Method and system for generating an educational report based on learning behavior data

Abstract

The application discloses a learning behavior data-based education report generation method and system, and belongs to the technical field of data processing. The method comprises the following steps: collecting learning behavior events; mapping event types and event intensities into initial potential energy values; obtaining residual potential energy values; forming transient behavior vectors; generating transient descriptive labels, and generating a state change trend sequence according to a historical sequence of the transient behavior vectors; outputting an education report, and performing a destruction operation on the transient descriptive labels after the output of the education report. The application maps learning behavior events into potential energy values that decay over time and accumulates to form transient behavior vectors, and destroys the temporarily generated descriptive labels immediately after the output of the education report, thereby solving the problems of user privacy anxiety, intensified data avoidance behavior and report deviation caused by excessive collection and static solidification in the prior art.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for generating educational reports based on learning behavior data. Background Technology

[0002] With the popularization of online education and the advancement of smart campus construction, methods for generating educational reports based on learning behavior data have gradually become a research hotspot in the field of educational informatization. These methods collect multi-source information such as student behavior logs on learning platforms, campus card records, and access control data, and utilize data mining and natural language processing technologies to automatically generate educational reports reflecting students' learning status, providing reference for teachers, parents, and students.

[0003] Existing educational report generation systems typically include a data acquisition module, a feature extraction module, a student profile building module, and a report generation module. The data acquisition module is responsible for obtaining raw behavioral data from various business systems; the feature extraction module cleans and aggregates the raw data, calculating statistical indicators such as learning time, completion rate, and accuracy rate; the student profile building module assigns static labels to students based on long-term accumulated statistical indicators, such as "diligent" or "struggling learner"; and the report generation module uses template filling or natural language generation technology to convert the profile labels and statistical indicators into text reports.

[0004] The existing technology has the following problems in practical applications: First, in order to achieve a comprehensive student profile, the system often needs to collect multi-dimensional behavioral data of students during their time at school, including learning activities, consumption records, social interactions, and even after-school activities. This kind of "panoramic monitoring" can easily cause students and parents to worry about privacy and security, generate resistance to being monitored, and lead to users developing a resistance to the system.

[0005] Secondly, in response to privacy concerns, some students have begun to consciously avoid data collection, such as by studying offline, clicking randomly, or using virtual private networks to hide their real behavior. These avoidance behaviors result in a large amount of missing and noisy data collected, which is seriously deviating from the students' actual learning status.

[0006] Third, reports generated based on distorted data are bound to make misjudgments. For example, genuine hard work may be interpreted as "low activity" or intentional random clicking may be misjudged as "strong interest in learning". When students find that the report does not match their actual situation, their trust in the system will further decrease, and they will become more determined to avoid data collection, forming a vicious cycle of "privacy concerns → data distortion → report bias → less trust → stronger avoidance".

[0007] Fourth, existing technologies typically use static, long-term accumulated labels to describe students. Once a label is generated, it remains attached to the student profile for an extended period, failing to reflect dynamic changes in the student's status in a timely manner. For example, when a student makes progress through effort, the label may remain at the "struggling learner" stage due to the excessive weighting of historical data. This leads to subsequent reports continuously outputting negative evaluations, potentially triggering a self-fulfilling prophecy and further exacerbating user distrust and resistance.

[0008] Therefore, there is an urgent need for an educational report generation method that can protect user privacy, dynamically reflect student status, and avoid fixed labels, in order to improve the accuracy of the reports and user acceptance. Summary of the Invention

[0009] This application provides a method and system for generating educational reports based on learning behavior data. This solves the vicious cycle problem in the prior art where excessive data collection and static labeling lead to mutually reinforcing issues of user privacy concerns, data distortion, and report bias. It achieves the technical effects of dynamically representing learning states based on transient behavior vectors, getting rid of long-term static label attachment, and breaking the cycle of privacy concerns and lack of trust, thereby improving the accuracy of educational reports and user acceptance.

[0010] This application provides a method for generating educational reports based on learning behavior data, including: collecting learning behavior events and extracting the event type, event intensity, and event timestamp of the learning behavior events; According to the preset type-intensity mapping rules, the event type and event intensity are mapped to the initial potential energy value; Assign a corresponding time decay factor according to the event type, and perform time decay calculation on the initial potential energy value according to the event timestamp and the time decay factor to obtain the remaining potential energy value. The remaining potential energy values ​​are accumulated into the corresponding dimension of the multidimensional behavioral potential energy field to form a transient behavioral vector; The values ​​of each dimension in the transient behavior vector are compared with preset dimension thresholds to generate transient descriptive labels, and a state change trend sequence is generated based on the historical sequence of the transient behavior vector. The transient behavior vector, state change trend sequence, and transient descriptive labels are populated into the report generation template to output the educational report. After the educational report is output, the transient descriptive labels are destroyed.

[0011] Furthermore, according to the preset type-intensity mapping rules, the step of mapping event type and event intensity to initial potential energy values ​​includes: Initialize the mapping relationship data table, configure the predefined type identifier as the primary key index of the mapping relationship data table, and store the basic weight value corresponding to the primary key index; Extract the event type, compare the event type with the primary key index, and extract the basic weight value corresponding to the primary key index that successfully matches the event type. Obtain the raw physical measurement data corresponding to the event intensity, and convert the raw physical measurement data into floating-point quantized values; The basic weight value is multiplied by the floating-point quantized value to generate the initial potential energy value, which is then assigned to a preset potential energy variable for storage.

[0012] Furthermore, the steps for performing time decay calculations on the initial potential energy value based on the event timestamp and time decay factor to obtain the remaining potential energy value include: Call the underlying clock interface to read the current time in standard time format; Read the event timestamp and calculate the absolute time difference between the standard time format data and the event timestamp; Locate and extract the time decay factor bound to the event type from the global configuration table; The remaining potential energy is calculated using the formula for calculating remaining potential energy. The formula is as follows: ; in, This represents the remaining potential energy value. This represents the initial potential energy value. The time decay factor, Refers to the absolute time difference. Indicates the base of the natural logarithm; After the remaining potential energy is calculated, the remaining potential energy value is saved in the memory data stack in double precision format.

[0013] Furthermore, the steps of accumulating the remaining potential energy values ​​into the corresponding dimension of the multidimensional behavioral potential energy field to form the transient behavioral vector include: Initialize the multidimensional behavioral potential energy field, which contains at least two independent data nodes, each data node being bound to a uniquely determined behavioral type feature code; Read the behavior type feature code corresponding to the event type and traverse it in the multidimensional behavior potential energy field; When a data node with the same behavior type feature code is encountered, it is identified as the target dimension. Read the historical accumulated values ​​recorded in the target dimension, execute an addition instruction between the historical accumulated values ​​and the remaining potential energy values ​​to obtain the updated dimension value, and overwrite the updated dimension value into the target dimension; The values ​​of all data nodes in the multidimensional behavioral potential energy field are read sequentially according to a preset position order and then concatenated to construct a transient behavioral vector in the form of a one-dimensional array.

[0014] Furthermore, the steps of comparing the values ​​of each dimension in the transient behavior vector with preset dimension thresholds to generate transient descriptive labels include: Traverse the transient behavior vector in the form of a one-dimensional array, and extract the updated dimension value stored therein as the current comparison object; Load the threshold determination rule table and locate and extract the preset dimension threshold corresponding to the array position of the current comparison object; Call the logical comparator to determine whether the current comparison object is numerically greater than the preset dimension threshold; If it is greater than the preset dimension threshold, it is determined to be true, and the address of the text string library that has a mapping relationship with the preset dimension threshold is queried in the threshold determination rule table; The corresponding structured character sequence data is retrieved based on the address of the text string library. The structured character sequence data is assigned to a temporary variable to instantiate the transient descriptive label, and the transient descriptive label is stored in a volatile memory area for later use.

[0015] Furthermore, the steps for generating a state change trend sequence based on the historical sequence of transient behavior vectors include: An internal timer is started, and the transient behavior vector generated each time is continuously appended to the dedicated database as historical node data according to the set sampling period frequency; Read the backtracking time length parameter, extract all historical node data that meet the conditions of the backtracking time length parameter from the dedicated database, and arrange them in chronological order to construct a multi-dimensional time series matrix; By traversing the multidimensional time series matrix, the transient behavior vector of the next adjacent time node is subtracted from the transient behavior vector of the previous time node in one dimension to obtain the set of numerical changes in each dimension. All the numerical change sets are sequentially concatenated and combined to construct the state change trend sequence, and the state change trend sequence is output to the memory cache in the form of a structured data stream.

[0016] Furthermore, the steps to populate the report generation template with transient behavior vectors, state change trend sequences, and transient descriptive labels, and output the educational report, include: Load the report generation template into the working memory, and scan the vector placeholders, trend placeholders, and label placeholders within the report generation template; Extract transient behavior vectors and overwrite vector placeholders; extract state change trend sequences from memory cache and overwrite trend placeholders; extract transient descriptive labels from volatile memory regions and overwrite label placeholders; concatenate to generate a request string sequence. Call the network communication interface to send the request string sequence to the large language model server, receive the parsed text data returned by the large language model server, encapsulate the parsed text data in a layout format to generate an educational report file stream, and persistently save the educational report file stream.

[0017] Furthermore, the steps before loading the threshold determination rule table also include: Set a fixed-length threshold evaluation period; When the runtime reaches the trigger node of the threshold evaluation cycle, read all the historical cumulative values ​​of the corresponding dimension of the multidimensional behavioral potential energy field in the previous cycle in the dedicated database. The arithmetic mean calculation command is executed on all historical cumulative values ​​to obtain the baseline average value for that dimension. Read the control float proportional constant within the configuration environment, multiply the baseline average value by the control float proportional constant, and calculate the dynamic evaluation boundary value. The dynamic evaluation boundary value is overwritten into the threshold determination rule table to replace the original data and serve as a new preset dimension threshold, so that the generation benchmark of transient descriptive labels strictly follows the dynamic shift of the accumulated data of the multidimensional behavioral potential energy field.

[0018] This application provides an educational report generation system based on learning behavior data, which implements a method for generating educational reports based on learning behavior data, including: Event acquisition module, potential energy numerical mapping module, decay calculation module, numerical accumulation module, tag generation module, and report output module; The event acquisition module is used to collect learning behavior events and extract the event type, event intensity and event timestamp of the learning behavior events. The potential energy value mapping module is used to map event type and event intensity to initial potential energy value according to a preset type-intensity mapping rule. The decay calculation module is used to allocate a corresponding time decay factor according to the event type, and to perform time decay calculation on the initial potential energy value according to the event timestamp and the time decay factor to obtain the remaining potential energy value. The numerical accumulation module is used to accumulate the remaining potential energy value into the corresponding dimension of the multidimensional behavioral potential energy field to form a transient behavioral vector. The tag generation module is used to compare the values ​​of each dimension in the transient behavior vector with a preset dimension threshold to generate transient descriptive tags, and to generate a state change trend sequence based on the historical sequence of the transient behavior vector. The report output module is used to fill the transient behavior vector, state change trend sequence and transient descriptive label into the report generation template, output the education report, and perform a destruction operation on the transient descriptive label after the education report is output.

[0019] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages: By employing a data dynamic mapping method based on behavioral potential energy fields, the original learning behavior is transformed into an instantaneous potential energy value that decays naturally over time. This avoids the long-term retention and static qualitative analysis of students' behavioral trajectories, reduces the sensitivity of the data, and helps alleviate students' and parents' concerns about privacy leaks.

[0020] Furthermore, when constructing transient behavior vectors, a time decay factor is introduced to make recent behavior dominate the state representation. Even if students occasionally exhibit avoidance behavior due to privacy concerns, resulting in single data loss or anomalies, no long-term negative qualitative judgment will be made against the students. Instead, the abnormal fluctuations can be quickly covered by the continuous accumulation of subsequent normal behaviors, thereby reducing the students' motivation to deliberately falsify or avoid data.

[0021] Furthermore, when generating transient descriptive labels, a judgment threshold that dynamically floats with the overall data distribution is adopted, combined with a mechanism that destroys labels immediately after the report is output. This ensures that each report only reflects the student's true state at the current moment. Even if a report is biased due to data distortion in the early stages, the bias will disappear as the label is destroyed, and will not accumulate to form a persistent wrong impression of the student. This gradually restores the student's trust in the accuracy of the report, guides the student from deliberate avoidance to natural participation, and breaks out of the vicious cycle of privacy concerns and data distortion. Attached Figure Description

[0022] Figure 1 A flowchart illustrating the method for generating educational reports based on learning behavior data provided in this application embodiment; Figure 2 This is a schematic diagram of the structure of an education report generation system based on learning behavior data provided in an embodiment of this application. Detailed Implementation

[0023] This application provides a method and system for generating educational reports based on learning behavior data. This solves the vicious cycle problem in the prior art where excessive data collection and static labeling lead to user privacy concerns, data distortion, and report bias that reinforce each other. By mapping learning behavior events to potential energy values ​​that decay over time and accumulating them to form transient behavior vectors, and immediately destroying temporary labels after generating the report, this method achieves the technical effects of dynamically representing students based on instantaneous states, avoiding long-term retention of sensitive data, breaking the cycle of privacy concerns and lack of trust, and improving the accuracy of educational reports and user acceptance.

[0024] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0025] like Figure 1 The diagram shows a flowchart of an educational report generation method based on learning behavior data provided in this application embodiment. This method is applied to an educational report generation system based on learning behavior data and includes the following steps: collecting learning behavior events and extracting the event type, event intensity, and event timestamp of the learning behavior events. The event types are used to distinguish different categories of learning behaviors, including but not limited to homework completion events, video viewing events, forum posting events, and quiz answering events. Each event type corresponds to a specific dimension in the multidimensional behavioral potential field.

[0026] The event intensity is a measure of the quantification of the behavioral event. For an assignment completion event, the event intensity is the score of that assignment, expressed as a percentage or a five-point scale; for a video viewing event, the event intensity is the duration of continuous viewing, expressed in minutes; for a forum posting event, the event intensity is the length of the post content, expressed in characters; and for a quiz answering event, the event intensity is the accuracy rate of the quiz, expressed as a percentage.

[0027] The event timestamp is the system time recorded by the system at the moment the event occurred, accurate to the second.

[0028] According to the preset type-intensity mapping rules, the event type and event intensity are mapped to the initial potential energy value; Assign a corresponding time decay factor according to the event type, and perform time decay calculation on the initial potential energy value according to the event timestamp and the time decay factor to obtain the remaining potential energy value. A time decay factor is configured independently for each type of behavioral event. This time decay factor controls the rate at which the influence of that type of behavioral event decays over time, and its unit is negative first second. A larger time decay factor indicates a faster decay of the potential energy of that type of behavioral event, and a more significant dominance of recently occurring behavioral events in the cumulative result. The time decay factor is set based on the relationship between behavior type and learning state representation. For events such as test accuracy, since the knowledge mastery they reflect has greater fluctuations, a larger time decay factor value is configured so that recent test results quickly dominate the potential energy field. For events such as learning habits, such as the timeliness of homework submission, since the behavioral patterns they reflect are relatively stable, a smaller time decay factor value is configured so that historical behaviors still retain some influence.

[0029] The remaining potential energy values ​​are accumulated into the corresponding dimension of the multidimensional behavioral potential energy field to form a transient behavioral vector; The values ​​of each dimension in the transient behavior vector are compared with preset dimension thresholds to generate transient descriptive labels, and a state change trend sequence is generated based on the historical sequence of the transient behavior vector. The transient behavior vector, state change trend sequence, and transient descriptive labels are populated into the report generation template to output the educational report. After the educational report is output, the transient descriptive labels are destroyed.

[0030] Furthermore, according to the preset type-intensity mapping rules, the step of mapping event type and event intensity to initial potential energy values ​​includes: Initialize the mapping relationship data table, configure the predefined type identifier as the primary key index of the mapping relationship data table, and store the basic weight value corresponding to the primary key index; Extract the event type, compare the event type with the primary key index, and extract the basic weight value corresponding to the primary key index that successfully matches the event type. Obtain the raw physical measurement data corresponding to the event intensity, and convert the raw physical measurement data into floating-point quantized values; The basic weight value is multiplied by the floating-point quantized value to generate the initial potential energy value, which is then assigned to a preset potential energy variable for storage.

[0031] Furthermore, the steps for performing time decay calculations on the initial potential energy value based on the event timestamp and time decay factor to obtain the remaining potential energy value include: Call the underlying clock interface to read the current time in standard time format; Read the event timestamp and calculate the absolute time difference between the standard time format data and the event timestamp; The traversal calculation operation is performed at fixed time intervals. Let the current time be... The event timestamp of a certain learning behavior event is The time decay factor for the event type corresponding to this event is The initial potential energy value of the event is Calculate the time difference between the current time and the event timestamp. , The calculation formula is , The unit is seconds.

[0032] Locate and extract the time decay factor bound to the event type from the global configuration table; The remaining potential energy is calculated using the formula for calculating remaining potential energy. The formula is as follows: ; in, This represents the remaining potential energy value. This represents the initial potential energy value. The time decay factor, Refers to the absolute time difference. Indicates the base of the natural logarithm; After the remaining potential energy is calculated, the remaining potential energy value is saved in the memory data stack in double precision format.

[0033] Furthermore, the steps of accumulating the remaining potential energy values ​​into the corresponding dimension of the multidimensional behavioral potential energy field to form the transient behavioral vector include: Initialize the multidimensional behavioral potential energy field, which contains at least two independent data nodes, each data node being bound to a uniquely determined behavioral type feature code; Read the behavior type feature code corresponding to the event type and traverse it in the multidimensional behavior potential energy field; When a data node with the same behavior type feature code is encountered, it is identified as the target dimension. Read the historical accumulated values ​​recorded in the target dimension, execute an addition instruction between the historical accumulated values ​​and the remaining potential energy values ​​to obtain the updated dimension value, and overwrite the updated dimension value into the target dimension; The values ​​of all data nodes in the multidimensional behavioral potential energy field are read sequentially according to a preset position order and then concatenated to construct a transient behavioral vector in the form of a one-dimensional array.

[0034] When a new learning behavior event is collected, the following operations are performed: First, extract the event type and determine the specific dimension coordinates bound to it in the multidimensional behavioral potential energy field based on the event type. These dimension coordinates are positive integers, ranging from 1 to the total number of dimensions. Second, add the remaining potential energy value of the event calculated in step two to the current accumulated value at that specific dimension coordinate to obtain the updated dimension value. The calculation formula is as follows: ,in The updated dimension values. This is the current cumulative value before the dimension was updated. This represents the remaining potential energy value for the event; finally, the updated dimension value is written back to the corresponding dimension position of the multidimensional behavioral potential energy field.

[0035] By continuously performing traversal calculation operations and cumulative update operations, the value of each dimension in the multidimensional behavioral potential field always reflects the cumulative impact of all historical behavioral events in that dimension after attenuation up to the current moment.

[0036] At any given moment, the current values ​​of all dimensions in the multidimensional behavioral potential energy field are extracted and arranged according to the order of the dimension coordinates to form a multidimensional vector, which is called the transient behavioral vector. Let the total number of dimensions be... Then the transient behavior vector is represented as ,in Indicates the first The current cumulative value in each dimension. The characteristic of the transient behavior vector is that recent high-frequency behavior events contribute significantly to the vector components, while the contribution of earlier behavior events gradually weakens until it approaches zero, thus ensuring that the vector always represents the student's latest behavior state.

[0037] Furthermore, the steps of comparing the values ​​of each dimension in the transient behavior vector with preset dimension thresholds to generate transient descriptive labels include: Traverse the transient behavior vector in the form of a one-dimensional array, and extract the updated dimension value stored therein as the current comparison object; Load the threshold determination rule table and locate and extract the preset dimension threshold corresponding to the array position of the current comparison object; Call the logical comparator to determine whether the current comparison object is numerically greater than the preset dimension threshold; If it is greater than the preset dimension threshold, it is determined to be true, and the address of the text string library that has a mapping relationship with the preset dimension threshold is queried in the threshold determination rule table; The corresponding structured character sequence data is retrieved based on the address of the text string library. The structured character sequence data is assigned to a temporary variable to instantiate the transient descriptive label, and the transient descriptive label is stored in a volatile memory area for later use.

[0038] For each dimension in the transient behavior vector, perform the following operation: extract the current value in that dimension, and denote it as the target dimension value. Read the target determination threshold corresponding to this dimension. ;judge Is it greater than ;like Greater than If the condition is met, the tag text string bound to that condition is extracted and added as a transient descriptive tag to the temporary tag set of this report; if Less than or equal to If not, no tag will be generated.

[0039] The target judgment threshold is determined dynamically. A judgment period is preset, for example, seven days. At the beginning of each judgment period, the current values ​​of all dimensions in the multidimensional behavioral potential energy field are recorded; at the end of each judgment period, the average value of the values ​​in each dimension during that period is calculated to obtain the periodic average value of that dimension. Preset a floating coefficient. , This is a dimensionless constant greater than zero, ranging from 0.8 to 1.2. For the next judgment period, the target judgment threshold for each dimension... Calculated using the following formula: ; in, The threshold is used to determine the target. This is the periodic average value of this dimension within the previous judgment period. This is the preset floating coefficient.

[0040] Furthermore, the steps for generating a state change trend sequence based on the historical sequence of transient behavior vectors include: An internal timer is started, and the transient behavior vector generated each time is continuously appended to the dedicated database as historical node data according to the set sampling period frequency; Read the backtracking time length parameter, extract all historical node data that meet the conditions of the backtracking time length parameter from the dedicated database, and arrange them in chronological order to construct a multi-dimensional time series matrix; By traversing the multidimensional time series matrix, the transient behavior vector of the next adjacent time node is subtracted from the transient behavior vector of the previous time node in one dimension to obtain the set of numerical changes in each dimension. All the numerical change sets are sequentially concatenated and combined to construct the state change trend sequence, and the state change trend sequence is output to the memory cache in the form of a structured data stream.

[0041] When preparing to generate an educational report each time, all historical vectors from the most recent reporting period are extracted from the transient behavior vector history sequence to form a transient vector set. Let the number of extracted vectors be... Each vector contains If there are multiple dimensions, then the set of transient vectors is represented as follows: ,in The vector corresponding to the earliest time step, The vector corresponding to the current moment.

[0042] Calculate the difference between any two adjacent transient behavior vectors in the transient vector set. For any adjacent... and , The value range is 1 to Calculate the difference in each dimension to obtain the difference vector. , The formula for calculating each component is as follows: ,in express In the Values ​​in each dimension express In the The values ​​in each dimension.

[0043] Arrange all the difference vectors in chronological order to form a sequence of state change trends. This sequence is represented as follows: In a state change trend sequence, positive components indicate that the potential energy in the corresponding dimension is increasing during that time period, while negative components indicate a decreasing trend. The magnitude of the absolute value of the components indicates the degree of drastic change.

[0044] Furthermore, the steps to populate the report generation template with transient behavior vectors, state change trend sequences, and transient descriptive labels, and output the educational report, include: Load the report generation template into the working memory, and scan the vector placeholders, trend placeholders, and label placeholders within the report generation template; Extract transient behavior vectors and overwrite vector placeholders; extract state change trend sequences from memory cache and overwrite trend placeholders; extract transient descriptive labels from volatile memory regions and overwrite label placeholders; concatenate to generate a request string sequence. Call the network communication interface to send the request string sequence to the large language model server, receive the parsed text data returned by the large language model server, encapsulate the parsed text data in a layout format to generate an educational report file stream, and persistently save the educational report file stream.

[0045] Furthermore, the steps before loading the threshold determination rule table also include: Set a fixed-length threshold evaluation period; When the runtime reaches the trigger node of the threshold evaluation cycle, read all the historical cumulative values ​​of the corresponding dimension of the multidimensional behavioral potential energy field in the previous cycle in the dedicated database. The arithmetic mean calculation command is executed on all historical cumulative values ​​to obtain the baseline average value for that dimension. Read the control float proportional constant within the configuration environment, multiply the baseline average value by the control float proportional constant, and calculate the dynamic evaluation boundary value. The dynamic evaluation boundary value is overwritten into the threshold determination rule table to replace the original data and serve as a new preset dimension threshold, so that the generation benchmark of transient descriptive labels strictly follows the dynamic shift of the accumulated data of the multidimensional behavioral potential energy field.

[0046] like Figure 2 The diagram shown is a structural schematic of an educational report generation system based on learning behavior data provided in this application embodiment. The educational report generation system based on learning behavior data provided in this application embodiment includes: an event acquisition module, a potential energy numerical mapping module, an attenuation calculation module, a numerical accumulation module, a tag generation module, and a report output module. The event acquisition module is used to collect learning behavior events and extract the event type, event intensity and event timestamp of the learning behavior events. The potential energy value mapping module is used to map event type and event intensity to initial potential energy value according to a preset type-intensity mapping rule. The decay calculation module is used to allocate a corresponding time decay factor according to the event type, and to perform time decay calculation on the initial potential energy value according to the event timestamp and the time decay factor to obtain the remaining potential energy value. The numerical accumulation module is used to accumulate the remaining potential energy value into the corresponding dimension of the multidimensional behavioral potential energy field to form a transient behavioral vector. The tag generation module is used to compare the values ​​of each dimension in the transient behavior vector with a preset dimension threshold to generate transient descriptive tags, and to generate a state change trend sequence based on the historical sequence of the transient behavior vector. The report output module is used to fill the transient behavior vector, state change trend sequence and transient descriptive label into the report generation template, output the education report, and perform a destruction operation on the transient descriptive label after the education report is output.

[0047] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0048] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0049] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0050] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0051] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0052] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for generating educational reports based on learning behavior data, characterized in that, Includes the following steps: Collect learning behavior events and extract the event type, event intensity, and event timestamp of each event. According to the preset type-intensity mapping rules, the event type and event intensity are mapped to the initial potential energy value; Assign a corresponding time decay factor according to the event type, and perform time decay calculation on the initial potential energy value according to the event timestamp and the time decay factor to obtain the remaining potential energy value. The remaining potential energy values ​​are accumulated into the corresponding dimension of the multidimensional behavioral potential energy field to form a transient behavioral vector; The values ​​of each dimension in the transient behavior vector are compared with preset dimension thresholds to generate transient descriptive labels, and a state change trend sequence is generated based on the historical sequence of the transient behavior vector. The transient behavior vector, state change trend sequence, and transient descriptive labels are populated into the report generation template to output the educational report. After the educational report is output, the transient descriptive labels are destroyed.

2. The method for generating educational reports based on learning behavior data as described in claim 1, characterized in that, The steps for mapping event type and event intensity to initial potential energy values ​​according to the preset type-intensity mapping rules include: Initialize the mapping relationship data table, configure the predefined type identifier as the primary key index of the mapping relationship data table, and store the basic weight value corresponding to the primary key index; Extract the event type, compare the event type with the primary key index, and extract the basic weight value corresponding to the primary key index that successfully matches the event type. Obtain the raw physical measurement data corresponding to the event intensity, and convert the raw physical measurement data into floating-point quantized values; The basic weight value is multiplied by the floating-point quantized value to generate the initial potential energy value, which is then assigned to a preset potential energy variable for storage.

3. The method for generating educational reports based on learning behavior data as described in claim 1, characterized in that, The steps for performing time decay calculations on the initial potential energy value based on the event timestamp and time decay factor to obtain the remaining potential energy value include: Call the underlying clock interface to read the current time in standard time format; Read the event timestamp and calculate the absolute time difference between the standard time format data and the event timestamp; Locate and extract the time decay factor bound to the event type from the global configuration table; The remaining potential energy is calculated using the formula for calculating remaining potential energy. The formula is as follows: ； in, This represents the remaining potential energy value. This represents the initial potential energy value. The time decay factor, Refers to the absolute time difference. Indicates the base of the natural logarithm; After the remaining potential energy is calculated, the remaining potential energy value is saved in the memory data stack in double precision format.

4. The method for generating educational reports based on learning behavior data as described in claim 1, characterized in that, The steps to accumulate the remaining potential energy values ​​into the corresponding dimension of the multidimensional behavioral potential energy field to form the transient behavioral vector include: Initialize the multidimensional behavioral potential energy field, which contains at least two independent data nodes, each data node being bound to a uniquely determined behavioral type feature code; Read the behavior type feature code corresponding to the event type and traverse it in the multidimensional behavior potential energy field; When a data node with the same behavior type feature code is encountered, it is identified as the target dimension. Read the historical accumulated values ​​recorded in the target dimension, execute an addition instruction between the historical accumulated values ​​and the remaining potential energy values ​​to obtain the updated dimension value, and overwrite the updated dimension value into the target dimension; The values ​​of all data nodes in the multidimensional behavioral potential energy field are read sequentially according to a preset position order and then concatenated to construct a transient behavioral vector in the form of a one-dimensional array.

5. The method for generating educational reports based on learning behavior data as described in claim 1, characterized in that, The steps for generating transient descriptive labels by comparing the values ​​of each dimension in the transient behavior vector with preset dimension thresholds include: Traverse the transient behavior vector in the form of a one-dimensional array, and extract the updated dimension value stored therein as the current comparison object; Load the threshold determination rule table and locate and extract the preset dimension threshold corresponding to the array position of the current comparison object; Call the logical comparator to determine whether the current comparison object is numerically greater than the preset dimension threshold; If it is greater than the preset dimension threshold, it is determined to be true, and the address of the text string library that has a mapping relationship with the preset dimension threshold is queried in the threshold determination rule table; The corresponding structured character sequence data is retrieved based on the address of the text string library. The structured character sequence data is assigned to a temporary variable to instantiate the transient descriptive label, and the transient descriptive label is stored in a volatile memory area for later use.

6. The method for generating educational reports based on learning behavior data as described in claim 1, characterized in that, The steps for generating a state change trend sequence based on the historical sequence of transient behavior vectors include: An internal timer is started, and the transient behavior vector generated each time is continuously appended to the dedicated database as historical node data according to the set sampling period frequency; Read the backtracking time length parameter, extract all historical node data that meet the conditions of the backtracking time length parameter from the dedicated database, and arrange them in chronological order to construct a multi-dimensional time series matrix; By traversing the multidimensional time series matrix, the transient behavior vector of the next adjacent time node is subtracted from the transient behavior vector of the previous time node in one dimension to obtain the set of numerical changes in each dimension. All the numerical change sets are sequentially concatenated and combined to construct the state change trend sequence, and the state change trend sequence is output to the memory cache in the form of a structured data stream.

7. The method for generating educational reports based on learning behavior data as described in claim 1, characterized in that, The steps to populate the report generation template with transient behavior vectors, state change trend sequences, and transient descriptive labels to output an educational report include: Load the report generation template into the working memory, and scan the vector placeholders, trend placeholders, and label placeholders within the report generation template; Extract transient behavior vectors and overwrite vector placeholders; extract state change trend sequences from memory cache and overwrite trend placeholders; extract transient descriptive labels from volatile memory regions and overwrite label placeholders; concatenate to generate a request string sequence. Call the network communication interface to send the request string sequence to the large language model server, receive the parsed text data returned by the large language model server, encapsulate the parsed text data in a layout format to generate an educational report file stream, and persistently save the educational report file stream.

8. The method for generating educational reports based on learning behavior data as described in claim 5, characterized in that, The steps before loading the threshold determination rule table also include: Set a fixed-length threshold evaluation period; When the runtime reaches the trigger node of the threshold evaluation cycle, read all the historical cumulative values ​​of the corresponding dimension of the multidimensional behavioral potential energy field in the previous cycle in the dedicated database. The arithmetic mean calculation command is executed on all historical cumulative values ​​to obtain the baseline average value for that dimension. Read the control float proportional constant within the configuration environment, multiply the baseline average value by the control float proportional constant, and calculate the dynamic evaluation boundary value. The dynamic evaluation boundary value is overwritten into the threshold determination rule table to replace the original data and serve as a new preset dimension threshold, so that the generation benchmark of transient descriptive labels strictly follows the dynamic shift of the accumulated data of the multidimensional behavioral potential energy field.

9. An educational report generation system based on learning behavior data, used to implement the educational report generation method based on learning behavior data as described in any one of claims 1-8, characterized in that, include: Event acquisition module, potential energy numerical mapping module, decay calculation module, numerical accumulation module, tag generation module, and report output module; The event acquisition module is used to collect learning behavior events and extract the event type, event intensity and event timestamp of the learning behavior events. The potential energy value mapping module is used to map event type and event intensity to initial potential energy value according to a preset type-intensity mapping rule. The decay calculation module is used to allocate a corresponding time decay factor according to the event type, and to perform time decay calculation on the initial potential energy value according to the event timestamp and the time decay factor to obtain the remaining potential energy value. The numerical accumulation module is used to accumulate the remaining potential energy value into the corresponding dimension of the multidimensional behavioral potential energy field to form a transient behavioral vector. The tag generation module is used to compare the values ​​of each dimension in the transient behavior vector with a preset dimension threshold to generate transient descriptive tags, and to generate a state change trend sequence based on the historical sequence of the transient behavior vector. The report output module is used to fill the transient behavior vector, state change trend sequence and transient descriptive label into the report generation template, output the education report, and perform a destruction operation on the transient descriptive label after the education report is output.

Images