A learning plan recommendation method, system and educational robot

Abstract

The invention discloses a learning plan recommendation method, system and educational robot, which relates to the technical field of educational robots. The method uses a trained personalized learning demand identification model to identify the demand, outputs the identified personalized learning demand; matches learning resources and learning plans according to the personalized learning needs of target students, and outputs the optimal learning path; collects feedback data for comprehensive evaluation and generates a state coefficient from the evaluation data; if the state coefficient does not exceed the expectation, uses a multi-objective optimization algorithm to optimize the personalized optimization learning path of the target student, and classifies the target students based on a pre-trained density clustering algorithm; after identifying abnormal learning behavior, a multi-level alarm mechanism is enabled, and according to the abnormal learning behavior, the corresponding response strategy is output from the emergency response strategy library. The current learning path is adjusted and optimized to make the student's learning path more matched with the current learning resources and learning plan.

Description

Technical Field

[0001] The present invention relates to the technical field of educational robots, and in particular to a learning plan recommendation method, system and educational robot. Background Art

[0002] Educational robots are high-tech intelligent devices designed specifically for the education field. They deeply integrate cutting-edge technologies from multiple disciplines, such as artificial intelligence, robotics, education, and cognitive science. These robots are usually equipped with sophisticated sensor systems, flexible robotic arms or mobile chassis, and powerful computing and data processing units, enabling them to perceive the environment, understand instructions, perform complex tasks, and interact naturally and smoothly with users.

[0003] In educational scenarios, educational robots play multiple roles: they can be students' personal tutors, providing customized learning plans and instant feedback based on their learning progress and abilities; they can also be interesting teaching assistants, stimulating students' interest and motivation through gamified learning methods; in addition, they can also assist teachers in classroom management, reduce teachers' burden, and improve teaching efficiency and quality.

[0004] In the Chinese invention patent with the authorization announcement number CN109377430B, a learning plan recommendation method and a learning client are disclosed, including: generating test questions for each knowledge point of the target learning material selected by the first student, obtaining the target test questions for which the student's knowledge point test score in each knowledge point test question is lower than the preset score threshold, obtaining the learning content for learning the target knowledge point corresponding to the target test question from the target learning material; after the student completes the study of the learning content, outputting the learning ability test question corresponding to the target knowledge point, which is different from the target test question; obtaining the learning ability test result obtained by the student's learning ability test on the learning ability test question, which is used to reflect the student's learning ability; and recommending to the student a target learning plan for the target learning material that is compatible with the learning ability test result. By implementing the embodiment of the present invention, the student's learning ability can be used as the basis for recommending a learning plan to the student, thereby facilitating the purpose of improving learning efficiency.

[0005] Combining the above applications and the prior art:

[0006] After students enter the learning state, especially when they have learning difficulties in interdisciplinary project learning and need additional tutoring and support, various methods are usually used to improve students' learning outcomes, such as formulating targeted learning strategies and plans, and matching learning resources in a targeted manner.

[0007] In this scenario, the existing study plan recommendation method is usually formulated by the teacher or oneself according to actual learning needs and goals after the learning task is launched. However, when the environment in the learning area is not suitable for learning, for example, there is a lot of noise in the learning area and the students' attention is distracted, the students' learning state may be greatly affected. In this learning scenario, if the study plan is still formulated under ideal conditions, students who are not very focused may have certain difficulties in executing it, and the matching degree between the original study plan and the students will also be reduced. It is also impossible to adjust and optimize the learning path in real time according to the current learning state, which may ultimately affect the learning effect.

[0008] To this end, the present invention provides a learning plan recommendation method, system and educational robot. Summary of the invention

[0009] 1. Technical issues to be resolved

[0010] In view of the deficiencies of the prior art, the present invention provides a learning plan recommendation method, system and educational robot. When used, the learning resources and learning plans are matched according to the personalized learning needs of the target students, and the optimal learning path is output. Feedback data is collected for comprehensive evaluation and a state coefficient is generated from the evaluation data. If the state coefficient does not exceed the expectation, a multi-objective optimization algorithm is used to optimize the personalized optimized learning path of the target students, and the target students are classified based on the pre-trained density clustering algorithm; a multi-level alarm mechanism is enabled after identifying abnormal learning behavior, and the corresponding response strategy is output from the emergency response strategy library according to the abnormal learning behavior. The current learning path is adjusted and optimized so that the student's learning path is more matched with the current learning resources and learning plan; thereby solving the technical problems recorded in the background technology.

[0011] (II) Technical solution

[0012] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0013] A learning plan recommendation method, comprising: when the noise level in a learning area exceeds expectations, using the collected attention data as input, screening out students with low concentration and taking them as target students; if the learning status of the target students is worse than expected, issuing a data collection instruction to the outside;

[0014] After collecting the student portrait data of the target students, the student portrait is generated and dynamically updated, and the trained personalized learning needs identification model is used to identify the needs and output the identified personalized learning needs;

[0015] Match learning resources and learning plans according to the personalized learning needs of target students, output the optimal learning path, collect feedback data for comprehensive evaluation and generate status coefficients from the evaluation data. If the status coefficient does not exceed expectations, send feedback adjustment instructions to the outside.

[0016] Use a multi-objective optimization algorithm to optimize the personalized optimization learning path of the target students. After collecting the learning data of the target students, classify the target students based on the pre-trained density clustering algorithm;

[0017] After identifying abnormal learning behavior, a multi-level alarm mechanism is enabled. According to the correspondence between abnormal learning behavior and emergency response, the emergency response strategy library outputs the corresponding response strategy. Among them, after feature extraction of the abnormal learning behavior obtained by detection, the corresponding abnormal features are obtained, and an emergency response strategy library containing response strategies for different types of events is pre-built.

[0018] It includes collecting noise data in the learning area by pre-arranged noise sensors, and generating the noisiness of the sound based on the noise data under dimensionless conditions. , in the obtained noisiness When the noise threshold is exceeded, an attention evaluation instruction is sent to the outside world; the noise level of the sound is generated based on the noise data. The way is as follows:

[0019] ;

[0020] Where: is the number of time nodes, is the total number of location nodes, are the noise intensities at location node i and time node j, respectively, is the weight coefficient at the corresponding time node and location node.

[0021] This includes collecting student attention data through a sensor network arranged in the learning area. After receiving the attention evaluation instruction, the student attention data is used as input and the trained attention evaluation model is used to evaluate the degree of concentration. If the student's attention concentration does not exceed expectations, he or she will be identified as a target student.

[0022] Including real-time tracking of target students' learning status data, including learning behavior, progress and learning performance data, and obtaining the similarity between the target learning status data and the corresponding qualification standards through similarity analysis; if the similarity data does not exceed the similarity threshold, a data collection instruction is issued to the outside.

[0023] Including, after receiving the data collection instruction, collecting the student portrait data of the target student, including the target student's learning behavior, status data and social and emotional fluctuation data in the learning area, and generating a student portrait data set after aggregation;

[0024] After using the Transformer model to analyze the data in the student portrait data set, the LSTM neural network is used to process the time series data, and the multimodal data is fused to generate the student's high-dimensional feature vector;

[0025] Taking high-dimensional feature vectors as input, a pre-trained hybrid neural network generates student profiles, which are dynamically updated through continuous data input.

[0026] Including collecting diverse learning resources and summarizing them to generate a learning resource library; based on the personalized learning needs of target students, the learning resource library recommends and outputs personalized learning resources, learning plans and strategies;

[0027] After building the knowledge graph of the learning content, use the reinforcement learning algorithm to obtain the optimal learning path, and dynamically update the learning path based on learning progress and feedback.

[0028] This includes collecting multi-dimensional evaluation indicators after a preset observation period, including the target students' knowledge mastery, learning enthusiasm, participation and emotional state, and summarizing them to generate a set of learning indicators;

[0029] Taking the multi-dimensional evaluation indicators of the current stage in the learning indicator set as input, the post-training learning state evaluation model is used to perform a comprehensive evaluation and generate a learning state value.

[0030] Including marking each sub-stage with a learning state value and generating a state coefficient from a continuous learning state value The way is as follows:

[0031] ;

[0032] Where: is the learning state value at time t, is the independent variable, Indicates the start time To current time Every moment in history, is the control coefficient, which controls the growth rate of the time weight. is the time decay factor, a, b and c are constants, Initial time To current time .

[0033] This includes, after receiving feedback adjustment instructions, using the knowledge tracking model to indicate the students' mastery of different knowledge points, and drawing a learning curve by analyzing the students' performance at different time points;

[0034] After predicting students' learning performance and knowledge mastery status based on the knowledge tracking model, a multi-objective optimization algorithm is used to optimize the current personalized optimization learning path.

[0035] It includes collecting various data of target students, including student behavior status data, learning interaction data and learning feedback data; after preprocessing and feature extraction, several target students are classified into different student classes, and the current learning path of the target students is optimized based on the learning paths of similar student classes.

[0036] This includes collecting feedback data from students and teachers, and using natural language processing algorithms to identify the subject and sentiment of the feedback data; using feedback data as input, using trained anomaly detection models to identify abnormal learning behaviors, and if any, issuing a first-level alarm command to the outside world;

[0037] If after responding to the feedback data, the number of consecutive first-level alarm instructions received still exceeds the expected number, a feedback coefficient is generated based on the status data of the first-level alarm instructions received. , if the feedback coefficient obtained is When the preset feedback threshold is exceeded, a secondary alarm command is issued to the outside.

[0038] Including, generating feedback coefficients based on the status data of the received first-level alarm command The way is as follows:

[0039] ;

[0040] Where: To observe the cycle length, , is the exponential decay kernel function, is the time attenuation coefficient, indicating the time The degree of influence of the event at time t; , is the current abnormality degree function, is the current value, is a qualified value, and The sum of the weight coefficients is 1. is the noise level at time t.

[0041] A learning plan recommendation system includes an attention analysis unit, which uses the collected attention data as input to screen out students with low concentration levels and select them as target students when the noise level in the learning area exceeds expectations, and sends a data collection instruction to the outside if the learning status of the target students is worse than expected;

[0042] The learning needs identification unit collects the student portrait data of the target students, generates the student portrait and updates it dynamically, uses the trained personalized learning needs identification model to identify the needs, and outputs the identified personalized learning needs;

[0043] The learning plan recommendation unit matches learning resources and learning plans according to the personalized learning needs of target students, outputs the optimal learning path, collects feedback data for comprehensive evaluation and generates a state coefficient based on the evaluation data. If the state coefficient does not exceed expectations, it issues feedback adjustment instructions to the outside.

[0044] The classification unit uses a multi-objective optimization algorithm to optimize the personalized optimization learning path of the target students. After collecting the learning data of the target students, the target students are classified based on the pre-trained density clustering algorithm;

[0045] The emergency response unit activates a multi-level alarm mechanism after identifying abnormal learning behavior. According to the correspondence between abnormal learning behavior and emergency response, the emergency response strategy library outputs the corresponding response strategy.

[0046] A learning plan recommendation educational robot includes at least one processor;

[0047] The memory is used to store executable instructions, where the instructions are executed by at least one of the processors to enable at least one of the processors to perform the steps of the learning plan recommendation method.

[0048] (III) Beneficial effects

[0049] The present invention provides a learning plan recommendation method, system and educational robot, which have the following beneficial effects:

[0050] 1. Based on the noise level obtained It is possible to judge whether the learning environment in the learning area will affect the students' learning status, and after determining the corresponding standards, the impact of the noise situation in the learning area can be timely assessed to ensure the learning status.

[0051] 2. Through the collected attention data, we can judge the degree of concentration of students when they are studying. If students are not focused in a noisy environment or their learning state is poor, we can adjust their learning state.

[0052] 4. Construct a student portrait data set and merge it to generate a student portrait, which quantitatively describes the student's current learning status and progress, which is conducive to evaluating and describing the student's current learning status. When the current learning status is poor, targeted evaluation can be quickly achieved.

[0053] 5. By identifying students' learning needs, if students are in a state of independent learning, the corresponding learning resources can be quickly matched to users to improve their current learning status; the current learning path can be optimized to guide students to learn according to the new learning path so that they can better adapt to the current learning scenario.

[0054] 6. By generating a status coefficient from several learning status values, a comprehensive evaluation of the learning status of some students with poor learning status screened out from the learning area can be achieved, and the degree of improvement of the students' learning status can be judged based on the status coefficient; the current learning path can be adjusted and optimized to make the students' learning path more compatible with the current learning resources and learning plans; several students can be classified and processed separately, and they can be switched between them when the matching between the learning path and the students' learning habits is insufficient.

[0055] 7. By collecting and analyzing students' feedback data, if any abnormality is identified, an alarm command is issued to the outside to achieve reminder and warning effects; by generating feedback coefficients Complete the construction of the secondary alarm instructions to avoid a single alarm method being ignored. When students are in a poor learning environment, abnormal feedback can be handled in a timely manner to avoid the poor learning environment and learning effect having a greater negative impact on students. The emergency response strategy library outputs the corresponding response strategy to handle abnormal feedback from students in a timely manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 A flow chart of a learning plan recommendation method of the present invention;

[0057] Figure 2 A system structure diagram of a learning plan recommendation method of the present invention;

[0058] Figure 3 The present invention is a schematic diagram of the structure of a learning plan recommendation educational robot. DETAILED DESCRIPTION

[0059] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0060] See also Figure 1 The present invention provides a learning plan recommendation method, comprising:

[0061] Step 1: When the noise level in the learning area exceeds expectations, the collected attention data is used as input to screen out students with low concentration and select them as target students. If the learning status of the target students is worse than expected, a data collection instruction is issued to the outside.

[0062] The step 1 includes the following contents:

[0063] Step 101: After defining the student's study area, the noise data in the study area is collected by the pre-arranged noise sensor, and the noise level of the sound is generated according to the noise data at each location node and each time node. , dimensionless condition, according to the obtained noisiness To determine the noise level of the learning environment, the following methods are used:

[0064] ;

[0065] Where: is the number of time nodes, is the total number of location nodes, are the noise intensities at location node i and time node j, respectively, is the weight coefficient of the corresponding time node and location node, and its value falls between 0 and 1;

[0066] Pre-set noise thresholds based on historical data and expectations for sound management in the learning area;

[0067] The noise level obtained When the noise level exceeds the preset threshold, it means that the noise level in the learning area is high, which may affect the students' learning concentration. At this time, an attention evaluation instruction is issued to the outside.

[0068] When in use, after entering the learning state in the learning area, the corresponding noise level is constructed based on the monitoring data by monitoring the noise in the learning area. , according to the obtained noise It can judge whether the learning environment in the learning area will affect the students' learning state, and after determining the corresponding standards, it can timely deal with the impact of the noise situation in the learning area to ensure the learning state;

[0069] Step 102: Arrange a sensor network in the learning area, such as a noise sensor, a posture sensor, a facial recognition sensor, and a camera; collect student attention data, such as noise, posture, and expression data, through the sensor network; train a machine learning algorithm with the labeled sample data to obtain a trained attention evaluation model;

[0070] After receiving the attention evaluation instruction, the student's attention data is used as input, and the trained attention evaluation model is used to evaluate the degree of concentration, so as to obtain the concentration of the student's attention. The student's learning attention is evaluated based on the concentration. If the student's concentration does not exceed expectations, it means that the student may have a problem of inattention in school, and he or she will be regarded as a target student.

[0071] When in use, as a further content, when students are in a relatively noisy state, considering that the noisy environment will have a certain impact on the students' learning state, at this time, the collected attention data can be used to judge the students' concentration level during learning. If the concentration level is insufficient, it needs to be processed in a timely manner.

[0072] Step 103: real-time tracking of the target student's learning status data, including learning behavior, progress and learning performance data, after setting corresponding qualification standards for the learning status data, similarity analysis is performed to obtain the similarity between the target learning status data and the corresponding qualification standards;

[0073] If the similarity data obtained does not exceed the preset similarity threshold, it means that the target student’s current learning status is not good. In order to improve his learning effect, he may need to receive auxiliary learning. At this time, a data collection instruction is issued to the outside.

[0074] When using, combine the contents in steps 101 to 103:

[0075] As a further content, after screening out the students with inattention, we take them as the target and analyze the students' learning status through similarity analysis. In a noisy environment, when students are inattentive, if their learning status is poor, it is necessary to adjust their learning status and adjust their learning strategies.

[0076] In this scenario, the existing study plan recommendation method is usually formulated by the teacher or oneself according to actual learning needs and goals after the learning task is launched. However, when the environment in the learning area is not suitable for learning, for example, there is a lot of noise in the learning area and the students' attention is distracted, the students' learning state may be greatly affected. In this learning scenario, if the study plan is still formulated under ideal conditions, students who are not very focused may have certain difficulties in executing it, and the matching degree between the original study plan and the students will also be reduced. It is also impossible to adjust and optimize the learning path in real time according to the current learning state, which may ultimately affect the learning effect.

[0077] Step 2: After collecting the student portrait data of the target students, generate the student portrait and dynamically update it, use the trained personalized learning needs identification model to identify the needs, and output the identified personalized learning needs;

[0078] The step 2 includes the following contents:

[0079] Step 201: After receiving the data collection instruction, collect the student portrait data of the target student, including learning behavior status data: integrating various learning activities including video watching, homework submission, online tests, etc.; social and emotional data: analyzing facial expressions, voice intonation and social media interaction content to obtain emotional fluctuation information; aggregating the target student's learning behavior, status data and social and emotional fluctuation data in the learning area to generate a student portrait data set;

[0080] Step 202: After using the Transformer model to analyze the data in the student portrait data set, use the LSTM neural network to process the time series data to obtain dynamic change information, and fuse the multimodal data to generate a high-dimensional feature vector of the student, such as text, numerical, and categorical data; wherein the high-dimensional feature vector represents the student's current learning status, interest areas, emotional state, and knowledge mastery;

[0081] Taking high-dimensional feature vectors as input, a pre-trained hybrid neural network generates student portraits, which are dynamically updated through continuous data input;

[0082] When in use, a student portrait data set is constructed based on the collected data and then merged to generate a student portrait, which can quantitatively describe the student's current learning status and progress, which is conducive to evaluating and describing the student's current learning status. When the current learning status is poor, targeted evaluation can be quickly achieved;

[0083] Step 203: Use the labeled sample data to train a deep LSTM neural network to obtain a trained personalized learning needs recognition model, use the student portrait data as input, use the trained personalized learning needs recognition model to perform needs recognition, and output the recognized personalized learning needs;

[0084] When using, combine the contents in steps 201 to 203:

[0085] After obtaining the student portrait data, the student learning needs are analyzed and identified. In the current learning scenario, by identifying the student's learning needs, if the student is in a state of autonomous learning, the corresponding learning resources can be quickly matched to the user to improve the current learning status.

[0086] Step 3: Match learning resources and learning plans according to the personalized learning needs of target students, output the optimal learning path, collect feedback data for comprehensive evaluation, and generate a state coefficient from the evaluation data. If the state coefficient does not exceed expectations, send feedback adjustment instructions to the outside;

[0087] The step three includes the following contents:

[0088] Step 301: collect diverse learning resources, including learning strategies, micro-courses, interactive exercises, experimental simulations, and academic articles, and aggregate them to generate a learning resource library; dynamically recommend target learning based on the personalized learning needs of target students, and output personalized learning resources, learning plans, and strategies from the learning resource library to complete the recommendation process of learning plans and learning resources;

[0089] Step 302: construct a knowledge graph of learning content, present learning content and concepts in the form of nodes and edges, where nodes represent specific knowledge points and edges represent the relationship between them, use a reinforcement learning algorithm to guide the agent to explore the optimal learning path, obtain the optimal learning path, and dynamically update the learning path based on the student's learning progress and feedback;

[0090] When using it, after matching students with corresponding learning resources and learning plans, the current learning path is optimized, and students are guided to learn according to the new learning path, which can be more adapted to the current learning scenario;

[0091] Step 303: After the target student receives the personalized learning resources and learning strategies, after a preset observation period, multi-dimensional evaluation indicators are collected, including the target student's knowledge mastery, learning enthusiasm, participation, and emotional state, and the learning indicator set is generated after aggregation;

[0092] After training the convolutional neural network with the labeled sample data, a post-training learning state evaluation model is obtained;

[0093] Taking the multi-dimensional evaluation indicators of the current stage in the learning indicator set as input, the post-training learning state evaluation model is used to perform comprehensive evaluation and generate a learning state value;

[0094] Step 304: After marking each sub-stage with a learning state value, a state coefficient is generated from the continuous learning state value. , the way is as follows:

[0095] ;

[0096] Where: is the learning state value at time t, is the independent variable, Indicates the start time To current time Every moment in history, is the control coefficient, which controls the growth rate of the time weight, and its value is between 0 and 1. is the time decay factor, with a value between 0 and 1, a, b and c are constants, with a value between 1 and 10, Initial time To current time ;

[0097] If the state coefficient If the threshold is not exceeded, it means that the students are not very focused at the current stage. After recommending corresponding learning strategies and resources to the target students with poor learning status, the improvement of learning status and effect has not reached the expected level. At this time, feedback adjustment instructions are sent to the outside.

[0098] When used, combine the contents in steps 301 to 304:

[0099] As a further step, multi-dimensional evaluation indicators are collected after the observation period. As a form of feedback, the students' current learning status can be evaluated, and by generating a status coefficient from several learning status values, a comprehensive evaluation of the learning status of some students with poor learning status screened out in the learning area can be achieved. The degree of improvement of the students' learning status can be judged based on the status coefficient to verify whether further assistance is needed for the students' learning process.

[0100] Step 4: Use a multi-objective optimization algorithm to optimize the personalized optimization learning path of the target students. After collecting the learning data of the target students, classify the target students based on the pre-trained density clustering algorithm;

[0101] The step 4 includes the following contents:

[0102] Step 401: After receiving the feedback adjustment instruction, the knowledge tracking model is used to indicate the students' mastery of different knowledge points, and the learning curve is drawn by analyzing the students' performance at different time points to identify the learning bottleneck and the changes in knowledge mastery; after predicting the students' possible future learning performance and knowledge mastery status based on the knowledge tracking model, the multi-objective optimization algorithm is used to optimize the current personalized optimization learning path, and the students can be given priority to review weak links or challenge new knowledge areas, so as to adjust the learning plan in advance; and the reinforcement learning algorithm is used to adjust the recommendation strategy according to the students' real-time feedback to improve the learning path and content adaptation degree;

[0103] When used, as a further content, when the student's learning status still fails to meet expectations, the student's learning status is predicted based on historical data, and the current learning path is adjusted and optimized to make the student's learning path more compatible with the current learning resources and learning plan;

[0104] Step 402: Considering that there are many target students who are not focused and have poor learning status, and there are certain differences between different target students, various data of target students are collected, including student behavior status data, such as learning time, frequency, completed tasks, test scores, etc.; learning interaction data, including clicks, browsing records, types of resources used, etc.; and students' learning feedback data on learning content and learning process;

[0105] After preprocessing and feature extraction of behavior data, interaction data, and feedback data, a number of target students are classified based on a pre-trained density clustering algorithm to obtain different student classes. The current learning path of the target students is optimized based on the learning paths of similar student classes.

[0106] It can identify the learning styles of different groups of students, such as those who prefer visual, auditory or kinesthetic learning, fast learners and slow learners, etc. It can identify different learning styles and needs, provide more targeted recommendations, monitor user usage patterns and preference changes, and adjust system functions and content provision in a timely manner;

[0107] When using, combine the contents in steps 401 to 402:

[0108] As a further improvement, when there are many students with poor learning status in the current stage, considering that each student has a certain difference in the learning path, several students can be classified and dealt with separately. As a further supplement, in order to increase the diversity and variability of learning paths, when the current learning path and students' learning habits are not well matched, they can be switched between them.

[0109] Step 5: After identifying abnormal learning behavior, a multi-level alarm mechanism is enabled. According to the correspondence between abnormal learning behavior and emergency response, the emergency response strategy library outputs the corresponding response strategy;

[0110] The step five includes the following contents:

[0111] Step 501: After designing questionnaires and interviews, collect feedback data from students, teachers and parents, and use natural language processing algorithms to identify the subject and sentiment of the feedback data; train a machine learning algorithm with the labeled sample data to obtain a trained anomaly detection model; use the feedback data as input and use the trained anomaly detection model to identify abnormal learning behavior, and if any, issue a first-level alarm command to the outside;

[0112] When in use, after completing the learning resources, learning strategies and paths for the user, and adjusting the learning plan, if the student's learning scenario changes significantly, by collecting and analyzing the student's feedback data, if any abnormality is identified in the student, an alarm command is issued to the outside to achieve a reminder and warning effect.

[0113] Step 502: If the number of consecutive first-level alarm instructions received after responding to the feedback data still exceeds the expectation, the alarm mechanism is activated. At this time, under dimensionless conditions, a feedback coefficient is generated based on the state data of the first-level alarm instruction received. , as follows:

[0114] ;

[0115] Where: To observe the cycle length, , is the exponential decay kernel function, is the time attenuation coefficient, with a value between 0 and 1, indicating that time The degree of influence of the event at time t; , is the current abnormality degree function, is the current value, is a qualified value, and The sum of the weight coefficients is 1. is the noise level at time t;

[0116] If the feedback coefficient is obtained If the pre-set feedback threshold is exceeded, the alarm rule is met, indicating that the learning status of the target students who are not focused and have poor learning status has not improved after obtaining new learning strategies and learning resources. At this time, a secondary alarm instruction is issued to the outside.

[0117] When used, as a further context, by generating the feedback coefficient Complete the construction of secondary alarm instructions to avoid single alarm mode being ignored. When students are in a poor learning environment, abnormal feedback can be handled in time, avoiding the negative impact of poor learning environment and learning effect on students and protecting students' health.

[0118] Step 503: extract features from the abnormal learning behavior detected and obtain corresponding abnormal features, and pre-build an emergency response strategy library containing response strategies for different types of events, such as adjusting learning plans and psychological counseling services; after receiving the secondary alarm instruction, the emergency response strategy library can output the corresponding response strategy based on the correspondence between the abnormal features and the emergency response;

[0119] When using, combine the contents in steps 501 to 503:

[0120] In response to abnormal feedback from students, the emergency response strategy library outputs corresponding response strategies. When students have abnormal feedback, it can be discovered and handled in time, at least ensuring that when the current learning environment, learning status and learning plan are not suitable, the student's mental state is healthy, which can protect the student's status.

[0121] See also Figure 2 The present invention provides a learning plan recommendation system, comprising:

[0122] The attention analysis unit uses the collected attention data as input to screen out students with low concentration levels and select them as target students when the noise level in the learning area exceeds expectations. If the learning status of the target students is worse than expected, a data collection instruction is issued to the outside.

[0123] The learning needs identification unit collects the student portrait data of the target students, generates the student portrait and updates it dynamically, uses the trained personalized learning needs identification model to identify the needs, and outputs the identified personalized learning needs;

[0124] The learning plan recommendation unit matches learning resources and learning plans according to the personalized learning needs of target students, outputs the optimal learning path, collects feedback data for comprehensive evaluation and generates a state coefficient based on the evaluation data. If the state coefficient does not exceed expectations, it issues feedback adjustment instructions to the outside.

[0125] The classification unit uses a multi-objective optimization algorithm to optimize the personalized optimization learning path of the target students. After collecting the learning data of the target students, the target students are classified based on the pre-trained density clustering algorithm;

[0126] The emergency response unit activates a multi-level alarm mechanism after identifying abnormal learning behavior. According to the correspondence between abnormal learning behavior and emergency response, the emergency response strategy library outputs the corresponding response strategy.

[0127] See also Figure 3 The present invention provides a learning plan recommendation educational robot, comprising:

[0128] at least one processor;

[0129] The memory is used to store executable instructions, where the instructions are executed by at least one of the processors to enable at least one of the processors to perform the steps of the learning plan recommendation method.

[0130] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example 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 performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0131] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0132] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only some logical function divisions. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0133] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0134] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for recommending a learning plan, characterized in that: include, When the noise level in the learning area exceeds expectations, the collected attention data is used as input to screen out students with low concentration and use them as target students. If the learning status of the target students is worse than expected, a data collection instruction is issued to the outside. The noise level of the sound is generated based on the noise data. The way is as follows: ； Where: is the number of time nodes, is the total number of location nodes, are the noise intensities at location node i and time node j, respectively, is the weight coefficient at the corresponding time node and location node; After collecting the student portrait data of the target students, the student portrait is generated and dynamically updated, and the trained personalized learning needs identification model is used to identify the needs and output the identified personalized learning needs; Match learning resources and learning plans according to the personalized learning needs of target students, output the optimal learning path, collect feedback data for comprehensive evaluation and generate status coefficients from the evaluation data If the state coefficient does not exceed the expectation, a feedback adjustment instruction is sent to the outside; each sub-stage is marked with a learning state value, and the state coefficient is generated by continuous learning state values. The way is as follows: ； Where: is the learning state value at time t, is the independent variable, Indicates the start time At every historical moment to the current time T, is the control coefficient, which controls the growth rate of the time weight. is the time decay factor, , and is a constant, Initial time To the current time T; Use a multi-objective optimization algorithm to optimize the personalized optimization learning path of the target students. After collecting the learning data of the target students, classify the target students based on the pre-trained density clustering algorithm; After identifying abnormal learning behavior, a multi-level alarm mechanism is enabled. According to the correspondence between abnormal learning behavior and emergency response, the emergency response strategy library outputs the corresponding response strategy. The abnormal learning behavior detected is subjected to feature extraction to obtain the corresponding abnormal features, and an emergency response strategy library containing response strategies for different types of events is pre-built. The feedback data of students and teachers is collected, and the natural language processing algorithm is used to identify the theme and emotional tendency of the feedback data. The feedback data is used as input, and the trained anomaly detection model is used to identify abnormal learning behavior. If it exists, a first-level alarm instruction is issued to the outside. If after responding to the feedback data, the number of consecutive first-level alarm instructions received still exceeds the expected number, a feedback coefficient is generated based on the status data of the first-level alarm instructions received. , if the feedback coefficient obtained is When the preset feedback threshold is exceeded, a secondary alarm command is issued to the outside; Generate feedback coefficient based on the status data of the first-level alarm command received The way is as follows: ； Where: To observe the cycle length, , is the exponential decay kernel function, is the time attenuation coefficient, indicating the time The degree of influence of the event at time t; , is the current abnormality degree function, is the current value, is a qualified value, and The sum of the weight coefficients is 1. is the noise level at time t.

2. A learning plan recommendation method according to claim 1, characterized in that: The noise data in the learning area is collected by the pre-arranged noise sensors. Under dimensionless conditions, the noisiness of the sound is generated based on the noise data. , in the obtained noisiness When the preset noise threshold is exceeded, an attention evaluation instruction is issued to the outside.

3. A learning plan recommendation method according to claim 2, characterized in that: The sensor network arranged in the learning area collects students' attention data. After receiving the attention evaluation instruction, the student's attention data is used as input and the trained attention evaluation model is used to evaluate the degree of concentration. If the student's attention concentration does not exceed expectations, he or she will be regarded as a target student.

4. A learning plan recommendation method according to claim 3, characterized in that: Track the target students' learning status data in real time, including learning behavior, progress and learning performance data, and obtain the similarity between the target learning status data and the corresponding qualification standards through similarity analysis; if the similarity data does not exceed the similarity threshold, issue a data collection instruction to the outside.

5. A learning plan recommendation method according to claim 4, characterized in that: After receiving the data collection instruction, the student portrait data of the target student is collected, including the target student's learning behavior, status data, and social and emotional fluctuation data in the learning area, and the student portrait data set is generated after aggregation; After using the Transformer model to analyze the data in the student portrait data set, the LSTM neural network is used to process the time series data, and the multimodal data is fused to generate a high-dimensional feature vector of the students; using the high-dimensional feature vector as input, the pre-trained hybrid neural network generates the student portrait, and the student portrait is dynamically updated through continuous data input.

6. A learning plan recommendation method according to claim 5, characterized in that: Collect diverse learning resources and compile them into a learning resource library; based on the personalized learning needs of target students, the learning resource library recommends and outputs personalized learning resources, learning plans and strategies; After building the knowledge graph of the learning content, use the reinforcement learning algorithm to obtain the optimal learning path, and dynamically update the learning path based on learning progress and feedback.

7. A method for recommending a learning plan according to claim 6, characterized in that: After a preset observation period, multi-dimensional evaluation indicators are collected, including the target students' knowledge mastery, learning enthusiasm, participation and emotional state, and summarized to generate a set of learning indicators; Taking the multi-dimensional evaluation indicators of the current stage in the learning indicator set as input, the post-training learning state evaluation model is used to perform a comprehensive evaluation and generate a learning state value.

8. A learning plan recommendation method according to claim 7, characterized in that: After receiving feedback adjustment instructions, the knowledge tracking model is used to indicate the students’ mastery of different knowledge points, and the learning curve is drawn by analyzing the students’ performance at different time points; After predicting students' learning performance and knowledge mastery status based on the knowledge tracking model, a multi-objective optimization algorithm is used to optimize the current personalized optimization learning path.

9. A learning plan recommendation method according to claim 8, characterized in that: Various data of target students are collected, including student behavior status data, learning interaction data and learning feedback data; after preprocessing and feature extraction, several target students are classified to obtain different student classes, and the current learning path of the target students is optimized based on the learning paths of similar student classes.

10. A learning plan recommendation system, applying the learning plan recommendation method according to any one of claims 1 to 9, characterized in that: include, The attention analysis unit uses the collected attention data as input to screen out students with low concentration levels and select them as target students when the noise level in the learning area exceeds expectations. If the learning status of the target students is worse than expected, a data collection instruction is issued to the outside. The learning needs identification unit collects the student portrait data of the target students, generates the student portrait and updates it dynamically, uses the trained personalized learning needs identification model to identify the needs, and outputs the identified personalized learning needs; The learning plan recommendation unit matches learning resources and learning plans according to the personalized learning needs of target students, outputs the optimal learning path, collects feedback data for comprehensive evaluation and generates a state coefficient based on the evaluation data. If the state coefficient does not exceed expectations, it issues feedback adjustment instructions to the outside. The classification unit uses a multi-objective optimization algorithm to optimize the personalized optimization learning path of the target students. After collecting the learning data of the target students, the target students are classified based on the pre-trained density clustering algorithm; The emergency response unit activates a multi-level alarm mechanism after identifying abnormal learning behavior. According to the correspondence between abnormal learning behavior and emergency response, the emergency response strategy library outputs the corresponding response strategy.

11. A learning plan recommendation educational robot, characterized in that: include, at least one processor; A memory for storing executable instructions, wherein the instructions are executed by at least one of the processors to enable at least one of the processors to perform the steps of the method according to any one of claims 1 to 9.

Images