable the learner's mouse behavior is.

[0158] The page switching frequency calculation sub-module collects the page access records in adjacent time periods based on the learning behavior data set, calculates the number of page switches, statistically analyzes the change rate of page switches per unit time, and obtains the page switching frequency value.

[0159] Based on the learning behavior data set, extract the page access records in different time periods, set the time window to 10 seconds, count the number of page switches per unit time, calculate the change rate, and screen the high-frequency switching time periods. Suppose a learner's page switching situation in a time period is as follows: access page A at the 1st second, access page B at the 2nd second, still on page B at the 3rd second, switch to page C at the 5th second, switch to page D at the 7th second, and return to page A at the 9th second. Then the number of page switches is 4 times, and the page switching frequency per unit time is calculated as:

[0160]

[0161] Among them, F represents the page switching frequency, E represents the number of page switches, and T represents the time window length. The calculated page switching frequency value is 0.4.

[0162] The pause time fluctuation analysis sub-module filters the pause time series of the same knowledge point according to the page switching frequency value and the analysis result of the mouse trajectory overlap degree, calculates the difference between the maximum value and the minimum value, determines the change range of the pause time, judges the pause time fluctuation situation of the learner on the same knowledge point, and obtains the record of the stability state of the behavior trajectory.

[0163] Combined with the page switching frequency value and the analysis result of the mouse trajectory overlap degree, filter the pause time series of the same knowledge point, set the detection time window to 30 seconds, count the time interval of the learner's pause on the same knowledge point, and calculate the difference between the maximum pause time and the minimum pause time. Suppose a learner's pause times on a knowledge point are: 12 seconds, 15 seconds, 18 seconds, 10 seconds, 20 seconds. Then calculate the change range of the pause time:

[0164] D = T max -T min = 20 - 10 = 10

[0165] Among them, D represents the change range of the pause time, T max represents the maximum pause time, and T min represents the minimum pause time. The calculated change range of the pause time is 10 seconds.

[0166] According to the change range of the pause time, set the pause time fluctuation threshold D th, the threshold is set based on the average pause time fluctuation range of knowledge points in the normal state of learners. This value is calculated through large-scale learning behavior data statistics. After analyzing the pause time data of different learners, the fluctuation of the pause time on the same knowledge point is statistically calculated, and the 75th percentile of the pause time change range of all learners on multiple knowledge points is taken as the benchmark. The calculation formula is as follows:

[0167] D th = Q3(D)

[0168] Among them, Q3(D) represents the third quartile of the pause time change range D data of all learners on all knowledge points. Suppose in the statistical data of 1000 learners, the quartiles of the pause time change range are calculated as follows:

[0169] The first quartile Q1(D) = 4 seconds;

[0170] The second quartile (median) Q2(D) = 6 seconds;

[0171] The third quartile Q3(D) = 8 seconds;

[0172] Then the pause time fluctuation threshold D th Take Q3(D) = 8 seconds, that is, among 75% of the learners, the pause time change range does not exceed 8 seconds. If it exceeds this value, it is considered that there is a large fluctuation in learning behavior.

[0173] According to the pause time change range D of learners on the same knowledge point for state judgment, if D ≤ D th , then the learner's behavior trajectory is stable, otherwise it is determined to be unstable. For the previously calculated pause time change range D = 10 seconds, make a judgment:

[0174] D = 10 > D th = 8

[0175] Since the pause time fluctuation of the learner exceeds the set threshold, the behavior trajectory is determined to be in an unstable state.

[0176] By recording the pause time fluctuations of multiple knowledge points, the following state table is finally formed:

[0177] Table 3.2 Record Table of Learner Behavior Trajectory Stability Status

[0178]

[0179] As shown in Table 3.2, the behavior trajectories of learners on knowledge points 101 and 104 are determined to be unstable, while the pause time changes on knowledge points 102 and 103 are small and are determined to be in a stable state, thus obtaining the record of the learner's behavior trajectory stability status.

[0180] Please refer to Figure 5 , the knowledge point backtracking index calculation module includes a learning behavior recording sub-module, a knowledge point backtracking calculation sub-module, and a learning path backtracking analysis sub-module;

[0181] The learning behavior recording sub-module obtains the access data of the learner on the same knowledge point according to the learning behavior data set, records the access times, access order, and jump path, detects the behavior data of the learner at each learning stage, counts the repeated access records of the same knowledge point and the jump trajectory of the learning path, and obtains the knowledge point access trajectory data;

[0182] To obtain the access data of the learner on the same knowledge point, including the access times, access order, and jump path, in order to specifically determine the access behavior of each learner, it is first necessary to set the acquisition method of the learning behavior data. For example, the access data can be obtained through the log records of the online learning platform, including the access timestamp, user ID, knowledge point ID, etc. of each knowledge point. Suppose the access time records of a learner on a certain knowledge point in a day are as follows:

[0183] Table 4.1 Learner Knowledge Point Access Record Table

[0184]

[0185] In Table 4.1, it can be seen that the learner (U001) accessed the knowledge point K101 multiple times in a day, and the access order was jumpy. This behavior indicates that the learner may have encountered problems in understanding the knowledge point K101. The system needs to record all access times, access order, and jump paths, and obtain the access trajectory data of the knowledge point by counting the repeated access records of the same knowledge point. The jump access path can be represented by a directed graph. For example:

[0186] K101→K102→K101→K103→K101

[0187] By analyzing the backtracking situation of the directed graph, the learning path pattern of the learner can be further judged to form the knowledge point access trajectory data.

[0188] Based on the knowledge point access trajectory data, the knowledge point backtracking calculation sub-module counts the access order in different learning stages, analyzes the repeatability in the learning path, quantifies the knowledge point backtracking degree in combination with the access data, and uses the formula:

[0189]

[0190] Calculate the backtracking frequency R of the learner on the same knowledge point f , and obtain the backtracking frequency data, where q irepresents the number of times the knowledge point is accessed for the i-th time, W i represents the importance weight of this access, P i represents the degree of path deviation at the i-th time in the learning path, and Z represents the total number of times the learner accesses this knowledge point;

[0191] Based on the knowledge point access trajectory data, calculate the backtracking frequency of the learner on the same knowledge point, count the access order in different learning stages, and analyze the repeatability in the learning path. First, define the calculation method of the backtracking frequency and calculate the knowledge point backtracking frequency R f The number of accesses, importance weight, and degree of path deviation need to be considered, where:

[0192] q i : the number of times the knowledge point is accessed for the i-th time. Assume that the learner U001 accesses K101 3 times, then q1 = 1, q2 = 1, q3 = 1;

[0193] z i (Access weight) is assigned proportionally within the range of 5 - 15 minutes according to the learning duration. If the learning duration is less than 5 minutes, the value is assigned 0.5; if it is 5 - 10 minutes, the value is assigned 1.0; if it is more than 10 minutes, the value is assigned 1.5. For example, in this case, the learning durations of the learner are 4 minutes, 10 minutes, and 6 minutes respectively, then:

[0194] z1 = 0.5 (4 minutes);

[0195] z2 = 1.0 (10 minutes);

[0196] z3 = 1.0 (6 minutes);

[0197] P i : degree of path deviation, which is set as the number of steps deviated from the recommended path. For example, if the recommended path is K101 → K102 → K103, and the actual path is K101 → K102 → K101 → K103, then the degree of path deviation is calculated as follows:

[0198] The first backtracking (K101 → K102 → K101): deviated 1 time;

[0199] The second backtracking (K101 → K103 → K101): deviated 2 times;

[0200] Finally, substitute into the formula:

[0201]

[0202] The calculation result shows that the backtracking frequency of the learner on the knowledge point K101 is 0.53. If this value is higher than 0.5 (indicating that the backtracking frequency reaches 50%), it means that the learner has a deviation in understanding this knowledge point, and finally, the backtracking frequency data is obtained.

[0203] The learning path backtracking analysis sub-module analyzes the backtracking degree of the learning path during the knowledge point backtracking process of learners according to the backtracking frequency data, calculates the intensity of the learning path backtracking, compares the backtracking behavior patterns of different learners, combines the path backtracking frequency and backtracking level, summarizes the understanding deviations of learners on different knowledge points, and obtains the knowledge point backtracking status record;

[0204] According to the backtracking frequency data, analyze the backtracking degree of the learning path during the knowledge point backtracking process of learners, calculate the intensity of the learning path backtracking, compare the backtracking behavior patterns of different learners, and first obtain the backtracking sequence of the learning path, as shown in Table 2:

[0205] Table 4.2 Learning Path Backtracking Situation Table

[0206]

[0207] As can be seen from Table 4.2, the backtracking frequencies of different learners on different knowledge points are different. Among them, the backtracking frequency of U003 on K103 is relatively high, indicating that there may be a large understanding deviation. By calculating the average backtracking frequency of all learners:

[0208]

[0209] If the backtracking frequency of a certain learner is greater than the average value of 0.48, it means that there is a large understanding deviation on this knowledge point (U001 and U003 have large understanding deviations). Here, 0.48 is set as the backtracking reference value, and its calculation method is based on the average value of the historical backtracking frequencies of all learners. This reference value will be dynamically adjusted as the sample data increases. The higher the reference value, the higher the backtracking degree of the overall learning group on the knowledge point, thus affecting the evaluation of the individual learning path. On this basis, combined with the path backtracking frequency and backtracking level, summarize the understanding of learners on different knowledge points, and finally obtain the knowledge point backtracking status record.

[0210] Please refer to Figure 6 , the portrait dynamic adjustment module includes a portrait update trigger sub-module, a portrait adjustment calculation sub-module, and a portrait construction plan generation sub-module;

[0211] The portrait update trigger sub-module analyzes the changing trend of learners' behaviors based on the learning rhythm mutation status record, the behavior trajectory stability status record, and the knowledge point backtracking status record, judges whether the learning state has mutated, compares the changing rates of the learning rhythm before and after, combines the behavior trajectory stability evaluation value and the knowledge point backtracking frequency, calculates the trigger conditions for portrait update, and obtains the portrait update requirement record;

[0212] Obtain the records of the learning rhythm mutation state, the behavior trajectory stability state, and the knowledge point backtracking state. By collecting the access frequencies, jump paths, and backtracking times of learners on different knowledge points, establish a behavior dataset of learners. Set the time window to 7 days, compare the differences in access times between the previous and current time windows, and define the calculation method of learning rhythm mutation as follows:

[0213]

[0214] Among them, O t represents the total number of learning accesses within the current time window, and O t-1 represents the total number of learning accesses within the previous time window. When ΔO exceeds the preset threshold θ O = 0.3, it is determined as the learning rhythm mutation state. The setting basis of this threshold is that if the learning rhythm mutation amplitude exceeds 30%, it may represent a large change in the learner's knowledge absorption speed, learning goal, or habit. This value is set based on the data analysis of the learning group, and 0.3 is selected as the critical value between stability and mutation. If it is lower than 0.3, it is considered that the learning behavior is still within the normal fluctuation range. Based on the knowledge point backtracking state record, count the backtracking frequencies of learners on different knowledge points, and set the backtracking threshold R f When it is lower than 0.2, it indicates that the learner has a high degree of mastery of the knowledge point. When the backtracking threshold is higher than 0.5, it indicates that there are deviations in the mastery of the knowledge point. The setting of this threshold comes from the analysis of the data of multiple groups of learners. It is found that when the backtracking frequency is lower than 0.2, the subsequent test correct rate can be stabilized above 85%, while when it is higher than 0.5, the test correct rate is usually lower than 60%. Therefore, 0.2 and 0.5 are used as the boundaries of the knowledge mastery degree. Then calculate the standard deviation σ p of the learning path. When σ p > 0.5, the setting of this threshold is based on the observation of the path deviation data. If the standard deviation is lower than 0.5, it indicates that the learning path is relatively stable, and the deviation degrees of most access paths are maintained within a reasonable range. If it exceeds 0.5, it indicates that the learning path fluctuates greatly, there are abnormalities in the learning order of knowledge points, which affects knowledge absorption, and further indicates that the learning path stability is low and the portrait update needs to be triggered. Call the calculation results of the learning rhythm mutation state, the behavior trajectory stability state, and the knowledge point backtracking state, comprehensively analyze the learning behavior characteristics of learners, and obtain the portrait update requirement record.

[0215] Table 5.1 Example of learning rhythm mutation calculation

[0216]

[0217] As shown in Table 5.1, when ΔS = 0.6, it exceeds the threshold of 0.3, and it is determined that the learning rhythm has mutated, triggering the portrait update requirement record.

[0218] The image adjustment calculation sub-module quantifies the amplitude of image update according to the image update requirement record, in combination with the knowledge point backtracking frequency and the learning path deviation rate, and in combination with the learning rhythm mutation state and the behavior trajectory stability, using the formula:

[0219]

[0220] Calculate the image dynamic adjustment parameter U p , where K i represents the knowledge point backtracking frequency for the i-th time, M i represents the importance weight of this knowledge point, D i represents the learning path deviation rate for the i-th time, L i represents the learning path recommended value, I i represents the learning rhythm mutation state value, o i represents the behavior trajectory stability coefficient, and zn represents the total number of learning stages of the learner;

[0221] Call the image update requirement record, and calculate the image dynamic adjustment parameter based on the knowledge point backtracking frequency and the learning path deviation rate. K i is the knowledge point backtracking frequency, with a value range of [0, 1], and M i is the importance weight of this knowledge point, using a standardized score, with a range of [0, 1]. This weight is set according to the centrality of the knowledge point in the curriculum structure. For example, the weight value of the core knowledge point is close to 1, while the weight value of the marginal knowledge point is less than 0.5. The weight calculation is weighted based on the knowledge point correlation, the appearance frequency in the learning path, and the error rate.

[0222] Set the following data:

[0223] Table 5.2 Image Adjustment Calculation Example

[0224]

[0225] Substitute into the formula for calculation:

[0226]

[0227] Finally, calculate the image dynamic adjustment parameter U p ≈0.86, and obtain the image adjustment parameter record.

[0228] This result indicates that when the image dynamic adjustment parameter is greater than 0.8, a relatively large adjustment needs to be made to the learner's image, otherwise the adjustment amplitude can be reduced.

[0229] The portrait construction plan generation sub-module, based on the portrait adjustment parameter record, compares the adjustment amplitude of the difference learner portraits, calculates the portrait change trend, analyzes the stability of the learner portraits, adjusts the portrait weight distribution, summarizes the portrait update strategy, and obtains the learner portrait construction plan;

[0230] Call the portrait adjustment parameter record, analyze the adjustment amplitude of different learner portraits, and calculate the portrait change trend. First, set the portrait weight adjustment rule, and the weight change formula is:

[0231] W new =W old +α×(U p -U th )

[0232] Among them, W new is the adjusted portrait weight, W old is the current portrait weight, U p is the calculated portrait dynamic adjustment parameter, U th is the portrait adjustment threshold. Set U th =0.75. The setting of this value is based on historical data analysis. Usually, when the portrait dynamic adjustment parameter exceeds 0.75, it means that the learner's knowledge structure has changed significantly and the portrait needs to be adjusted greatly. Otherwise, the adjustment range can be controlled within a smaller range.

[0233] α is the adjustment coefficient. Set α = 0.2. This coefficient determines the adjustment amplitude of the portrait weight. If it is set too high, it may cause the portrait to change too much. If it is set too low, the adjustment effect is not obvious. After comprehensively comparing the effects of different adjustment coefficients, set 0.2 to balance the portrait update frequency and stability.

[0234] Taking the portrait weight W old =0.6 of a certain learner for calculation:

[0235] W new =0.6+0.2×(0.86 - 0.75)

[0236] W new =0.6+0.2×0.11=0.6+0.022=0.622

[0237] The finally adjusted portrait weight is 0.622. By comparing the adjustment amplitudes of different learner portraits, calculating the portrait change trend, summarizing the portrait update strategy, the learner portrait construction plan is obtained.

[0238] This result shows that when the portrait dynamic adjustment parameter is close to the threshold U th , the portrait adjustment amplitude is small, and when U p is much larger than U thWhen this happens, the adjustment range of the image weight will increase accordingly to dynamically adapt to the learner's knowledge mastery situation.

[0239] A method for constructing a learner behavior portrait based on data analysis, comprising the following steps:

[0240] S1: Collect the course learning logs of the online learning platform, record the time points of knowledge point switching and the task completion timestamps, obtain the interaction behavior logs, extract the mouse trajectory coordinate sequence and the page stay duration, and segment and store the learning path according to the time window to generate a learning behavior data set;

[0241] S2: Call the learning behavior data set, calculate the learning behavior volatility within adjacent time periods, judge the change of the jump frequency between knowledge points, obtain the time interval fluctuation situation, compare it with the set volatility threshold, screen the behavior patterns with fluctuating learning rhythms, and obtain the learning rhythm fluctuation state record;

[0242] S3: Call the learning behavior data set, calculate the overlap degree of the mouse movement trajectories, the difference in page switching frequencies, and the change range of the pause time within adjacent time periods, compare the fluctuation of the pause time for the same knowledge point, and obtain the behavior trajectory stability state record;

[0243] S4: Call the learning behavior data set, count the number of repeated visits of the learner to the same knowledge point, record the knowledge point jump path, calculate the knowledge point backtracking frequency, obtain the deviation degree of the knowledge point backtracking in the learning path, and obtain the knowledge point backtracking state record;

[0244] S5: According to the learning rhythm fluctuation state record, the behavior trajectory stability state record, and the knowledge point backtracking state record, judge the portrait update requirement, adjust the trigger condition for updating the learner portrait, and optimize the learner behavior characteristics in combination with the learning path deviation rate to generate a learner portrait construction plan.

[0245] The above is only a preferred embodiment of the present invention, and does not limit the present invention in other forms. Any person skilled in the art may use the disclosed technical content to make changes or modifications into equivalent embodiments with equivalent changes and apply them to other fields. However, as long as it does not depart from the technical solution content of the present invention, any simple modification, equivalent change, and modification made to the above embodiments based on the technical essence of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. A learner behavior portrait construction system based on data analysis, characterized in that: The system includes: The learning behavior collection module, based on the online learning platform, combines the course learning log and the interaction behavior log to record the learner's learning path sequence and store it, generating a learning behavior dataset; The learning rhythm mutation detection module calculates the learning behavior volatility of adjacent time periods according to the learning behavior dataset, judges the learning behavior mutation situation, analyzes the change of the jump frequency between knowledge points, detects the time interval fluctuation situation, screens the behavior patterns with rhythm mutations, and obtains the learning rhythm mutation status record; The behavior trajectory stability evaluation module calculates the overlap degree of the behavior paths in different time periods according to the learning behavior dataset, analyzes the stability of the learner's behavior pattern, and obtains the behavior trajectory stability status record; The knowledge point backtracking index calculation module calculates the learner's same knowledge point backtracking frequency and learning path deviation rate according to the learning behavior dataset, and obtains the knowledge point backtracking status record; The portrait dynamic adjustment module combines the learning rhythm mutation status record, the behavior trajectory stability status record and the knowledge point backtracking status record, adjusts the portrait update trigger condition and judges the portrait update requirement, and generates a learner portrait construction plan.

2. The learner behavior portrait construction system based on data analysis according to claim 1, characterized in that The learning behavior dataset includes learning behavior sequences, knowledge point switching records, task completion timestamps, mouse trajectory coordinate sequences, page stay durations, and learning path sequences; the learning rhythm mutation status record includes learning behavior volatility, knowledge point jump frequency change values, task completion interval fluctuation values, and rhythm mutation behavior pattern identifiers; the behavior trajectory stability status record includes behavior path overlap degrees, page switching frequency change values, pause time change ranges, and learning behavior pattern stability identifiers; the knowledge point backtracking status record includes knowledge point backtracking frequencies, learning path deviation rates, knowledge point repeated access times, and jump path backtracking degrees; the learner portrait construction plan includes portrait update trigger conditions, portrait update requirement determination values, learning path deviation adjustment strategies, knowledge point backtracking adjustment strategies, and learner portrait adjustment plans.

3. The system for constructing a learner behavior portrait based on data analysis according to claim 1, characterized in that, The learning behavior collection module includes a knowledge point switching record sub-module, an interaction behavior data processing sub-module, and a learning path storage sub-module; The knowledge point switching record sub-module collects the knowledge point switching information in the course learning log, obtains the knowledge point access records of each learner at different time points, calculates the time interval and switching frequency of adjacent knowledge point switches, aggregates the data of all learners, and generates knowledge point switching feature data; The interaction behavior data processing sub-module extracts the mouse trajectory coordinate sequence based on the interaction behavior log, calculates the average speed, acceleration and number of stop points of the mouse movement, counts the page stay duration, and uses the formula: Calculate the interactive behavior feature value S and obtain the interactive behavior feature data, where x i ,y i Represents the coordinate value of the mouse at the i-th time point, t i Represents the timestamp of the time point, and N represents the total number of operations; The learning path storage sub-module serializes and stores the learner's learning path according to the set time window based on the knowledge point switching feature data and the interaction behavior feature data, records the knowledge point learning order and interaction behavior pattern of the learner in different time periods, and obtains the learning behavior dataset.

4. The learner behavior portrait construction system based on data analysis according to claim 1, characterized in that The learning rhythm mutation detection module includes a learning behavior volatility calculation sub-module, a knowledge point jump frequency analysis sub-module, and a rhythm mutation behavior screening sub-module; The learning behavior volatility calculation sub-module extracts the behavior data of the learner in adjacent time periods based on the learning behavior data set, calculates the learning activity frequency in each time period, and uses the formula: Calculate the learning behavior volatility W, analyze it in combination with the time series, and generate learning behavior volatility data, where B i represents the number of learning behavior activities in the i-th time period, B i+1 represents the number of learning behavior activities in the (i + 1)-th time period, t i represents the timestamp corresponding to the time point i, t i+1 represents the timestamp corresponding to the time point (i + 1), and w is the number of calculation samples; The knowledge point jump frequency analysis sub-module analyzes the knowledge point switching situation of the learner in each time period according to the learning behavior volatility data, counts the number of knowledge point jumps of the learner in consecutive time periods, calculates the knowledge point switching frequency, and judges the change trend to obtain the knowledge point jump frequency data; The rhythm mutation behavior screening sub-module detects the time interval fluctuation situation of the learner based on the knowledge point jump frequency data, calculates the change rate of the time interval between adjacent time periods, compares it with the set fluctuation threshold, screens the learner behavior patterns with rhythm mutations, and obtains the learning rhythm mutation status record.

5. The system for constructing a learner behavior portrait based on data analysis according to claim 1, wherein The behavior trajectory stability evaluation module includes a mouse trajectory comparison sub-module, a page switching frequency calculation sub-module, and a pause time fluctuation analysis sub-module; The mouse trajectory comparison sub-module obtains the mouse movement trajectory data in the learning behavior data set, and according to the mouse trajectory coordinate point sequence in adjacent time periods, uses the formula: Calculate the mouse trajectory offset value S t , obtain the mouse trajectory overlap degree analysis result, where X i,t , Y i,t respectively represent the coordinates of the i-th trajectory point at time t, X i,t+1 , Y i,t+1 represent the corresponding coordinates at time t + 1, and G represents the total number of trajectory points; The page switching frequency calculation sub-module collects the page access records in adjacent time periods based on the learning behavior data set, calculates the number of page switches, counts the change rate of page switches per unit time, and obtains the page switching frequency value; The pause time fluctuation analysis sub-module screens the pause time series of the same knowledge point according to the page switching frequency value and the mouse trajectory overlap degree analysis result, calculates the difference between the maximum value and the minimum value, determines the pause time change range, judges the pause time fluctuation situation of the learner at the same knowledge point, and obtains the behavior trajectory stability status record.

6. The system for constructing a learner behavior portrait based on data analysis according to claim 1, wherein The knowledge point backtracking index calculation module includes a learning behavior record sub-module, a knowledge point backtracking calculation sub-module, and a learning path backtracking analysis sub-module; The learning behavior record sub-module obtains the access data of the learner at the same knowledge point according to the learning behavior data set, records the access times, access order, and jump path, detects the behavior data of the learner in each learning stage, counts the repeated access records of the same knowledge point and the jump trajectory of the learning path, and obtains the knowledge point access trajectory data; The knowledge point backtracking calculation sub-module counts the access order in different learning stages based on the knowledge point access trajectory data, analyzes the repeatability in the learning path, and quantifies the degree of knowledge point backtracking in combination with the access data, using the formula: Calculate the backtracking frequency R of the learner on the same knowledge point f , and obtain the backtracking frequency data, where q i represents the number of times the knowledge point is visited for the i-th time, W i represents the importance weight of this visit, P i represents the degree of path deviation at the i-th time in the learning path, and Z represents the total number of times the learner visits this knowledge point; The learning path backtracking analysis sub-module analyzes the degree of learning path backtracking of the learner during the knowledge point backtracking process according to the backtracking frequency data, calculates the intensity of learning path backtracking, compares the backtracking behavior patterns of different learners, combines the path backtracking frequency and backtracking level, and summarizes the understanding deviation of the learner at different knowledge points to obtain the knowledge point backtracking status record.

7. The system for constructing a learner behavior portrait based on data analysis according to claim 1, wherein The portrait dynamic adjustment module includes a portrait update trigger sub-module, a portrait adjustment calculation sub-module, and a portrait construction scheme generation sub-module; The image update trigger sub-module analyzes the changing trend of learners' behaviors based on the recorded learning rhythm mutation state, the recorded behavioral trajectory stability state, and the recorded knowledge point backtracking state, determines whether there is a mutation in the learning state, compares the changing rates of the learning rhythm before and after, combines the behavioral trajectory stability evaluation value and the knowledge point backtracking frequency, calculates the trigger condition for image update, and obtains the image update requirement record; The image adjustment calculation sub-module quantifies the image update amplitude according to the image update requirement record, combines the knowledge point backtracking frequency and the learning path deviation rate, and combines the learning rhythm mutation state and the behavioral trajectory stability, using the formula: Calculate the dynamic adjustment parameter U of the image p , where K i represents the knowledge point backtracking frequency at the i-th time, M i represents the importance weight of this knowledge point, D i represents the learning path deviation rate at the i-th time, L i represents the learning path recommendation value, I i represents the learning rhythm mutation state value, o i represents the behavior trajectory stability coefficient, and zn represents the total number of learning stages of the learner; The image construction plan generation sub-module, based on the recorded image adjustment parameters, compares the image adjustment amplitudes of different learners, calculates the image change trend, analyzes the stability of the learner image, adjusts the image weight distribution, summarizes the image update strategy, and obtains the learner image construction plan.

8. A method for constructing a learner behavior portrait based on data analysis, characterized in that, Executed according to the system described in any one of claims 1-7, including the following steps: S1: Collect the course learning logs of the online learning platform, record the time points of knowledge point switching and the task completion timestamps, obtain the interaction behavior logs, extract the mouse trajectory coordinate sequence and the page stay duration, segment and store the learning path according to the time window, and generate the learning behavior data set; S2: Call the learning behavior data set, calculate the learning behavior volatility within adjacent time periods, judge the change situation of the jump frequency between knowledge points, obtain the time interval fluctuation situation, compare it with the set fluctuation threshold, screen the behavior patterns with fluctuating learning rhythm, and obtain the learning rhythm fluctuation state record; S3: Call the learning behavior data set, calculate the overlap degree of the mouse movement trajectories, the difference in page switching frequency, and the change range of the pause time within adjacent time periods, compare the fluctuation situation of the pause time for the same knowledge point, and obtain the behavioral trajectory stability state record; S4: Call the learning behavior data set, count the number of repeated visits of the learner to the same knowledge point, record the knowledge point jump path, calculate the knowledge point backtracking frequency, obtain the deviation degree of the knowledge point backtracking in the learning path, and obtain the knowledge point backtracking state record; S5: According to the learning rhythm fluctuation state record, the behavioral trajectory stability state record, and the knowledge point backtracking state record, judge the image update requirement, adjust the learner image update trigger condition, optimize the learner behavior characteristics in combination with the learning path deviation rate, and generate the learner image construction plan.

Images