AI-driven online education system and method based on user behavior analysis

Abstract

The invention relates to the technical field of online education, and particularly discloses an AI-driven online education system and method based on user behavior analysis. And learning efficiency and personalized experience are improved through a data-driven intelligent means. The system comprises the following core modules: a data acquisition module for collecting basic information and learning behavior data of a user; the learning path generation module generates a learning path based on a machine learning algorithm; the recommendation adjustment module optimizes the path according to the user use feedback; and the optimization module evaluates the learning effect through the long-term learning data and further optimizes the path. According to the system, personalized customization and real-time updating of a learning path are realized through dynamic adjustment and a closed-loop optimization mechanism, and the problems that a traditional online education system is insufficient in flexibility and poor in adaptability are effectively solved. The system can obviously improve the learning efficiency, enhances the user participation degree, and provides an intelligent solution for an online education platform.

Description

Technical Field

[0001] The present invention relates to the field of online education technology, and in particular to an AI-driven online education system and method based on user behavior analysis. Background Art

[0002] With the popularization of Internet technology, online education has become an important part of the global education system. However, the current mainstream online education platforms still rely on preset course structures and fixed learning paths, which cannot effectively adapt to the personalized needs of learners. Traditional systems usually adopt a "one-size-fits-all" teaching model, ignoring the obvious differences in students' knowledge base, learning habits and interest preferences. For example, most platforms only provide a linear learning process in chapter order, lacking a dynamic adjustment mechanism, which forces students to repeat the content they have already mastered, or cause learning difficulties due to excessive jumps in difficulty. This model not only reduces learning efficiency, but also limits user participation and satisfaction.

[0003] Although some platforms try to optimize content distribution through simple recommendation algorithms, existing methods mainly rely on explicit user feedback (such as ratings, favorites) or shallow behavioral data (such as clicks), which makes it difficult to deeply explore users' potential learning needs. For example, rule-based recommendation systems cannot respond to dynamic changes in user behavior in real time, while collaborative filtering-based methods face data sparsity and cold start problems. In addition, traditional systems lack a closed-loop evaluation mechanism for learning effects and cannot adjust teaching strategies based on users' real-time performance. These technical bottlenecks have led to a significant gap between the level of personalization of online education and offline education, making it difficult to meet users' urgent needs for efficient and customized learning experiences.

[0004] At present, although some research in the industry has attempted to combine artificial intelligence technology to improve the intelligence level of online education, existing solutions generally have problems such as insufficient technical integration and single application scenarios. For example, some systems only optimize content recommendations and fail to organically combine functions such as learning path planning and dynamic difficulty adjustment. In addition, existing methods mostly rely on offline data analysis and cannot achieve true real-time interaction. Summary of the invention

[0005] In view of this, an embodiment of the present invention provides an AI-driven online education system and method based on user behavior analysis to solve at least one of the above technical problems.

[0006] To achieve the above objectives, in a first aspect, an AI-driven online education system based on user behavior analysis is provided, comprising: Data acquisition module, used to obtain user basic data and user learning data; A preprocessing module, used for preprocessing the user basic data and the user learning data to obtain preprocessed data, and performing feature extraction on the preprocessed data to obtain a number of learning features; A learning path generation module, used to generate a learning path using a machine learning algorithm according to the learning features; A recommendation adjustment module, used to obtain first learning data after the user has learned along the learning path, and to adjust the learning path using a recommendation algorithm according to the first learning data to obtain a recommended learning path; The optimization module is used to obtain second learning data after the user learns according to the recommended learning path, generate a user learning situation evaluation result based on the second learning data, and optimize the recommended learning path based on the user learning situation evaluation result to obtain an optimized learning path.

[0007] In a second aspect, the present invention provides an AI-driven online education method based on user behavior analysis, comprising the following steps: Obtain user basic data and user learning data; Preprocessing the user basic data and the user learning data to obtain preprocessed data, and performing feature extraction on the preprocessed data to obtain a number of learning features; Generate a learning path using a machine learning algorithm according to the learning features; Acquire first learning data after the user has learned along the learning path, and adjust the learning path using a recommendation algorithm according to the first learning data to obtain a recommended learning path; Acquire second learning data after the user has learned along the recommended learning path, generate a user learning situation evaluation result based on the second learning data, optimize the recommended learning path based on the user learning situation evaluation result, and obtain an optimized learning path.

[0008] In a third aspect, an electronic device is provided, comprising: one or more processors; a storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement an AI-driven online education method based on user behavior analysis as described in the second aspect.

[0009] In a fourth aspect, a computer-readable storage medium is provided, on which a computer program is stored, and when the program is executed by a processor, an AI-driven online education method based on user behavior analysis as in the second aspect is implemented.

[0010] The above technical solution has the following beneficial technical effects: Through in-depth analysis of user behavior data, the system can generate customized learning paths based on learners' knowledge level, interest preferences, etc.

[0011] The recommendation adjustment module and optimization module are based on real-time user feedback and long-term learning data respectively to achieve adaptive updates of the learning path, ensuring that the content difficulty and learning progress always match student needs.

[0012] Accurate path recommendations and dynamic optimization reduce ineffective learning time, help students quickly master weak links, and improve learning efficiency.

[0013] The system is continuously optimized, combined with the predictive capabilities of machine learning algorithms, to provide users with a smarter and more flexible learning experience, improving satisfaction and platform stickiness. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The accompanying drawings are used to better understand the present invention and do not constitute an improper limitation of the present invention. Figure 1 It is a structural block diagram of an AI-driven online education system based on user behavior analysis according to an embodiment of the present invention; Figure 2 is a structural block diagram of a preprocessing module in an embodiment of the present invention; Figure 3 is a structural block diagram of a feature extraction unit in an embodiment of the present invention; Figure 4 is a structural block diagram of a learning path generation module in an embodiment of the present invention; Figure 5 is a structural block diagram of a recommended adjustment module in an embodiment of the present invention; Figure 6 is a structural block diagram of an optimization module in an embodiment of the present invention; Figure 7 is a flow chart of an AI-driven online education method based on user behavior analysis according to an embodiment of the present invention; Figure 8 It is a schematic diagram of the structure of a computer system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0015] The following is a description of exemplary embodiments of the present invention in conjunction with the accompanying drawings, including various details of the embodiments of the present invention to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0016] The embodiments of the present invention achieve dynamic generation of personalized learning paths and adaptive adjustment of content recommendations by constructing a deep analysis framework based on user behavior data, thereby solving the technical problems of insufficient flexibility and low degree of personalization in traditional online education systems and providing users with a learning experience that is more in line with their cognitive laws.

[0017] Embodiment 1 like Figure 1 As shown, this embodiment provides an AI-driven online education system based on user behavior analysis, the system comprising: Data acquisition module, used to obtain user basic data and user learning data; A preprocessing module, used for preprocessing the user basic data and the user learning data to obtain preprocessed data, and performing feature extraction on the preprocessed data to obtain a number of learning features; A learning path generation module, used to generate a learning path using a machine learning algorithm according to the learning features; A recommendation adjustment module, used to obtain first learning data after the user has learned along the learning path, and to adjust the learning path using a recommendation algorithm according to the first learning data to obtain a recommended learning path; The optimization module is used to obtain second learning data after the user learns according to the recommended learning path, generate a user learning situation evaluation result based on the second learning data, and optimize the recommended learning path based on the user learning situation evaluation result to obtain an optimized learning path.

[0018] Specifically, the user's basic data is left when the user registers and uses the online education system for the first time, including but not limited to basic information such as name, age, educational background, learning goals, and hobbies. The user fills in this information to establish his or her preliminary personal portrait. In addition, the system can also conduct preliminary analysis based on the user's device information and login habits to further improve the user's preliminary portrait.

[0019] Specifically, the user learning data is obtained in real time through the user terminal device (for example, computer, tablet or mobile phone, etc.), including learning time, learning frequency, interactive behavior (for example, video playback, pause, comment, like and question, etc.), learning progress (for example, course completion rate, completed learning content and homework submission, etc.), test scores, answer accuracy, etc. All collected data are uploaded to the system database in real time through the Application Programming Interface (API).

[0020] Specifically, Figure 2 As shown, in the preprocessing module, it specifically includes: A first preprocessing unit, used to remove duplicate data from the user basic data and the user learning data to obtain first data; A second preprocessing unit, configured to remove outliers in the first data by using an outlier detection algorithm, and fill in the removed outliers to obtain second data; The feature extraction unit is used to process the second data using a feature extraction algorithm to obtain a plurality of learning features.

[0021] Specifically, in the first preprocessing unit, by traversing the user basic data and the user learning data, marking the duplicate data therein, deleting the redundant duplicate data, and retaining only one copy of each data, the accuracy and consistency of the data are ensured. For example, when a user repeatedly performs an operation (such as watching the same video multiple times), the system only retains one record to avoid the impact of multiple calculations on the analysis results. The deduplication algorithm is based on the timestamp, user ID, operation type and other fields of the record to ensure that each record is unique.

[0022] Specifically, in the second preprocessing unit, the anomaly detection algorithm includes the Z-score method and the interquartile range method. Anomalies are data points that are significantly different from normal data patterns. Anomalies are caused by user behavior errors or system failures. For example, if the user's learning time exceeds 24 hours, or the frequency of a certain behavior is too abnormal, the system should automatically identify and mark it as an anomaly. Taking the Z-score detection method as an example, the calculation formula is as follows: ; In the formula, z is the Z value representing the standard score, x is the data point to be detected, μ is the mean of the data, and σ is the standard deviation of the data. If |z|>3, the data point is considered to be an outlier.

[0023] Specifically, in the second preprocessing unit, deletion or filling processing is adopted for abnormal values. The filling processing includes mean filling or adjacent value filling, etc. Mean filling is to calculate the average value of other data and fill the average value to the position of abnormal value. Missing values ​​or data with inconsistent formats are also corrected by filling or deletion processing.

[0024] Specifically, in the feature extraction unit, the learning features are used for subsequent analysis. The goal of feature extraction is to identify key factors that can reflect the user's learning status and preferences. For example, the system can calculate the user's learning frequency, the duration of each learning session, the active learning period, etc. as features to measure learning intensity and participation. Feature extraction can also combine user behavior with historical data for in-depth mining, such as learning ability, interest change trends, etc. Taking the user's learning frequency as an example, the calculation formula for the user's learning frequency is as follows: F=T total / Tperiod ; Where F is the learning frequency, T total T is the total learning time (in hours) of the user in a specified learning cycle. period Indicates the total number of days in a specified learning cycle (e.g., 1 week or 1 month, etc.). The user learning frequency reflects the user's activity level over a period of time.

[0025] Specifically, Figure 3 As shown, in the feature extraction unit, it includes: An initial feature extraction subunit, used to extract a number of initial features of the user from the second data; A clustering analysis subunit, used for selecting a plurality of first features from the plurality of initial features by using a clustering algorithm; A regression analysis subunit, used for selecting a plurality of second features from the plurality of initial features by using a linear regression algorithm; A time series analysis subunit, used for selecting a plurality of third features from the plurality of initial features by using an autoregressive integrated moving average algorithm; The plurality of learning features include a plurality of the first features, a plurality of the second features, and a plurality of the third features.

[0026] Specifically, in the cluster analysis subunit, by dividing users into different groups based on the similarity of their learning behaviors, the system can understand the overall characteristics of users and provide a basis for personalized recommendations. The K-means clustering algorithm is used to divide users into several learning groups based on characteristics such as learning progress, learning intensity, and participation, so as to identify user groups with similar learning behaviors. Each user group can have different learning paths and content recommendations. The objective function of the K-means clustering algorithm is as follows: ; In the formula, μ i is the mean of the ith cluster, is the jth data point belonging to the i-th cluster, m is the total number of data points in the cluster, j is used to represent the index of the specific data point in the i-th cluster, j=1,2,...,m, where m is the total number of data points in cluster i.

[0027] Specifically, in the regression analysis subunit, by establishing the relationship between learning progress and related factors (such as learning time, learning frequency, etc.), the system can predict the user's learning situation at a certain point in time, thereby providing data support for subsequent personalized recommendations. Linear regression is one of the regression analysis methods. The system can predict future learning outcomes through regression analysis of historical learning data. The linear regression formula is as follows: Y=β0+β1x1+β2x2+…+β n x n+ε; Where Y is the predicted learning performance or progress, x1, x2, ..., x n are characteristic variables that affect learning performance, β0, β1, …, β n is the regression coefficient, is the error term. The characteristic variables affecting academic performance represent different factors or attributes used to predict the target variable (e.g., academic performance or learning progress) in the regression model. The characteristic variables affecting academic performance are obtained in the following ways: Method 1 is to extract the following feature variables from user learning data: Learning time, which is the length of time a user spends learning each time, can be used as a feature variable (e.g., x1).

[0028] Learning frequency, which is the frequency at which a user learns within a certain period of time (e.g., x2).

[0029] Job completion status, which indicates the completion status of the user's job, such as the job score, the speed of job submission, etc. (for example, x3).

[0030] A quiz grade, which is the user's quiz score or exam grade in the course (for example, x4).

[0031] Interaction behaviors, such as the frequency of likes, comments, questions, etc. (e.g., x5).

[0032] The second method is to obtain the following characteristic variables based on the user's basic data: Age, which has a certain relationship with learning progress (e.g., x6).

[0033] Educational background, for example, the user's academic level or learning experience (for example, x7).

[0034] Learning objectives, such as a user-set learning goal (for example, x8).

[0035] Method 3 is to obtain the following feature variables based on the features generated by historical data: The learning trend is a trend indicator (eg, x9) generated based on the user's past learning behavior, which may be the rate of change of the user's learning speed, etc.

[0036] Learning participation, which measures the user's activity in the learning process, such as participation in discussions, video viewing time, etc. (e.g., x 10 ).

[0037] The characteristic variables that affect learning performance are obtained by collecting and processing user behavior data, personal data, learning activities and other information, and they are obtained through data exploration and feature engineering in the preparation stage of regression analysis. Finally, these characteristic variables will be passed into the linear regression model as input and compared with the corresponding regression coefficients β1, β2, …, β n work together to predict learning performance or progress Y.

[0038] Specifically, in the time series analysis subunit, time series analysis is used to evaluate the user's learning trend and analyze his learning progress over a period of time. By modeling the time series of the user's learning behavior, the user's future learning behavior and progress can be predicted. This is very useful for dynamically adjusting the learning path and content recommendation, especially for long-term learning projects, where the time series model can be used to evaluate the user's learning bottleneck and potential. The formula of the autoregressive integrated moving average algorithm is as follows: ; Where Y t is the time series data, μ is a constant term, and θ are the autoregressive coefficient and the moving average coefficient respectively, is the error term. p is the autoregressive order, indicating that the model contains observations from the past p time points (e.g., Y t-1 ,Y t-2 ,…,Y t-p ); q is the moving average order, indicating that the model contains error terms for the past q time points (e.g. ). The subscripts t-1, t-2, ... represent the lag term of the time series, for example, Y t-1 is the previous observation at the current time point t, is the prediction error at the previous time point. reflects the weight of the autoregressive term on the current value, θ1, θ2, …, θ q Reflects the correction effect of the moving average term on the current value. Error term is a white noise sequence. This model captures the long-term trend and periodic fluctuation of the time series by combining historical data and historical errors, and is suitable for predicting the dynamic changes of user learning behavior.

[0039] Specifically, Figure 4 As shown, in the learning path generation module, it specifically includes: A learning goal acquisition unit, used to acquire a user's learning goal; A rule setting unit, used for setting path generation constraints according to the user learning goal; The first path generation unit is used to generate a learning path by using a machine learning algorithm according to the learning features, the user learning goals and the path generation constraints.

[0040] Specifically, the user learning goal is filled in by the user when registering in the online education system, or automatically generated according to the user's learning behavior and feedback. The path generation constraints are set according to the user's learning goals and user learning data. For example, the system will determine which knowledge points need to be learned first and which can be learned later based on the course structure. It will also define how to skip related content and recommend more difficult materials when the user has a good grasp of a certain learning module; when the user makes slow progress in a certain learning module, how to extend the learning time or provide more review materials.

[0041] Specifically, in the first path generation unit, machine learning algorithms, such as deep learning and reinforcement learning, are applied to dynamically adjust the learning path. The deep learning model predicts the user's future learning needs by analyzing the user's learning characteristics, such as academic performance, participation, etc., and generates a personalized learning path accordingly. Reinforcement learning adjusts the learning path according to the user's learning results through a feedback mechanism. The system optimizes each decision in the learning path based on the reward and punishment mechanism, thereby continuously improving the recommendation strategy of learning content to ensure that the learning path can meet the user's learning goals.

[0042] Specifically, the first path generation unit includes: A deep learning subunit, used for inputting the learning features into a pre-trained deep learning model, taking the user learning goal as a goal, taking the path generation constraint as a constraint condition, and obtaining an initial path; A reinforcement learning subunit, used to optimize the initial path using a reinforcement learning algorithm to obtain a reinforcement path; An evaluation subunit, used to obtain reinforcement learning data after the user learns according to the reinforcement path, and generate an evaluation result of the reinforcement path according to the reinforcement learning data; an optimization subunit, configured to optimize the reinforcement path according to the evaluation result of the reinforcement path to obtain a learning path; The evaluation result of the reinforcement path includes several evaluation items, and the several evaluation items include: current learning progress, homework grades, test scores and learning participation.

[0043] Specifically, in the deep learning subunit, learning features (such as historical scores, learning participation, learning time, etc.) are input into a pre-trained deep learning model, and the deep learning model learns these learning features to identify the user's learning patterns and behavioral characteristics. For example, the system evaluates the user's mastery of certain knowledge points through the user's past test scores and homework submissions, and predicts the user's possible needs in future learning. The deep learning model generates a set of prediction values ​​to indicate the user's learning trends in different knowledge points or modules, thereby providing a basis for subsequent learning path adjustments. The deep learning model analyzes the user's learning characteristics, predicts the user's future learning needs, and generates preliminary learning path adjustment suggestions. The expression of the deep learning model is as follows: c=o (W·h+b); Among them, c is the predicted learning performance or progress, h is the user's learning characteristics (for example, historical grades, homework grades, participation, etc.), W is the weight matrix of the deep learning model, b is the bias term, and o is the activation function (for example, a sigmoid function or a ReLU function).

[0044] Specifically, in the reinforcement learning subunit, the learning path is further optimized through the reinforcement learning algorithm. Reinforcement learning continuously adjusts strategies through interaction with the environment, and its goal is to use reward and punishment mechanisms to enable each decision in the learning path to drive users to better achieve their learning goals. The system evaluates the effectiveness of the learning path adjustment based on user feedback at various learning stages (such as homework grades, test performance, participation, etc.), and uses this feedback to optimize each decision in the learning path. For example, if the user's learning progress in a certain module is slow, the system will increase the learning time of the module or recommend more review materials. Through the reinforcement learning algorithm, the system continuously adjusts each decision in the learning path to achieve the best learning effect and ensure that the learning path can better adapt to the user's learning progress and needs. The expression of the reinforcement learning algorithm is as follows: ; In the formula, For state s t Take action a t The value or expected cumulative reward value (Q value), r t To perform action a t The immediate reward obtained after α is the learning rate, γ is the discount factor, and s t+1 is the next state, and a′ is the next action. In the Q learning algorithm in the reinforcement learning algorithm, the formula Indicates that in state s t+1The maximum Q value obtained among all possible actions a′ under the condition . The Q value here represents the expected cumulative reward obtained by taking a certain action in a certain state.

[0045] Specifically, the evaluation subunit includes several evaluation items, including: current learning progress, homework grades, test scores, and learning participation. If the evaluation results of the reinforcement path show that the user has learning difficulties in certain content (for example, low grades or frequent learning interruptions), the system will adjust the learning path. For example, adding review materials for certain modules or adjusting the learning order of the path so that users can better master what they have learned. The evaluation process ensures that the path adjustment can effectively improve the user's learning outcomes. By evaluating the effectiveness of the learning path, the personalized learning path is further optimized to ensure that the recommended content can better meet the user's learning needs and goals.

[0046] Specifically, in the optimization subunit, when a user completes a learning module or passes a test, the reinforcement path is optimized according to the evaluation result of the reinforcement path, and the order and difficulty of the recommended learning content can be adjusted to ensure that the learning content at each stage meets the current needs of the user. If it is found that the user's learning progress in a certain module is lagging behind, the system will adjust the path, recommend more learning materials or increase the learning time to help the user catch up.

[0047] Specifically, Figure 5 As shown, in the recommendation adjustment module, it includes: A first acquisition unit, used to acquire first learning data after the user has learned according to the learning path; A recommendation unit, configured to use a recommendation algorithm according to the learning path and the first learning data to add a number of learning contents to the learning path to obtain a first path; A monitoring unit, configured to obtain first monitoring learning data after the user learns the additional learning content along the first path; The first output unit is used to adjust the first path according to the first monitoring learning data to obtain a recommended learning path.

[0048] Specifically, the learning content in the recommendation unit includes teaching videos, reading materials, exercises and tests. These materials are set according to the user's learning progress, knowledge mastery and goals. For example, when the user is in a certain learning module, the system will recommend related teaching videos to help the user deepen his understanding; when the system finds that the user has mastered a certain knowledge point and passed the relevant test, the system will recommend more difficult content. The system ensures that the recommended content not only meets the requirements of the learning path, but also meets the user's current learning needs, helping them to achieve their learning goals more effectively.

[0049] Specifically, in the monitoring unit, the first monitoring learning data includes the ratio of the number of completed learning modules to the total number of modules, the statistics of learning time, and the user's interaction with the learning materials (for example, video viewing time, exercise answering, etc.). At the same time, the system will collect the user's learning feedback, especially test scores, homework scores, and participation. This information will help the system determine whether the user has completed the study as planned, whether it has encountered difficulties, and the user's understanding of certain content, so as to make adjustments for the next recommendation.

[0050] Specifically, in the first output unit, the recommended content is adaptively adjusted based on the user's learning progress and feedback. If the user has mastered the relevant knowledge in the current learning module and passed the relevant test, the system will skip this content and recommend more challenging learning materials. On the contrary, if the user performs poorly in a certain learning module or often interrupts learning, the system will automatically recommend more review materials or more basic learning content. The system adjusts the difficulty and content of the recommendation based on each user's learning feedback (such as low scores, learning interruptions, etc.) to ensure that each user can obtain the most appropriate resources in their own learning state. The recommendation algorithm is continuously optimized through a real-time feedback mechanism. Each interaction between the user and the recommended content (such as test scores, homework completion, etc.) will provide new data for the system, and the system will analyze and optimize the recommendation strategy based on these data. If the user fails to achieve the expected results in a certain learning module, the system will recommend more relevant review materials or adjust the recommended learning path; if the user feedback indicates that some content is too lengthy or too difficult, the system will also make corresponding adjustments based on these feedbacks. In addition, the system will make dynamic recommendations based on user engagement (such as frequent skipping of content or too short a study time) to ensure that the learning content is more attractive and targeted. After completing the recommendation and adjustment, the system will generate a learning report and provide it to users and administrators. The learning report contains information such as the user's learning progress, performance feedback, learning time, and engagement, which helps users understand their learning status and provides data support for subsequent optimization of the system. Administrators can evaluate the user's learning situation through the report and provide additional support as needed. The report can also help the system identify which learning content is popular or unpopular, thereby optimizing the algorithm for content recommendation and learning path generation.

[0051] Specifically, the recommendation algorithms include content-based recommendation algorithms, collaborative filtering algorithms, matrix decomposition algorithms, deep learning recommendation algorithms and hybrid recommendation algorithms.

[0052] Specifically, the user basic data and user learning data generate an initial user portrait, obtain the third learning data after the user learns according to the optimized learning path, and update the initial user portrait according to the third learning data to obtain a dynamically updated user portrait. The similarity calculation formula is as follows: ; Where A and B are the feature vectors of the learning content.

[0053] Specifically, the collaborative filtering recommendation algorithm generates recommendations by analyzing the similarity of interests between users or between items. The algorithm is based on an assumption: if user u and user v have similar interests in the same item in the past, their interests in future items should also be similar. Collaborative filtering is divided into two methods: user-based and item-based, and the calculation method uses cosine similarity or Pearson correlation coefficient. The matrix decomposition algorithm identifies user interests and item features by decomposing the user-item rating matrix into multiple small matrices. The matrix decomposition algorithm can effectively extract the implicit relationships between users and items, and use these relationships to generate accurate recommendations, especially when the data is sparse. Decompose the original matrix R into a user latent feature matrix, a singular value matrix Σ, and an item latent feature matrix V T , the formula is: R≈U∑V T .

[0054] Specifically, the deep learning recommendation algorithm models user behavior data through a neural network to capture complex nonlinear relationships. The user's historical behavior, item features, and contextual information can be used as inputs to the network, and the deep neural network will learn the potential associations between users and items. Through the neural network, the system can generate the most appropriate recommended content for each user. The hybrid recommendation algorithm combines the advantages of different algorithms and can overcome the limitations of a single algorithm. By combining multiple recommendation strategies, the system can analyze user behavior and interests from multiple dimensions, thereby improving the accuracy and diversity of recommendations. Hybrid recommendation algorithms include weighted mixing, cascade mixing, feature-level mixing, or random selection. The weighted mixing assigns weights to each algorithm based on the performance of different algorithms, and then weights and merges their recommendation results. The weighted mixing method can effectively integrate the advantages of different algorithms to ensure that recommendations are more personalized. The cascade mixing first uses a recommendation algorithm (e.g., content-based recommendation) and passes its results as input to another algorithm (e.g., collaborative filtering). This ensures that the output of the previous algorithm is further optimized by the next algorithm, which is particularly suitable when user data is sparse. The feature-level blending combines the recommendation features from different algorithms to form a unified recommendation feature set, and uses a machine learning model (e.g., decision tree, random forest, etc.) for final prediction. This makes it possible to use the features of multiple algorithms at the same time to generate more comprehensive recommendations. The random selection generates a recommendation list based on multiple recommendation algorithms, and then selects one of the results randomly or according to a specific strategy. This method is suitable for introducing a certain degree of diversity and creativity in the recommendation results.

[0055] Specifically, Figure 6 As shown, in the optimization module, it includes: A second acquisition unit, used to acquire second learning data after the user learns according to the recommended learning path; An evaluation unit, configured to generate a user learning situation evaluation result from three aspects: knowledge mastery, learning progress, and learning initiative according to the second learning data; An optimization unit, configured to optimize the recommended learning path according to the evaluation result of the user's learning situation to obtain an optimized learning path; When executing the optimization unit, if the knowledge mastery level is greater than a first threshold, the knowledge points currently involved in the recommended learning path are reduced, and difficult knowledge points are added; If the knowledge mastery level is less than or equal to a preset first threshold, the knowledge points in the recommended learning path are not modified; The difficult knowledge points meet the following requirements: the error rate of answering questions exceeds a predetermined error rate threshold and / or the learning time exceeds a preset multiple of the benchmark time; If the knowledge mastery level is less than a preset second threshold or the learning progress is less than a preset third threshold, then increasing the learning time of each knowledge point in the recommended learning path; If the knowledge mastery level is greater than or equal to a preset second threshold, and the learning progress is greater than or equal to a preset third threshold, the learning time of each knowledge point in the recommended learning path remains unchanged; If the learning initiative is less than a fourth threshold, the user is regularly reminded to learn.

[0056] Specifically, the first threshold, the second threshold and the third threshold are artificially set empirical values.

[0057] Specifically, the second learning data includes homework scores, test results, and learning interactions. Homework scores reflect the user's mastery of the knowledge learned, and test results provide a quantitative basis for the user's mastery of a certain knowledge point. Learning interactions, such as the user's viewing time of video content, homework submission, forum questions, etc., reflect the user's participation and interest in the learning content. All of these data are collected in real time by the system and transmitted to the feedback and optimization module, providing data support for the subsequent optimization of learning paths and recommended content.

[0058] Specifically, in the evaluation unit, the knowledge mastery level is obtained by analyzing the user's homework scores and test results. For example, if the test score of a certain knowledge point is low, it means that the user has difficulty mastering this point. The learning progress of the user is obtained by monitoring the user's learning progress, checking whether the set learning path is followed, and whether there are any missing modules or content. The user's learning initiative is evaluated based on the user's interaction (for example, learning time or participation frequency, etc.).

[0059] Specifically, the knowledge mastery is mainly evaluated by analyzing the user's homework scores and test results. The method used includes calculating the correct answer rate, completion rate and retest score change trend of each knowledge point. The following formula can be used to quantify the knowledge mastery: ; Among them, K is the degree of knowledge mastery, S i is the score of the ith test or assignment (between 0 and 1, or a percentage), and n is the number of assignments or tests.

[0060] In addition, weight adjustments can be introduced to make corrections based on the importance or difficulty of assignments and quizzes. For example, the most recent quizzes better reflect the current level of mastery, so a time decay factor can be introduced, and the corresponding formula is as follows: ; Among them, w i is the time decay weight of the ith test or assignment, defined as: ; T current is the current time, T i is the time of the ith test, and λ is the decay coefficient. In this way, the changing trend and current status of knowledge mastery can be dynamically evaluated.

[0061] Specifically, the learning progress can be monitored by monitoring the user's position and progress speed in the learning path. The method used is to compare the deviation between the user's actual learning path and the recommended path or expected path set by the system. The formula is as follows: ; Among them, P represents the learning progress, C actual Indicates the number of knowledge points or modules that the user has completed, C expected Indicates the number of knowledge points or modules recommended or expected to be completed by the system. If P=1, it means that the user's progress is consistent with the recommended path; when P<1, it means that the user's learning speed is slower than expected, there are obstacles to understanding or external interference; when P>1, it means that the user's learning speed is faster than expected, there is a situation of leapfrogging learning or advanced mastery.

[0062] Specifically, learning initiative is evaluated through user interaction behaviors, including learning time, login frequency, access depth, Q&A, and participation. Evaluation indicators include: learning time proportion, learning frequency, and interaction depth.

[0063] The learning time ratio is the ratio of effective learning time to total time within a certain time window (such as a week or a day). The formula is as follows: ; Among them, T learning is the effective learning time, T total is the total time in this time window.

[0064] Learning frequency refers to the number of times a user learns or logs in within a specific period (such as a week). The formula is as follows: ; Among them, N active is the number of learning times or login times of a user in a specific period, and T is the time period.

[0065] Interaction depth refers to the number of interactions (such as asking questions, checking hints, and redoing questions) performed by users during the learning process. The formula is as follows: ; Where: I i is the number of interactions in the ith learning session, and n is the total number of interactions.

[0066] The final learning initiative can be comprehensively evaluated by weighted average or normalization: A=w1A1+w2A2+w3A3; Among them, w1, w2, and w3 are the weights of different interactive behaviors, reflecting the importance that the learning system attaches to different types of interactions.

[0067] Specifically, in the optimization unit, the results are evaluated according to the user's learning situation. If the user has a good grasp of certain knowledge points, the system will skip these contents and recommend more difficult learning materials. If the system finds that the user is making slow progress in certain modules or performs poorly in the test, the system will recommend more review materials, tutoring content or provide additional learning support. The system can also ensure that the user has enough time to master each knowledge point by dynamically adjusting the learning time of each module. Such adjustments can not only improve learning efficiency, but also help users choose the most suitable learning resources according to their current learning status. After each adjustment of the learning path and recommended content, the system will continue to collect the user's learning feedback and further optimize the learning path and recommendation strategy through new feedback data. Through this closed-loop feedback mechanism, the system can achieve continuous self-optimization to ensure that the recommended content and learning path always meet the user's learning progress and needs. Specifically, the system will update the learning path and recommended content in real time to ensure that the user always receives the most suitable materials for his current learning stage. The system continuously optimizes the personalized learning path and recommended content as the user's learning progress changes through continuous collection, analysis and feedback, and maximizes the learning effect. It can adjust the personalized learning path and recommended content in real time according to the user's learning feedback. Each adjustment and optimization is based on the user's mastery, learning progress, participation and other data to ensure that the learning content is highly consistent with the user's learning status. By forming a closed-loop feedback mechanism, the system can continuously optimize the learning path and recommendation strategy, improve the accuracy and adaptability of personalized learning, and effectively promote the user's learning process and enhance the user experience.

[0068] Specifically, the AI-driven online education system based on user behavior analysis also includes a user portrait updating unit, which is used to generate an initial user portrait according to the user basic data and the user learning data, obtain the third learning data after the user learns according to the optimized learning path, and update the initial user portrait according to the third learning data to obtain a dynamically updated user portrait. After each learning cycle, the system dynamically updates the user portrait based on the user's long-term learning behavior data and feedback information. The updated user portrait not only reflects the user's current learning status, but also can be predicted in combination with the user's changing trends, interest development and learning habits. The updated user portrait will provide support for the adjustment of the next round of learning paths and the optimization of content recommendation strategies. At the end of each learning cycle, the system will collect and integrate the user's learning behavior data and feedback information during the cycle. These data usually include the user's learning progress, learning time, learning frequency, learning interaction data (such as questions, comments, likes, etc.), performance feedback, and changes in user interests. Through these data, the system can fully understand the user's performance and preferences in the learning process, and provide a basis for subsequent user portrait updates and learning path adjustments. All these data will be transmitted to the user portrait construction and update module for further analysis and processing. The system identifies users' learning trends, interest changes, and behavior patterns by analyzing their long-term learning behavior data and feedback information.

[0069] The system will focus on analyzing the following aspects: changes in the user's learning ability, such as evaluating their learning progress and ability improvement through homework grades and test scores. Changes in user interests, the system predicts the development trend of their interests by analyzing the user's attention to different learning content and the length of time they study; changes in the user's learning habits, by monitoring the length of study, frequency of study, and time period of study, to understand whether the user's learning habits have changed. These analysis results provide a basis for updating user portraits and adjusting personalized learning paths.

[0070] Based on the analysis of users' long-term learning behaviors and feedback, the system will dynamically update the user portrait. The updated user portrait not only reflects the user's current learning status, but also combines its long-term learning trajectory and interest change trends. The updated content includes: the user's mastery of various knowledge points, especially the strong interest in certain knowledge areas or areas that need additional learning; the user's learning progress and goals, the system adjusts the learning path for the next stage according to the user's current learning status; the user's interest preferences, the system updates the user's preferred learning areas according to the user's learning content and feedback; and the user's learning behavior pattern, the system monitors changes in the user's learning habits, such as learning time, frequency, etc. These updates make the user portrait more accurate and can better adapt to the user's personalized learning needs. The updated user portrait will directly affect the adjustment of the learning path and the optimization of the content recommendation strategy in the next stage.

[0071] Specifically, the system will adjust the learning path based on the new user profile to ensure that the user's learning content matches their current learning needs. If the user has a good grasp of a certain knowledge point, the system will automatically skip that content and recommend more difficult learning materials; if the user progresses slowly on certain modules, the system will recommend more relevant review materials or tutoring content. The updated user profile will also optimize the recommendation strategy. The system will recommend the learning materials that best meet the user's current needs based on the user's changes in interests and learning habits. For example, if the user shows a high interest in a certain field, the system will prioritize related learning content.

[0072] Embodiment 2 like Figure 7 As shown, an AI-driven online education method based on user behavior analysis includes the following steps: S10: Obtain user basic data and user learning data; S20: preprocessing the user basic data and the user learning data to obtain preprocessed data, and extracting features from the preprocessed data to obtain a number of learning features; S30: Generate a learning path using a machine learning algorithm according to the learning features; S40: Acquire first learning data after the user has learned along the learning path, and adjust the learning path using a recommendation algorithm according to the first learning data to obtain a recommended learning path; S50: Obtain second learning data after the user has learned along the recommended learning path, generate a user learning situation evaluation result based on the second learning data, and optimize the recommended learning path based on the user learning situation evaluation result to obtain an optimized learning path.

[0073] Specifically, in step S20, the following steps are included: S21: removing duplicate data from the user basic data and the user learning data to obtain first data; S22: using an anomaly detection algorithm to remove outliers in the first data, and filling the removed outliers to obtain second data; S23: Process the second data using a feature extraction algorithm to obtain a number of learning features.

[0074] Specifically, in step S30, the following steps are included: S31: Obtaining user learning goals; S32: Setting path generation constraints according to the user learning goal; S33: Generate a learning path using a machine learning algorithm according to the learning features, the user learning goals and the path generation constraints.

[0075] Specifically, in step S33, the following steps are included: S331: Input the learning features into a pre-trained deep learning model, take the user learning goal as the goal, take the path generation constraint as the constraint condition, and obtain an initial path; S332: Optimizing the initial path using a reinforcement learning algorithm to obtain a reinforcement path; S333: Acquire reinforcement learning data after the user has learned the reinforcement path, and generate an evaluation result of the reinforcement path according to the reinforcement learning data; S334: Optimize the reinforcement path according to the evaluation result of the reinforcement path to obtain a learning path.

[0076] Specifically, in step S40, the following steps are included: S41: Acquire first learning data after the user has learned according to the learning path; S42: using a recommendation algorithm according to the learning path and the first learning data, adding a number of learning contents to the learning path to obtain a first path; S43: Acquire first monitoring learning data after the user learns the additional learning content according to the first path; S44: adjusting the first path according to the first monitored learning data to obtain a recommended learning path.

[0077] Specifically, in step S50, the following steps are included: S51: Acquire second learning data after the user has learned according to the recommended learning path; S52: generating a user learning situation evaluation result based on the second learning data from three aspects: knowledge mastery, learning progress and learning initiative; S53: Optimizing the recommended learning path according to the user learning situation evaluation result to obtain an optimized learning path; When executing step S53, if the knowledge mastery level is greater than a first threshold, the knowledge points currently involved in the recommended learning path are reduced, and difficult knowledge points are added; If the knowledge mastery level is less than or equal to a preset first threshold, the knowledge points in the recommended learning path are not modified; Among them, difficult knowledge points refer to knowledge points with an error rate of more than 50% for the answers to the knowledge points or knowledge points where the learning time for the knowledge points exceeds a preset multiple (e.g., 1.5 times) of the benchmark time. The benchmark time can be the average learning time. Alternatively, difficult knowledge points refer to knowledge points that correspond to multiple interdisciplinary subjects or cover multiple subjects.

[0078] If the knowledge mastery level is less than a second threshold or the learning progress is less than a third threshold, increasing the learning time of each knowledge point in the recommended learning path; If the knowledge mastery level is greater than or equal to a preset second threshold, and the learning progress is greater than or equal to a preset third threshold, the learning time of each knowledge point in the recommended learning path remains unchanged; If the learning initiative is less than a fourth threshold, the user is regularly reminded to learn.

[0079] Specifically, the first threshold, the second threshold and the third threshold are artificially set empirical values.

[0080] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0081] An embodiment of the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-mentioned AI-driven online education methods based on user behavior analysis.

[0082] If the integrated module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. Of course, there are other forms of readable storage media, such as quantum memory, graphene memory, etc. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media does not include electrical carrier signals and telecommunication signals.

[0083] The present invention also provides an electronic device. The electronic device of an embodiment of the present invention includes: one or more processors; a storage device for storing one or more programs, and when the one or more programs are executed by the one or more processors, the one or more processors implement an AI-driven online education method based on user behavior analysis provided by the present invention.

[0084] Reference below Figure 8 , which shows a schematic diagram of the structure of a computer system 800 of an electronic device suitable for implementing an embodiment of the present invention. Figure 8 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0085] like Figure 8As shown, the computer system 800 includes a central processing unit (CPU) 801, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 802 or a program loaded from a storage part 808 to a random access memory (RAM) 803. In the RAM 803, various programs and data required for the operation of the computer system 800 are also stored. The CPU 801, the ROM 802, and the RAM 803 are connected to each other through a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0086] The following components are connected to the I / O interface 805: an input section 806 including a keyboard, a mouse, etc.; an output section 807 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 808 including a hard disk, etc.; and a communication section 809 including a network interface card such as a LAN card, a modem, etc. The communication section 809 performs communication processing via a network such as the Internet. A drive 810 is also connected to the I / O interface 805 as needed. A removable medium 811, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 810 as needed so that a computer program read therefrom is installed into the storage section 808 as needed.

[0087] In particular, according to the embodiments disclosed in the present invention, the process described in the main step diagram above can be implemented as a computer software program. For example, an embodiment of the present invention includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program contains program code for executing the method shown in the main step diagram. In the above embodiment, the computer program can be downloaded and installed from the network through the communication part 809, and / or installed from the removable medium 811. When the computer program is executed by the central processing unit 801, the above functions defined in the system of the present invention are executed.

[0088] It should be noted that the computer-readable medium shown in the present invention may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present invention, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer readable signal media may also be any computer readable medium other than computer readable storage media, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0089] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the above-mentioned module, program segment or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0090] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may occur depending on design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An AI-driven online education system based on user behavior analysis, characterized in that: include: Data acquisition module, used to obtain user basic data and user learning data; A preprocessing module, used for preprocessing the user basic data and the user learning data to obtain preprocessed data, and performing feature extraction on the preprocessed data to obtain a number of learning features; A learning path generation module, used to generate a learning path using a machine learning algorithm according to the learning features; A recommendation adjustment module, used to obtain first learning data after the user has learned along the learning path, and to adjust the learning path using a recommendation algorithm according to the first learning data to obtain a recommended learning path; The optimization module is used to obtain second learning data after the user learns according to the recommended learning path, generate a user learning situation evaluation result based on the second learning data, and optimize the recommended learning path based on the user learning situation evaluation result to obtain an optimized learning path.

2. The AI-driven online education system based on user behavior analysis according to claim 1, characterized in that: The preprocessing module includes: A first preprocessing unit, used to remove duplicate data from the user basic data and the user learning data to obtain first data; A second preprocessing unit, configured to remove outliers in the first data by using an outlier detection algorithm, and fill in the removed outliers to obtain second data; The feature extraction unit is used to process the second data using a feature extraction algorithm to obtain a plurality of learning features.

3. The AI-driven online education system based on user behavior analysis according to claim 1, characterized in that: The learning path generation module includes: A learning goal acquisition unit, used to acquire a user's learning goal; A rule setting unit, used for setting path generation constraints according to the user learning goal; The first path generation unit is used to generate a learning path by using a machine learning algorithm according to the learning features, the user learning goals and the path generation constraints.

4. The AI-driven online education system based on user behavior analysis according to claim 1, characterized in that: The recommendation adjustment module includes: A first acquisition unit, used to acquire first learning data after the user has learned according to the learning path; A recommendation unit, configured to use a recommendation algorithm according to the learning path and the first learning data to add a number of learning contents to the learning path to obtain a first path; A monitoring unit, configured to obtain first monitoring learning data after the user learns the additional learning content along the first path; The first output unit is used to adjust the first path according to the first monitoring learning data to obtain a recommended learning path.

5. The AI-driven online education system based on user behavior analysis according to claim 1, characterized in that: The optimization module includes: A second acquisition unit, used to acquire second learning data after the user learns according to the recommended learning path; An evaluation unit, configured to generate a user learning situation evaluation result from three aspects: knowledge mastery, learning progress, and learning initiative according to the second learning data; An optimization unit, configured to optimize the recommended learning path according to the evaluation result of the user's learning situation to obtain an optimized learning path; When executing the optimization unit, if the knowledge mastery level is greater than a preset first threshold, the knowledge points currently involved in the recommended learning path are reduced and difficult knowledge points are added; If the knowledge mastery level is less than or equal to a preset first threshold, the knowledge points in the recommended learning path are not modified; The difficult knowledge points meet the following requirements: the error rate of answering questions exceeds a predetermined error rate threshold and / or the learning time exceeds a preset multiple of the benchmark time; If the knowledge mastery level is less than a preset second threshold or the learning progress is less than a preset third threshold, then increasing the learning time of each knowledge point in the recommended learning path; If the knowledge mastery level is greater than or equal to a preset second threshold, and the learning progress is greater than or equal to a preset third threshold, the learning time of each knowledge point in the recommended learning path remains unchanged; If the learning initiative is less than a fourth threshold, the user is regularly reminded to learn.

6. The AI-driven online education system based on user behavior analysis according to claim 1, characterized in that: It also includes a user portrait updating unit, which is used to generate an initial user portrait based on the user basic data and the user learning data, obtain the third learning data of the user after learning according to the optimized learning path, update the initial user portrait according to the third learning data, and obtain a dynamically updated user portrait.

7. The AI-driven online education system based on user behavior analysis according to claim 3, characterized in that: The first path generation unit includes: A deep learning subunit, used for inputting the learning features into a pre-trained deep learning model, taking the user learning goal as the goal, taking the path generation constraint as the constraint condition, and obtaining an initial path; A reinforcement learning subunit, used to optimize the initial path using a reinforcement learning algorithm to obtain a reinforcement path; An evaluation subunit, used to obtain reinforcement learning data after the user learns according to the reinforcement path, and generate an evaluation result of the reinforcement path according to the reinforcement learning data; an optimization subunit, configured to optimize the reinforcement path according to the evaluation result of the reinforcement path to obtain a learning path; The evaluation result of the reinforcement path includes several evaluation items, and the several evaluation items include: current learning progress, homework grades, test scores and learning participation.

8. The AI-driven online education system based on user behavior analysis according to claim 4, characterized in that: The recommendation algorithm is a content-based recommendation algorithm, which obtains the user's historical learning materials, calculates the similarity between the user's historical learning materials and the learning materials to be added, and selects the learning materials to be added with the highest similarity as the learning content.

9. The AI-driven online education system based on user behavior analysis according to claim 2, characterized in that: The feature extraction unit includes: An initial extraction subunit, used to extract a number of initial features of the user from the second data; A clustering analysis subunit, configured to select a plurality of first features from the plurality of initial features using a clustering algorithm; A regression analysis subunit, used for selecting a plurality of second features from the plurality of initial features by using a linear regression algorithm; A time series analysis subunit, used for selecting a plurality of third features from the plurality of initial features by using an autoregressive integrated moving average algorithm; The plurality of learning features include a plurality of the first features, a plurality of the second features, and a plurality of the third features.

10. An AI-driven online education method based on user behavior analysis, characterized in that: include: S10: Obtain user basic data and user learning data; S20: preprocessing the user basic data and the user learning data to obtain preprocessed data, and extracting features from the preprocessed data to obtain a number of learning features; S30: Generate a learning path using a machine learning algorithm according to the learning features; S40: obtaining first learning data after the user has learned along the learning path, and adjusting the learning path by using a recommendation algorithm according to the first learning data to obtain a recommended learning path; S50: Obtain second learning data after the user has learned along the recommended learning path, generate a user learning situation evaluation result based on the second learning data, and optimize the recommended learning path based on the user learning situation evaluation result to obtain an optimized learning path.

Images