Method and system for building big data model of learning service platform based on artificial intelligence

Abstract

The present invention relates to the field of model construction technology, and in particular to a method and system for building a big data model of a learning service platform based on artificial intelligence. The method comprises the following steps: based on a preset learning service platform, multi-dimensional learner data collection is performed on learners to obtain a multi-modal data set of learners; based on the multi-modal data set of learners, a learner portrait recognition model is constructed for learners to obtain a learner intelligent portrait recognition model; based on a preset learning service platform, multi-dimensional teacher data collection is performed on teachers to obtain a teacher multi-modal data set; based on the multi-modal data set of teachers, a teacher portrait recognition model is constructed for teachers to obtain a teacher intelligent portrait recognition model; based on a preset learning service platform, time series data collection of the teaching process of learners and teachers is performed to obtain a teaching process time series data set. The present invention significantly improves the accuracy and practicality of learner and teacher portraits.

Description

Technical Field

[0001] The present invention relates to the field of model building technology, and in particular to a method and system for building a big data model of a learning service platform based on artificial intelligence. Background Art

[0002] One of the existing technical challenges is how to achieve in-depth understanding and analysis of learners' behaviors and preferences in a large-scale learning environment, and how to provide customized teaching support to teachers.

[0003] Furthermore, educational platforms face challenges in processing and analyzing large amounts of multimodal data, including learners’ course choices, learning paths, interaction patterns, and teachers’ teaching content and styles. Existing data analysis methods often fail to fully tap the potential value of these data, resulting in insufficient educational decision-making support capabilities. In addition, the lack of real-time monitoring and early warning mechanisms during the teaching process makes it difficult for educators to promptly discover and respond to abnormal situations in the teaching process, which greatly affects the improvement of teaching quality and learning outcomes. At the same time, the evaluation and diagnosis of educational outcomes is also a difficult problem in existing technologies. Full-cycle educational effectiveness monitoring requires comprehensive consideration of multiple dimensions such as learners’ learning behaviors, outcomes, feedback, and teachers’ teaching behaviors and evaluations, but there is currently a lack of a method that can comprehensively and continuously evaluate educational outcomes. This limits educators’ ability to continuously optimize the educational process and also affects the effective formulation of teaching strategies. Summary of the invention

[0004] Based on this, it is necessary for the present invention to provide a method and system for building a big data model of a learning service platform based on artificial intelligence to solve at least one of the above technical problems.

[0005] To achieve the above purpose, a method for constructing a big data model of a learning service platform based on artificial intelligence includes the following steps:

[0006] Step S1: Based on the preset learning service platform, multi-dimensional learner data is collected for learners to obtain a multimodal data set of learners; based on the multimodal data set of learners, a learner portrait recognition model is constructed for learners to obtain a learner intelligent portrait recognition model;

[0007] Step S2: Based on the preset learning service platform, multi-dimensional teacher data is collected for teachers to obtain a teacher multimodal data set; based on the teacher multimodal data set, a teacher portrait recognition model is constructed for teachers to obtain a teacher intelligent portrait recognition model;

[0008] Step S3: Based on the preset learning service platform, the teaching process time series data of learners and teachers are collected to obtain a teaching process time series data set; based on the teaching process time series data set, a teaching early warning model is constructed for the learning-teaching interaction process to obtain a teaching abnormality intelligent early warning model;

[0009] Step S4: Based on the preset learning service platform, the full-cycle data of educational effects of learners and teachers are collected to obtain a full-cycle data set of educational effects; educational effect labels are extracted from the full-cycle data set of educational effects to obtain an educational effect label set; based on the full-cycle data set of educational effects and according to the educational effect label set, an educational effect diagnosis model is constructed for the educational process to obtain an educational effect diagnosis model;

[0010] Step S5: Integrate the learner intelligent portrait recognition model, the teacher intelligent portrait recognition model, the teaching abnormality intelligent early warning model and the education effect diagnosis model to obtain a big data model of the learning service platform.

[0011] The present invention significantly improves the accuracy and practicality of learner and teacher portraits through multi-dimensional data analysis capabilities, and provides richer data support for personalized learning services. Through the introduction of a real-time teaching early warning mechanism, abnormal situations in the teaching process can be effectively identified and responded to, greatly improving the teaching quality and learning effects. Through the implementation of full-cycle education effect evaluation, more real-time and comprehensive education monitoring and analysis are provided to continuously optimize the education process. Through the realization of data fusion and model integration, information silos are broken and the platform's in-depth understanding and decision-making support capabilities are enhanced. Through the ability of intelligent and adaptive learning, the model can quickly respond to changes in the education environment and improve the adaptability and foresight of education services. The provision of personalized education experience enhances learners' personalized learning experience through in-depth analysis of learners' behaviors and preferences, customized learning resources and paths.

[0012] Preferably, the present invention also provides a learning service platform big data model construction system based on artificial intelligence, which is used to execute the learning service platform big data model construction method based on artificial intelligence as described above, and the learning service platform big data model construction system based on artificial intelligence includes:

[0013] The learner portrait recognition model construction module is used to collect multi-dimensional learner data of learners based on the preset learning service platform to obtain a learner multimodal data set; based on the learner multimodal data set, a learner portrait recognition model is constructed for the learner to obtain a learner intelligent portrait recognition model;

[0014] The teacher portrait recognition model construction module is used to collect multi-dimensional teacher data on teachers based on the preset learning service platform to obtain a teacher multimodal data set; based on the teacher multimodal data set, a teacher portrait recognition model is constructed to obtain a teacher intelligent portrait recognition model;

[0015] The early warning model construction module is used to collect the teaching process time series data of learners and teachers based on the preset learning service platform to obtain the teaching process time series data set; based on the teaching process time series data set, the teaching early warning model is constructed for the learning-teaching interaction process to obtain the teaching abnormality intelligent early warning model;

[0016] The diagnostic model building module is used to collect the full-cycle data of educational effects of learners and teachers based on the preset learning service platform to obtain a full-cycle data set of educational effects; extract educational effect labels from the full-cycle data set of educational effects to obtain an educational effect label set; and construct an educational effect diagnostic model for the educational process based on the full-cycle data set of educational effects and according to the educational effect label set to obtain an educational effect diagnostic model;

[0017] The model integration module is used to integrate the learner intelligent portrait recognition model, the teacher intelligent portrait recognition model, the teaching abnormality intelligent early warning model and the education effect diagnosis model to obtain the big data model of the learning service platform.

[0018] In the present invention, through a highly integrated design, the key modules of learner and teacher portrait recognition, teaching warning and education effect diagnosis are integrated, and the whole process from data collection to model construction is realized. The automation processing is realized, and the work efficiency and accuracy of the system are significantly improved. The intelligent data processing capability of the system can efficiently process and analyze a large amount of multimodal data, reduce the interference of human factors, and ensure the speed and quality of data processing. At the same time, the system has real-time and forward-looking performance, can monitor the teaching process and education effect in real time, discover and warn problems in time, and update and optimize the model in real time according to changes in the education environment. The provision of personalized services, based on learner portrait recognition and teacher portrait recognition models, provides data support for customized teaching and learning plans, and meets the specific needs of different users. The education effect diagnosis model provides data-based decision support for education managers, and promotes the formulation of more effective teaching strategies. The system also realizes a comprehensive evaluation of education effects and provides more comprehensive and continuous education effect monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments thereof made with reference to the following drawings:

[0020] Figure 1A schematic flow chart of the steps of a method for constructing a big data model for a learning service platform based on artificial intelligence according to an embodiment is shown.

[0021] Figure 2 A detailed flow chart of step S4 of an embodiment is shown.

[0022] Figure 3 A detailed flow chart of step S5 of an embodiment is shown. DETAILED DESCRIPTION

[0023] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are 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 technicians in this field without creative work are within the scope of protection of the present invention.

[0024] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0025] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0026] To achieve this, please refer to Figures 1 to 3 The present invention provides a method for constructing a big data model of a learning service platform based on artificial intelligence, comprising the following steps:

[0027] Step S1: Based on the preset learning service platform, multi-dimensional learner data is collected for learners to obtain a multimodal data set of learners; based on the multimodal data set of learners, a learner portrait recognition model is constructed for learners to obtain a learner intelligent portrait recognition model;

[0028] Step S2: Based on the preset learning service platform, multi-dimensional teacher data is collected for teachers to obtain a teacher multimodal data set; based on the teacher multimodal data set, a teacher portrait recognition model is constructed for teachers to obtain a teacher intelligent portrait recognition model;

[0029] Step S3: Based on the preset learning service platform, the teaching process time series data of learners and teachers are collected to obtain a teaching process time series data set; based on the teaching process time series data set, a teaching early warning model is constructed for the learning-teaching interaction process to obtain a teaching abnormality intelligent early warning model;

[0030] Step S4: Based on the preset learning service platform, the full-cycle data of educational effects of learners and teachers are collected to obtain a full-cycle data set of educational effects; educational effect labels are extracted from the full-cycle data set of educational effects to obtain an educational effect label set; based on the full-cycle data set of educational effects and according to the educational effect label set, an educational effect diagnosis model is constructed for the educational process to obtain an educational effect diagnosis model;

[0031] Step S5: Integrate the learner intelligent portrait recognition model, the teacher intelligent portrait recognition model, the teaching abnormality intelligent early warning model and the education effect diagnosis model to obtain a big data model of the learning service platform.

[0032] In this embodiment, multidimensional data collection is first started, focusing on collecting learners' learning behavior data, learning achievement data and learning feedback data, which together constitute learners' multimodal data sets. The tracking and feedback mechanism of the preset learning service platform is used to automatically record learners' activities on the platform, such as course access paths, homework submission records and discussion area interactions. Subsequently, data cleaning and integration are performed through Python's Pandas library, and feature selection and dimensionality reduction are performed using the Scikit-learn Machine Learning Library, thereby constructing a learner portrait recognition model, which identifies learners' behavior patterns and learning preferences through machine learning algorithms, and generates a learner intelligent portrait recognition model. Next, multidimensional data of teachers are collected, including data on teachers' teaching methods, data on the use of teaching resources and data on the frequency of classroom interactions, to form a multimodal data set of teachers. Natural language processing technology is used to analyze teachers' teaching content and style, and social network analysis tools are used to determine teachers' influence in professional communities. Through these technologies, a teacher portrait recognition model is constructed to obtain a teacher intelligent portrait recognition model. Furthermore, the time series interaction data generated by learners and teachers in the teaching process are collected to form a teaching process time series data set. Time series analysis, such as the autoregressive integrated moving average model, is used to identify the trend and periodicity of teaching activities. Combined with pattern recognition technology, such as the hidden Markov model, a teaching warning model is constructed to obtain an intelligent warning model for teaching anomalies, which can monitor abnormal behaviors in the teaching process and issue warnings in time. In addition, full-cycle data collection of educational effects is carried out, including learning behavior data, learning outcome data, learning feedback data, teaching behavior data, and teaching evaluation data, to form a full-cycle data set for educational effects. Using a supervised learning algorithm, combined with an educational effect label set, an educational effect diagnosis model is trained to conduct a detailed evaluation of the educational process and generate an educational effect diagnosis model. Finally, through an integrated learning method, the learner intelligent portrait recognition model, the teacher intelligent portrait recognition model, the teaching anomaly intelligent warning model, and the educational effect diagnosis model are integrated to form a unified learning service platform big data model. In this process, a deep learning framework is used to build and optimize the model to ensure the accuracy and robustness of the model when processing complex educational data. Through these comprehensive steps, a learning service platform big data model is finally obtained.

[0033] Preferably, step S1 comprises the following steps:

[0034] Step S11: collecting the learning behavior data of the learners based on the preset learning service platform to obtain a learning behavior trajectory data set, wherein the learning behavior trajectory data set includes learning path data, learning time data, learning progress data and learning resource access frequency data;

[0035] Specifically, the tracking function of the preset learning service platform can be used to automatically record all interactive behaviors of learners on the platform. Through an integrated event tracking system, such as Google Analytics or a customized tracking script, every page visited by the learner, every button clicked, and every video watched can be captured in detail. These data are stored in the database to form a learning behavior trajectory dataset. Then, data mining tools, such as Python's Pandas library, are used to clean and process the collected data to extract key information, such as learning paths, learning time, learning progress, and frequency of access to learning resources.

[0036] Step S12: Based on the preset learning service platform, the learner's learning resource type data is collected, and the course selection category is collected to obtain the learner's resource preference data and the learner's course category data;

[0037] Specifically, a series of SQL queries can be executed through the database of the preset learning service platform to collect the category information of the courses selected by the learners. For example, a query is executed as follows: SELECT course_id, course_category FROM course_selections WHERE student_id = 'X', where 'X' represents the ID of a specific learner. This will return the unique identifier and category of each course selected by the learner. In this way, a list of course categories can be constructed for each learner, that is, the learner course category data. At the same time, the platform's logging system is used to collect learners' access and interaction data on different learning resources (such as videos, reading materials, and online exercises). By analyzing this data, the learner's preference for each resource type can be determined, and this information can be integrated into learner resource preference data.

[0038] Step S13: clustering learners' interest preferences according to learners' resource preference data and learners' course category data to obtain learners' interest preference data;

[0039] Specifically, an interest preference model can be defined, which uses the learner's course category data as classification features and the resource preference data as quantitative features. Then, a clustering algorithm, such as K-means clustering, is used to process these data. The K-means clustering algorithm will divide learners into different interest groups based on their course categories and resource preferences. For example, learners can be divided into "math enthusiasts" and "literature enthusiasts" groups. Finally, learner interest preference data is obtained, which can reflect learners' preference patterns for different learning fields and resource types.

[0040] Step S14: Based on the preset learning service platform, learning effect data and learning achievement data of the learner are collected, and basic identity data is collected to obtain learning effect data, learning achievement data and basic identity data of the learner;

[0041] Specifically, the built-in evaluation module of the preset learning service platform can be used to collect learners' learning effect data. This module can automatically record learners' scores in various online tests and exams. Use the data analysis tools provided by the platform, such as the reporting function of the learning management system, to export these performance data. Among them, the learning effect data also includes an evaluation dimension, that is, to collect the specific impact of learning on learners' work through questionnaires. This questionnaire can cover the frequency and scope of application of learning content in work, the impact of learning outcomes on work performance, the improvement of professional skills and knowledge structure, the acquisition of soft skills and their practical application, and the impact of learning on career satisfaction and work attitude. In order to improve the participation and data quality of the questionnaire, easy-to-understand questionnaire content can be designed, incentives can be provided to encourage learners to participate, the anonymity and privacy protection of the questionnaire can be ensured, and the data can be analyzed regularly to optimize the course content. These measures will help collect learners' direct feedback on the course, and transform learning outcomes into actual work capabilities, providing learners with more personalized and effective educational services. Then, through the platform's homework submission and feedback system, learners' learning outcome data are collected, including their papers, project reports, and teachers' comments. Basic identity data is collected through the personal information form filled out by learners when registering on the platform. This data includes but is not limited to the learner's name, age, gender and educational background information.

[0042] Step S15: Record the learning behavior trajectory data set, learner resource preference data, learner course category data, learner interest preference data, learning effect data, learning achievement data and learner basic identity data as a learner multimodal data set;

[0043] Specifically, data integration technology can be used to merge learning behavior trajectory data sets, learner resource preference data, learner course category data, learner interest preference data, learning effect data, learning outcome data and learner basic identity data, and unified storage of structured and unstructured data can be achieved through data lakes or data warehouses. Furthermore, through data fusion technology, such as Python's Pandas library, learning effect data and learning outcome data are associated with multimodal data sets, and data frames are merged to form a unified view. Finally, the learner's basic identity data is synchronized through the UPSERT operation of the database, the construction of the multimodal data set is completed, and finally the learner's multimodal data set is obtained.

[0044] Step S16: construct a learner portrait recognition model for the learner based on the learner multimodal data set to obtain a learner intelligent portrait recognition model.

[0045] Specifically, please refer to the sub-steps of step S16 for the detailed implementation process of this embodiment.

[0046] The present invention realizes a comprehensive analysis of learner behavior through meticulous step design, laying a solid data foundation for personalized learning experience. By deeply understanding the learners' interest preferences and clustering analysis, learners are provided with richer and more accurate resource recommendations and course arrangements. At the same time, personalized learning paths are customized in combination with learners' interests and course category data to effectively improve learners' learning motivation and efficiency. Through a comprehensive evaluation of learning effects and results, it helps to timely discover and solve difficulties encountered in learning. Through the integrated application of learners' basic identity data, customized teaching is provided with deeper support. Through the construction of an intelligent portrait recognition model, learner portraits can be dynamically updated, providing strong technical support for precise teaching and personalized learning. Finally, through in-depth analysis of learners' behaviors and preferences, accurate matching of teaching resources is achieved, and resource utilization efficiency and learning satisfaction are improved.

[0047] Preferably, step S16 comprises the following steps:

[0048] Step S161: constructing a portrait framework for the learner to obtain a multi-dimensional portrait framework of the learner, wherein the multi-dimensional portrait framework of the learner includes a learner basic attribute dimension, a learner behavior pattern dimension, a learner learning style dimension, a learner interest tendency dimension, and a learner performance dimension;

[0049] Specifically, the structure of the learner multidimensional portrait framework can be defined, which includes five key dimensions: basic attributes, behavioral patterns, learning styles, interest tendencies, and performance. Use data modeling tools, such as ERwin (ERwin DataModeler) or PowerDesigner, to create a data model and determine the attributes and data types of each dimension. For example, basic attributes include the learner's name, age, and education level; behavioral patterns involve the learner's login frequency and online time. Then, use database queries and data extraction scripts to extract relevant data from the multimodal data set. In this way, a learner multidimensional portrait framework is constructed.

[0050] Step S162: performing sequence pattern mining on the learning path data to obtain learner path feature data, and performing time series analysis on the learning time data to obtain learning time distribution feature data;

[0051] Specifically, data mining tools such as SPSS (Statistical Package for the SocialSciences) Modeler or Python's Scikit-learn library can be used to perform sequential pattern mining on learning path data. This involves identifying the most common paths that learners follow in the course and the order in which they complete the course units. By applying sequence mining algorithms, such as PrefixSpan or GSP (Generalized Sequential Pattern mining), common patterns and trends in learner behavior can be discovered. At the same time, time series analysis tools such as Python's Statsmodels (statistical modeling in Python) library are used to analyze learning time data. This includes identifying the time patterns of learning activities, for example, whether learners tend to study at a specific time of the day or a specific day of the week. Through these analyses, learner path feature data and learning time distribution feature data are finally obtained.

[0052] Step S163: Performing progress quantitative evaluation on the learning progress data to obtain learner progress quantitative data, and performing frequency pattern recognition on the learning resource access frequency data to obtain learning resource access pattern data;

[0053] Specifically, you can use the built-in tools of the learning management system to track and record learners' progress in the course. By calculating the percentage of course units completed by learners or the number of assessments passed, you can get quantitative data on learner progress. Then, use data analysis tools, such as Microsoft Excel or Google Sheets, to create progress tracking reports and use conditional formatting to visualize learners' progress status. In addition, use log analysis tools to monitor learners' access to learning resources and identify access patterns of learning resources by counting the number of visits and access intervals to each resource. By applying frequency distribution analysis and pattern recognition algorithms, such as the Apriori algorithm, you can finally get learning resource access pattern data.

[0054] Step S164: performing behavior pattern fusion on the learner path feature data, the learning time distribution feature data, the learner progress quantification data and the learning resource access pattern data to obtain learner behavior pattern feature data;

[0055] Specifically, data preprocessing tools, such as Python's Pandas library, can be used to clean and standardize learner path feature data, learning time distribution feature data, learner progress quantification data, and learning resource access pattern data. Next, association rule learning algorithms in machine learning, such as Apriori or FP-Growth (Frequent Pattern Growth), are applied to mine the potential connections and frequent patterns between these data to identify learners' behavior patterns. In addition, clustering algorithms, such as K-means clustering or hierarchical clustering, are used to group learners' behavior data to discover different categories of behavior patterns. Finally, through feature engineering, these patterns and grouping results are encoded into feature vectors to form learners' behavior pattern feature data.

[0056] Step S165: clustering the learner resource preference data into learning styles to obtain learner learning style feature data;

[0057] Specifically, data mining tools can be used to process learners' preference data for different types of learning resources, such as video, text, audio, and interactive modules. Then, clustering algorithms, such as K-means clustering or DBSCAN (Density-Based Spatial Clustering of Applications with Noise), are applied to group learners' resource preferences to identify different learning styles, such as visual, auditory, or hands-on. In addition, the model evaluation tool in Python's Scikit-learn library is used to evaluate the clustering effect to ensure that the classification of learning styles is both accurate and discriminative. Finally, learners' learning style feature data is obtained.

[0058] Step S166: performing learning topic identification on the learner's interest preference data to obtain learner's interest topic data, and performing dynamic interest tracking on the learner's interest topic data to obtain learner's interest evolution feature data;

[0059] Specifically, text analysis tools, such as Python's NLTK (Natural Language Toolkit) or spaCy (Industrial-strength Natural Language Processing (NLP) in Python) library, can be used to identify learning topics for learners' interest preference data. By performing word segmentation, part-of-speech tagging, and sentiment analysis on learners' language expressions in forum posts, homework submissions, and feedback, key topics reflecting learners' interests can be extracted. Next, topic modeling techniques, such as Latent Dirichlet Allocation (LDA) or Non-negative MatrixFactorization (NMF), are used to further identify and quantify multiple topics of learners' interests. Then, dynamic interest tracking methods are applied, such as using time series analysis to capture the evolution of learners' interests by tracking learners' behavior on specific topics over time. In addition, machine learning models, such as hidden Markov models, can be used to predict the transition probability of learners' interest states. Finally, learners' interest topic data and interest evolution feature data are finally obtained.

[0060] Step S167: performing effect quantification scoring on the learning effect data to obtain learner effect quantification data, and performing achievement quantification on the learning achievement data to obtain learner achievement quantification feature data;

[0061] Specifically, the assessment tools in the learning management system can be used to quantify learning effect data. This includes converting qualitative data such as test scores, homework scores, and teacher feedback into numerical scores. By writing automated scripts, these scoring data are extracted from the LMS (Learning Management System) database, and statistical methods are used to calculate statistical indicators such as mean scores and standard deviations to obtain quantitative data on learner effects. Next, the learning outcome data is quantified, for example, by analyzing the depth and breadth of project reports submitted by learners and their participation in forum discussions to evaluate the quality and quantity of learning outcomes. Use text analysis tools, such as Python's NLTK library, to quantify text data, extract keywords and topics, and obtain quantitative feature data on learner outcomes.

[0062] Step S168: Perform a comprehensive learning performance evaluation on the learner effect quantified data and the learner achievement quantified characteristic data to obtain learner performance characteristic data;

[0063] Specifically, multi-criteria decision analysis methods, such as weighted summation methods, can be used to comprehensively consider the importance of different indicators. For example, different weights can be assigned to test scores, homework quality, and participation to calculate the learner's comprehensive performance score and ultimately obtain learner performance characteristic data.

[0064] Step S169: Based on the learner's basic identity data, the learner's behavior pattern feature data, the learner's learning style feature data, the learner's interest evolution feature data and the learner's performance feature data, the learner's feature fusion and portrait construction are performed according to the learner's multidimensional portrait framework to obtain the learner's personalized multidimensional portrait; the learner's personalized multidimensional portrait is subjected to learning feature mapping and model training to obtain the learner's intelligent portrait recognition model.

[0065] Specifically, data integration techniques, such as the ETL (Extract, Transform, Load) process of the data warehouse, can be used to integrate learners' basic identity data, learners' behavioral pattern feature data, learners' learning style feature data, learners' interest evolution feature data, and learners' performance feature data into a unified data platform. Then, machine learning algorithms, such as K-nearest neighbor or support vector machine, are applied to analyze and learn these feature data to build learner portraits. Feature selection techniques, such as recursive feature elimination, are used to identify the most influential features and use these features to train classification models. In addition, deep learning frameworks, such as TensorFlow or PyTorch, are used to build neural network models that can capture nonlinear relationships and interaction effects in the data. Through model training and verification, a learner intelligent portrait recognition model is finally obtained, which can predict the learner's learning style and interests based on the learner's behavior and performance.

[0066] The present invention achieves an in-depth understanding of learners' basic attributes, behavior patterns, learning styles, interest tendencies and performance by constructing a comprehensive learner multidimensional portrait framework. By applying sequence pattern mining and time series analysis, learners' learning paths and time distribution can be deeply analyzed to optimize learning plans. By quantitatively evaluating learning progress and resource access frequency, learning progress can be accurately tracked, and resources can be adjusted in time to meet learners' needs. Comprehensive evaluation of learners' behavior patterns helps to provide personalized teaching support. Through learning style clustering, different learners' styles can be identified and adapted to enhance the learning experience. By dynamically tracking the evolution of learners' interests, learning motivation can be continuously stimulated. By comprehensively evaluating learners' performance, timely feedback and guidance can be provided to learners. By constructing personalized multidimensional portraits, deep insights into precise teaching and personalized learning can be provided. Through the training of intelligent portrait recognition models, learners' behaviors and needs can be intelligently predicted and responded to, thereby achieving continuous optimization of teaching strategies and continuous improvement of learning experience.

[0067] Preferably, step S2 comprises the following steps:

[0068] Step S21: collecting basic attribute data of teachers based on a preset learning service platform to obtain a static feature data set of teachers, wherein the static feature data set of teachers includes gender data, age data, professional background data and teaching field data of teachers;

[0069] Specifically, the basic attribute data of teachers can be collected through the teacher information management system of the preset learning service platform. This includes exporting the teacher's gender, age, professional background and teaching field information from the system. Use SQL query statements to retrieve the teacher's static feature data set. Finally, the collected data is stored in the data warehouse to build a teacher's static feature data set.

[0070] Step S22: Based on the preset learning service platform, the teacher's teaching behavior time series data is collected, and the teaching content data is collected to obtain the teacher's teaching behavior data and the teacher's teaching content data;

[0071] Specifically, the logging system of the preset learning service platform can be used to collect the time series data of teachers' teaching behavior. This involves using event tracking tools to record all the teaching activities of teachers on the platform, such as publishing homework, uploading teaching resources, and conducting online discussions. At the same time, the platform's content management system is used to collect teaching content data, including video lectures, presentations, and teaching texts uploaded by teachers. Finally, the teacher's teaching behavior data and teacher teaching content data are obtained.

[0072] Step S23: performing content separation processing on the teacher's teaching content data to obtain teaching content video data and teaching content text data;

[0073] Specifically, multimedia content analysis tools can be used to identify and classify the various components in teaching videos, such as lectures, experimental demonstrations, and question-and-answer sessions. Then, text mining techniques, such as natural language processing (NLP) tools, are applied to extract key information from teaching text data, including course outlines, learning objectives, and important concepts. Python's OpenCV (Open Source Computer Vision Library) library is used to process video data to achieve frame extraction and scene segmentation; the NLTK library is used to perform word segmentation, part-of-speech tagging, and named entity recognition of text data. Through these processes, the teaching content video data and teaching content text data are successfully separated.

[0074] Step S24: Based on the preset learning service platform, comprehensive data on the teaching quality and effectiveness of teachers is collected to obtain a teacher quality and effectiveness evaluation data set, wherein the teacher quality and effectiveness evaluation data set includes teacher achievement data, learning effect data, and teaching influence evaluation data;

[0075] Specifically, the research results data of teachers can be collected through the achievement management system of the preset learning service platform, which includes but is not limited to published papers, participated projects and awards received. Using the platform's data analysis tools, the influence and recognition of these achievements can be quantified. Next, learning effect data are collected through the student feedback system and the grade management system, which includes students' test scores, course evaluations and satisfaction survey results. In addition, social network analysis tools are used to evaluate the teaching influence of teachers, which involves analyzing the teacher's activity in professional communities, student evaluations and peer recommendations. By integrating these data, a teacher quality and effectiveness evaluation data set is constructed, among which the teacher quality and effectiveness evaluation data set includes teacher achievement data, learning effect data and teaching influence evaluation data.

[0076] Step S25: Record the teacher static feature data set, teacher teaching behavior data, teaching content video data, teaching content text data and teacher quality and effectiveness evaluation data set as the teacher multimodal data set;

[0077] Specifically, data integration technologies, such as ETL processes, can be used to integrate teacher static feature data sets, teacher teaching behavior data, teaching content video data, teaching content text data, and teacher quality and effectiveness evaluation data sets. Then, data modeling tools, such as ERwin, are used to design the architecture of the teacher multimodal data set and define the relationships and dependencies between data. In addition, database management systems, such as MySQL or MongoDB, are used to store and manage the integrated data. Through these steps, a teacher multimodal data set is constructed.

[0078] Step S26: construct a teacher portrait recognition model for teachers based on the teacher multimodal data set to obtain a teacher intelligent portrait recognition model.

[0079] Specifically, please refer to the sub-steps of step S26 for the detailed implementation process of this embodiment.

[0080] The present invention constructs a detailed teacher portrait by comprehensively capturing the basic attribute data of teachers. By deeply analyzing the teaching behavior and content of teachers, it is possible to reveal the teaching model and structure and guide the continuous optimization of teaching methods. By analyzing the teaching content in multiple dimensions, the system strengthens the understanding of teaching materials and enriches the presentation of teaching content. By comprehensively evaluating the teaching quality and effectiveness of teachers, the system provides substantial feedback for the professional development of teachers. The constructed multimodal data set integrates multiple key data dimensions of teachers and provides comprehensive information support for the in-depth analysis of teachers' teaching characteristics. Through the establishment of an intelligent teacher portrait recognition model, it is possible not only to dynamically update the teaching advantages of teachers, but also to support personalized teaching and self-improvement of teachers. In addition, the present invention can also provide optimization suggestions for teaching methods and content based on teacher portraits to improve the quality of teaching. Ultimately, these measures help to improve the level of personalization of teaching, meet the needs of different learners, and maximize teaching effectiveness.

[0081] Preferably, step S26 comprises the following steps:

[0082] Step S261: constructing a teacher portrait framework to obtain a teacher multi-dimensional portrait framework, wherein the teacher multi-dimensional portrait framework includes a teacher static attribute dimension, a teaching resource utilization dimension, a teaching style dimension, and a teaching effectiveness dimension;

[0083] Specifically, a multidimensional teacher portrait framework can be defined, which covers the dimensions of teachers' static attributes, teaching resource utilization, teaching style, and teaching effectiveness. Using data modeling software such as ERwin or PowerDesigner, a conceptual model is created to clarify the attributes contained in each dimension and the relationship between them. For example, the teacher static attribute dimension includes the teacher's gender, age, professional background, and teaching field information.

[0084] Step S262: extracting and encoding the teacher's gender data, teacher's age data, teacher's professional background data and teacher's teaching field data to obtain the teacher's basic information code;

[0085] Specifically, the gender, age, professional background and teaching field data of teachers were feature extracted and encoded. First, data preprocessing tools such as Python's Pandas library were used to clean the data, including removing invalid values ​​and standardizing the format. Then, feature engineering techniques were applied to transform the data, for example, using One-Hot Encoding to process categorical data of gender and teaching field, and using Label Encoding to process text data of professional background. For age data, the age group to which it belongs was calculated as a feature. Through these methods, the basic information coding of teachers was finally obtained.

[0086] Step S263: performing usage statistics on the teaching content video data and the teaching content text data respectively to obtain video resource usage data and text resource usage data, and performing resource utilization efficiency calculation on the video resource usage data and the text resource usage data to obtain teaching resource utilization rate data;

[0087] Specifically, the reporting tools of the learning management system can be used to track and record the number of visits, viewing / reading completion rates, and student interaction data of video and text resources. Then, data analysis software, such as Excel or Python's Pandas library, is used to calculate the frequency and popularity of these resources. Next, evaluation indicators for resource utilization efficiency are defined, such as resource access depth (the completeness of viewing or reading) and student engagement (the number of comments, ratings, and shares). By calculating these indicators, video resource usage data and text resource usage data can be obtained. Finally, statistical analysis methods, such as regression analysis, are used to evaluate the relationship between resource utilization efficiency and teaching outcomes, thereby obtaining teaching resource utilization data.

[0088] Step S264: Perform teaching style identification on the teaching content video data to obtain preliminary teaching style classification data; and annotate and verify the preliminary teaching style classification data to obtain teacher teaching style data, wherein the teacher teaching style data is any one or more combinations of directional, heuristic, personalized, interactive, and practical;

[0089] Specifically, video analysis tools can be used to process teaching content video data to identify the teacher's teaching style. Using machine learning libraries such as Python's Scikit-learn, a classification model is trained that can classify teaching styles into guided, heuristic, personalized, interactive, and practical categories based on features in the video content, such as the teacher's language use, interactive methods, and classroom activities. By learning from a large amount of labeled video data, the model can recognize and classify new teaching videos. Then, the model's preliminary classification results are reviewed and corrected by an expert annotation team to ensure the accuracy of the classification. Use annotation tools such as Brilliant Annotation Tool (BAT) to annotate the video content in detail. Finally, the teacher's teaching style data is obtained.

[0090] Step S265: Evaluate the teacher achievement data for achievement influence, and obtain the teacher achievement influence score;

[0091] Specifically, academic databases such as Google Scholar or Web of Science can be used to collect the number of citations of papers published by teachers, journal impact factors, and ranking data in the research field. Then, social network analysis tools can be used to evaluate the network position and influence of teachers in the academic community. In addition, peer evaluations of teachers' achievements can be collected through questionnaires and interviews. Statistical analysis software such as SPSS can be used to combine these data to calculate the impact score of teachers' achievements.

[0092] Step S266: performing a learning effectiveness analysis on the learning effectiveness data to obtain a student learning effectiveness indicator set, and performing a learning effectiveness score on the student learning effectiveness indicator set to obtain a student comprehensive learning effectiveness score;

[0093] Specifically, data mining tools, such as the CARM (Computer-Assisted Reasoning for Medical Diagnosis) package in R language, can be used to conduct in-depth analysis of learning effectiveness data to identify key factors that affect students' learning effectiveness. Then, based on educational psychology theories, such as Bloom's cognitive domain, a series of learning effectiveness indicators are defined, such as knowledge mastery, application ability, and analytical ability. Using educational data mining methods, such as the item response theory model, each student's learning effectiveness indicator is scored. Finally, these indicators are converted into a comprehensive learning effectiveness score for students through weighted summation or other comprehensive scoring methods. This score provides a comprehensive assessment of student learning effectiveness.

[0094] Step S267: Perform multi-dimensional influence evaluation on the teaching influence evaluation data to obtain a comprehensive teacher influence score, wherein the multi-dimensional influence evaluation specifically includes student satisfaction dimension, peer evaluation dimension, social influence dimension, and scientific research contribution dimension;

[0095] Specifically, the teaching impact evaluation data includes students' satisfaction evaluation of teachers, including but not limited to teaching methods, interactivity and course content. Next, use academic social platforms, such as ResearchGate or Academia.edu, to evaluate the peer evaluation of teachers by analyzing peer recommendations, comments and cooperation request data. In addition, social media analysis tools can be used to measure the influence of teachers on social media, including the number of times posts are shared, fan interactions and public mentions. For the scientific research contribution dimension, the number of papers published, the number of citations and the h-index index of teachers can be collected through academic databases such as Web of Science or Scopus. Then, use multi-criteria decision analysis methods, such as hierarchical analysis process, to integrate the data of these different dimensions, assign appropriate weights, and calculate the comprehensive influence score of teachers.

[0096] Step S268: Perform multi-dimensional effectiveness fusion on the teacher's achievement influence score, the student's comprehensive learning effect score and the teacher's comprehensive influence score to obtain the teacher's comprehensive effectiveness index;

[0097] Specifically, statistical analysis software, such as SPSS or R language, can be used to analyze and standardize the teacher's achievement impact score, the student's comprehensive learning effect score and the teacher's comprehensive impact score to ensure that they are comparable. Then, according to the educational goals and teacher evaluation criteria, weights are assigned to each scoring dimension, which reflects the importance of different dimensions in the evaluation of teacher effectiveness. Next, the weighted summation method is used to combine these weighted scores to obtain the teacher's comprehensive effectiveness index.

[0098] Step S269: Based on the teacher basic information coding, teaching resource utilization data, teacher teaching style data and teacher comprehensive effectiveness index, multi-dimensional features of teachers are fused according to the teacher multi-dimensional portrait framework to obtain the teacher comprehensive feature vector, and deep learning is performed on the teacher comprehensive feature vector to obtain the teacher intelligent portrait recognition model.

[0099] Specifically, based on the teacher multi-dimensional portrait framework, the teacher basic information coding, teaching resource utilization data, teacher teaching style data and teacher comprehensive effectiveness index can be integrated to construct a comprehensive feature vector of the teacher. Use feature engineering methods, such as feature scaling and feature selection, to process and optimize these data. Then, use deep learning techniques, such as neural networks, to train the teacher intelligent portrait recognition model. Use deep learning frameworks, such as TensorFlow or Keras, to build a neural network model with multiple hidden layers that can learn the relationship between teacher feature vectors and teaching effectiveness. Through the training and verification process, the model parameters are continuously adjusted, and cross-validation and grid search techniques are used to optimize the model performance. Finally, an intelligent portrait recognition model that can accurately identify and predict the multi-dimensional characteristics of teachers is obtained.

[0100] The present invention realizes the in-depth capture of the key features of teachers' static attributes, teaching resource utilization, teaching style and teaching effectiveness by constructing a comprehensive multi-dimensional teacher portrait framework. The efficiency of data storage and analysis is improved through refined teacher data management through feature extraction and encoding. By statistically analyzing the utilization of teaching resources and calculating the efficiency, it provides guidance for the rational allocation of teaching content and resource optimization. By automatically identifying and verifying the teaching style of teachers, it provides a basis for teachers' self-improvement and teaching innovation. By comprehensively evaluating the teaching achievements and influence of teachers, a comprehensive teaching performance evaluation system is formed. The comprehensive teacher effectiveness index obtained by multi-dimensional fusion reflects the teaching effect of teachers more comprehensively. The establishment of an intelligent portrait recognition model uses deep learning technology to enhance the intelligent recognition and prediction capabilities of teacher characteristics.

[0101] Preferably, step S3 comprises the following steps:

[0102] Step S31: Based on the preset learning service platform, the teaching process time series data of the learners and teachers are collected to obtain the teaching process time series data set, wherein the teaching process time series data set includes learner behavior data, teacher behavior data and learner-teacher interaction data, the learner behavior data includes learning progress data, learning time data, learning resource type data, learning resource access sequence data, learning interaction data, learning effectiveness feedback data, the teacher behavior data includes teacher teaching method data, classroom interaction frequency data and teaching resource usage data, and the learner-teacher interaction data includes teacher question-answer interaction data and classroom participation data;

[0103] Specifically, the built-in tracking system of the learning service platform can be used to collect the time series data of the teaching process of learners and teachers. This includes using the platform's API to automatically collect learners' login time, course access records, homework submissions and test scores to form learner behavior data. At the same time, the teacher's teaching method, classroom interaction frequency and teaching resource usage are captured to form teacher behavior data. In addition, the interaction between learners and teachers, including questions, answers and feedback, is recorded through forums, instant messaging and question-and-answer systems to form learner-teacher interaction data.

[0104] Step S32: performing learning trajectory recognition on the learning progress data and the learning duration data to obtain learning behavior trajectory data;

[0105] Specifically, data mining tools, such as SPSS Modeler, can be used to analyze learning progress data and learning time data. By setting parameters to define thresholds for learning activities, such as the percentage of course completion or daily learning time, learners' learning stages and speeds can be identified. Then, sequence analysis tools are applied to track learners' transition paths between different courses and modules, identifying common learning patterns and potential bottlenecks. These analysis results are integrated into learning behavior trajectory data.

[0106] Step S33: performing behavior pattern recognition on the learning interaction data to obtain learning interaction pattern data;

[0107] Specifically, learning interaction data includes data such as student questions, teacher replies, online discussion posts, and real-time chat conversations. Text analysis tools such as Python's NLTK library are used to preprocess these text data, perform word segmentation, remove stop words, and extract stems. Sentiment analysis technology is used to determine the emotional tendency of the interactive text, revealing the emotional color of the communication. The latent Dirichlet allocation topic modeling technology is further used to identify the main topics and subtopics of the discussion, clarifying the core content of the discussion. The social network analysis tool is used to construct and visualize the interaction network between students and teachers, and identify the key participants in the network. The K-means clustering machine learning algorithm is used to classify the interaction data, and different learning interaction modes are identified, such as "exploratory discussion", "problem-solving interaction" or "knowledge sharing communication". Finally, the learning interaction mode data is obtained.

[0108] Specifically, learning effectiveness feedback data includes but is not limited to test scores, homework scores, self-assessment, and peer evaluation. Then, use educational analysis tools, such as Bloom's Taxonomy framework, to define different levels and indicators of learning effectiveness. Use data mining techniques, such as decision trees or random forest algorithms, to analyze these feedback data to identify key factors that affect learning effectiveness. Use statistical analysis software, such as R or Python's Scikit-learn library, to calculate each learner's performance on each indicator, and generate learning effectiveness indicator data for each learner based on preset standards or benchmarks.

[0109] Step S35: performing multimodal learning behavior fusion on the learning behavior trajectory data, the learning interaction mode data and the learning effectiveness indicator data to obtain the learner's comprehensive learning behavior data;

[0110] Specifically, data preprocessing tools, such as Python's Pandas library, can be used to clean and standardize learning behavior trajectory data, learning interaction pattern data, and learning effectiveness indicator data. Then, multimodal learning analysis methods are applied to identify potential patterns and associations in the data through principal component analysis or t-distributed random neighbor embedding dimensionality reduction technology. Next, clustering algorithms, such as K-means clustering, are used to classify learners into different learning behavior types. Finally, a comprehensive data model is constructed that combines multiple behavioral dimensions of learners to generate comprehensive learning behavior data for learners.

[0111] Step S36: tracking the interest preference of the learning resource type data to obtain the learning interest preference data, and identifying the learning style of the learning resource type data and the learning resource access sequence data to obtain the learner's learning style data;

[0112] Specifically, association rule learning algorithms, such as Apriori or FP-Growth (Frequent Pattern Growth), can be applied based on learning resource type data to identify learners' preference patterns for different resources. Through these algorithms, it is possible to discover which resources are often accessed together and which resources are the most popular. Next, sequential pattern mining techniques are used to analyze the order in which learners access resources to identify their learning styles, such as linear learners, exploratory learners, or deep learners. Finally, combined with learners' resource preferences and access sequence data, classification algorithms, such as support vector machines or gradient boosting machines, are used to build learners' learning style models and generate learners' learning style data.

[0113] Step S37: Perform performance statistics on the learning performance feedback data to obtain learning performance effect data;

[0114] Specifically, learning performance effects can be extracted and counted from learning effectiveness feedback data. Initially, the performance data including test scores, homework scores, and project evaluations are exported from the learning management system, which is done by executing SQL queries or using API calls. Subsequently, data analysis tools such as Microsoft Excel and Python's Pandas library are used to process the performance data and calculate key statistical indicators such as mean score, median, standard deviation, and percentile ranking to quantify the distribution and variation of learning performance. Combining these methods, learning performance effect data that reflects the overall performance of learners is finally obtained.

[0115] Step S38: Based on the teacher behavior data and the learner-teacher interaction data and according to the learner's comprehensive learning behavior data, learning interest preference data, learner learning style data and learning performance effect data, a teaching early warning model is constructed for the learning-teaching interaction process to obtain an intelligent early warning model for teaching abnormalities.

[0116] Specifically, please refer to the sub-steps of step S38 for the detailed implementation process of this embodiment.

[0117] The present invention realizes in-depth analysis and real-time evaluation of teaching behaviors and interactions by comprehensively monitoring the learners' and teachers' time series data in the teaching process, and provides a data basis for personalized learning support. Through the fusion of multimodal learning behaviors, a comprehensive learner behavior profile is formed, and customized learning resources are recommended to learners through the identification of personalized learning interests and styles. By quantifying and statistically analyzing the learning performance effects, intuitive feedback is provided to learners and teachers, which promotes the adjustment of learning strategies. Through the construction of an intelligent early warning model, teaching anomalies can be identified in time and early warnings can be issued, which optimizes the teaching interaction process and improves the teaching quality. In addition, teachers can dynamically adjust teaching strategies according to the real-time feedback of the intelligent early warning model to meet the needs of different learners, thereby improving the interactivity and teaching effect of teaching as a whole.

[0118] Preferably, step S38 includes the following steps:

[0119] Step S381: Identify the teaching characteristics of the teacher's teaching method data to obtain the teacher's teaching characteristics data, calculate the interactive activity of the classroom interaction frequency data to obtain the classroom interactive activity data, and calculate the teaching resource utilization efficiency of the teaching resource usage data to obtain the teaching resource utilization efficiency data;

[0120] Specifically, you can collect teacher teaching videos, teaching plans, course outlines, and teaching activity record data. Use natural language processing (NLP) tools, such as Python's NLTK or spaCy library, to analyze the teaching language by combining the teacher's teaching method data, extract key words, phrases, and concepts, and reveal the teacher's teaching style. Then, evaluate the interactive activity by calculating the frequency of classroom interactions, such as the number of questions and answers. Use Python's SciPy library to calculate the frequency and interval time of classroom interactions to quantify the activeness of the interaction. In addition, analyze the use records of teaching resources, such as click-through rate and access time, to evaluate the efficiency of resource use. By calculating the concentration and depth of resource access, we can obtain the teaching resource utilization efficiency data.

[0121] Step S382: performing multi-dimensional behavior pattern clustering on the teacher's teaching characteristics data, classroom interaction activity data, and teaching resource utilization efficiency data to obtain teaching behavior pattern label data;

[0122] Specifically, clustering algorithms, such as K-means clustering or hierarchical clustering, can be used to cluster the behavior patterns of teachers' teaching characteristics data, classroom interaction activity data, and teaching resource utilization efficiency data. Through Python's Scikit-learn library, the characteristics and distance metrics of the data are defined to determine the appropriate number and categories of clusters. Then, data visualization techniques, such as scatter plots and heat maps, are used to display the teaching behavior patterns of different teachers. Through these clustering results, the teaching behavior pattern label data is obtained.

[0123] Step S383: Perform teaching style recognition on the teacher's question-answer interaction data to obtain the teacher's teaching style data;

[0124] Specifically, text analysis tools can be used to process the interactive text of teachers in the question-and-answer session, that is, the teacher's question-and-answer interaction data, including the answers to questions and the guidance of discussions. Use NLP techniques such as sentiment analysis and topic modeling to analyze the teacher's language use and topic coverage. Then, apply machine learning algorithms such as support vector machines or random forests to classify and identify different teaching styles, such as directive, exploratory, or collaborative teaching. By training the model to identify language patterns and interaction features, data reflecting the teacher's teaching style is obtained.

[0125] Step S384: performing a dialogue quality assessment on the teacher question-and-answer interaction data to obtain a teacher question-and-answer interaction quality score, and performing a student participation calculation on the classroom participation data to obtain student participation data;

[0126] Specifically, text analysis tools can be used to process teacher Q&A interaction data. Through natural language processing (NLP) techniques such as sentiment analysis and text classification, the quality of teacher questions, the accuracy of answers, and the enthusiasm of interaction are evaluated to obtain the teacher Q&A interaction quality score. For example, Python's NLTK (Natural Language Toolkit) library can be used to identify keywords and sentiment tendencies in questions and answers, and then quantify the quality of interaction. Then, student engagement is calculated by analyzing classroom participation data such as the number of student speeches, the number of discussion posts, and the timeliness of homework submission. Data aggregation and statistical analysis, such as in Excel or R language, are used to obtain student engagement data.

[0127] Step S385: Perform teaching quality efficiency evaluation on the teacher's question-and-answer interaction quality score and the student participation data to obtain teaching quality efficiency evaluation data;

[0128] Specifically, educational data mining methods, such as the Educational Production Function (EPF), can be used to analyze the impact of teacher interaction quality and student engagement on student learning outcomes. By constructing a regression model, the effect of teacher interaction on student engagement can be quantified and the efficiency of teaching quality can be evaluated. Using statistical software, such as SPSS or Stata, the teaching quality efficiency evaluation data is finally obtained, which reflects the effectiveness of the teacher's teaching method and the student's learning input-output ratio.

[0129] Step S386: Analyze the synchronization of learning behavior and teaching behavior on the learner's comprehensive learning behavior data and teaching behavior pattern label data to obtain the learning-teaching behavior consistency evaluation index, and calculate the semantic similarity and topic relevance of the learning interest preference data and the teacher's teaching feature data to obtain the interest-teaching feature matching matrix;

[0130] Specifically, data fusion technology can be used to integrate learners' comprehensive learning behavior data and teaching behavior pattern label data. By analyzing learners' behavior patterns and teachers' teaching styles, machine learning algorithms such as K-nearest neighbors or decision trees are used to evaluate the consistency of learning-teaching behaviors. Next, the semantic similarity and topic relevance of learning interest preference data and teachers' teaching characteristics data are calculated. Using NLP technology and semantic analysis tools such as Word2Vec or Gensim, text data is converted into vector representations and the similarity between them is calculated. Using topic modeling techniques such as LDA (Latent Dirichlet Allocation), the topic associations between learners' interests and teachers' characteristics are identified. Through these analyses, the learning-teaching behavior consistency evaluation index and the interest-teaching characteristics matching matrix are obtained.

[0131] Step S387: Performing style matching evaluation on the learner's learning style data and the teacher's teaching style data to obtain a style matching score, and performing effect-efficiency correlation difference analysis on the learning performance effect data and the teaching quality efficiency evaluation data to obtain learning-teaching effectiveness difference data;

[0132] Specifically, data matching tools can be used to evaluate the match between learners' learning style data and teachers' teaching style data. Use machine learning algorithms, such as cosine similarity or Jaccard similarity index, to quantify the similarity between the two styles and obtain a style matching score. Next, analyze the learning performance data and teaching quality and efficiency evaluation data, and use statistical methods, such as covariance analysis or multiple regression analysis, to identify the correlation differences between the two. It is possible to evaluate the impact of different teaching styles and learning style combinations on learning outcomes and obtain learning-teaching effectiveness difference data.

[0133] Step S388: extracting and fusing abnormal features of learner comprehensive learning behavior data, teaching behavior pattern label data, learning interest preference data, teacher teaching feature data, learner learning style data, teacher teaching style data, learning performance effect data and teaching quality efficiency evaluation data according to the learning-teaching behavior consistency evaluation index, interest-teaching feature matching matrix, style matching score and learning-teaching effectiveness difference data to obtain a learning-teaching abnormal feature vector set;

[0134] Specifically, a set of standards for evaluating the consistency of learning-teaching behavior, the matching degree of interest and teaching characteristics, the matching degree of style, and the difference in learning-teaching effectiveness can be established based on the statistical analysis results of historical data and the opinions of educational experts. Using Python's Pandas library, the relevant data of learners and teachers are analyzed and compared with predefined standards, and outlier detection algorithms such as box plot method based on IQR (Interquartile Range) or standard deviation method based on Z-score are used to identify data points that do not meet expectations. Then, feature engineering methods, including principal component analysis or autoencoders, are applied to extract features and reduce dimensions of these abnormal data, and a learning-teaching abnormal feature vector set containing key information is constructed, where historical data is an educational data set that integrates multiple dimensions. This data set specifically covers learning behavior data, which records learners' activities on the platform, such as login frequency, course access time, number of homework submissions, and participation in online discussions. Learning performance data is also included, which records learners' scores in various tests, exams, and assessments, as well as the scores of projects or assignments. Teaching behavior data tracks all teaching activities of teachers on the platform, including the frequency and time of publishing teaching resources, online teaching, answering questions and feedback. Teaching evaluation data consists of peer review, student evaluation and self-evaluation, reflecting the teaching effectiveness of teachers. Learner feedback data collects learners' direct feedback and suggestions on course content, teaching methods and learning experience. Teacher teaching style data determines the characteristics of teachers' teaching style by analyzing teachers' teaching videos, teaching designs and student feedback. Learner learning style data analyzes learners' preferred learning methods based on learners' course selection, learning path and interaction mode. Teaching resource use data tracks learners' and teachers' access to and use of various teaching resources such as e-books, video lectures and online tools. Learning effectiveness feedback data records learners' satisfaction surveys, self-evaluations of learning effectiveness and peer evaluations after the course. Finally, educational effectiveness evaluation data comes from long-term follow-up studies and provides an evaluation of the effectiveness of educational interventions.

[0135] Step S389: Perform feature learning and pattern recognition on the learning-teaching anomaly feature vector set according to the learning-teaching behavior consistency evaluation index, interest-teaching feature matching matrix, style matching score and learning-teaching effectiveness difference data to obtain an intelligent early warning model for teaching anomalies.

[0136] Specifically, the input features of the model can be defined based on the learning-teaching anomaly feature vector set, including learning-teaching behavior consistency evaluation indicators, interest-teaching feature matching matrix, style matching score and learning-teaching effectiveness difference data. Using TensorFlow or PyTorch as a machine learning framework, a neural network structure containing a recurrent neural network, especially a long short-term memory network (LSTM) is designed to process time series data and capture nonlinear relationships. The model consists of multiple LSTM layers and fully connected layers, and the ReLU (Rectified Linear Unit) activation function and batch normalization technology are used to optimize the training process. In order to train the model, it is necessary to prepare labeled training data, use the cross entropy loss function and Adam (Adaptive Moment Estimation) optimizer, and adjust the hyperparameters through cross validation and grid search. After training optimization, an intelligent early warning model for teaching anomalies is finally constructed, which can analyze real-time data streams and identify abnormal patterns.

[0137] The present invention realizes the accurate grasp of teaching characteristics by accurately identifying the uniqueness of the teacher's teaching method, and at the same time evaluates the activeness of classroom interaction and the efficiency of the use of teaching resources. Through multi-dimensional behavior pattern cluster analysis, the teacher's behavior data can be effectively classified to form a representative teaching behavior pattern label, so as to deeply understand the teaching behavior. By automatically identifying and evaluating teaching styles, combined with interaction quality scores and student participation, a comprehensive evaluation of teaching quality and efficiency is carried out to guide the optimization of the teaching process. Through the synchronization analysis of learning-teaching behavior, a reference is provided for personalized teaching, and the matching degree of teaching style and learning style is evaluated, which provides a basis for the docking of teaching methods and learning needs. Through in-depth analysis of learning-teaching effectiveness differences, teaching improvement points are identified, which promotes the improvement of teaching effects. Intelligent identification and early warning of abnormal features can discover and deal with problems in the teaching process in advance, build a teaching environment that can respond in real time and continuously optimize, and improve the personalization and adaptability of education.

[0138] Preferably, step S4 comprises the following steps:

[0139] Step S41: Based on the preset learning service platform, full-cycle data collection of educational effects is performed on learners and teachers to obtain a full-cycle data set of educational effects, wherein the full-cycle data set of educational effects includes learning behavior data, learning achievement data, learning feedback data, teaching behavior data and teaching evaluation data;

[0140] Specifically, the data analysis tools of the preset learning service platform can be used to comprehensively collect data on the entire cycle of educational effectiveness. First, learning behavior data is collected through the platform's database system, which includes learners' login frequency, course access duration, and homework submission records. Using the API interface provided by the platform, learning outcome data, such as test scores and project completion, can be automatically obtained. At the same time, learning feedback data is collected through the platform's feedback collection mechanisms, such as online surveys and instant feedback tools. For teaching behavior data, information is obtained through classroom observation records and teaching activity logs, while teaching evaluation data is collected through final evaluations and peer review systems. Finally, a full-cycle data set for educational effectiveness is obtained.

[0141] Step S42: extracting behavior pattern features from the learning behavior data and the teaching behavior data respectively to obtain a learner behavior pattern feature vector and a teacher behavior pattern feature vector;

[0142] Specifically, the learning behavior data and teaching behavior data can be deeply analyzed to extract the behavior pattern features. Using data mining software such as SPSS Modeler, sequence analysis and cluster analysis are applied to identify learners' behavior patterns. For example, learners' learning path preferences can be identified by analyzing their course access sequence. For teaching behavior data, topic modeling techniques are used to identify teachers' teaching strategies and styles. Using Python's Scikit-learn library, feature extraction is performed on the data, such as using the Bag of Words model to convert teaching behavior descriptions into numerical features, thereby obtaining learner behavior pattern feature vectors and teacher behavior pattern feature vectors.

[0143] Step S43: performing teaching mode recognition on the teaching behavior data to obtain teaching mode classification label data;

[0144] Specifically, text analysis tools can be used to process teaching behavior descriptions and evaluation texts to extract keywords and phrases. Then, classification algorithms, such as support vector machines or random forests, are applied to perform pattern recognition on these features. The model is trained to identify different teaching behavior patterns and to classify teaching behaviors into predefined categories, such as lecture-based, inquiry-based, or discussion-based teaching. Using Python's Natural Language Toolkit (NLTK) library, the text data can be further processed, and the Scikit-learn library can be used to train and evaluate classification models, ultimately obtaining teaching mode classification label data.

[0145] Step S44: performing a comprehensive effect evaluation on the education process according to the learner behavior pattern feature vector, the teacher behavior pattern feature vector, the learning feedback data teaching pattern classification label data and the teaching evaluation data to obtain the comprehensive evaluation result data of the education process, and extracting the education effect label from the comprehensive evaluation result data of the education process to obtain the education effect label set;

[0146] Specifically, the effect of the education process can be comprehensively evaluated by using multivariate analysis methods, combining learner behavior pattern feature vectors, teacher behavior pattern feature vectors, learning feedback data, teaching model classification label data, and teaching evaluation data. Statistical analysis software, such as Pandas and Statsmodels libraries in R or Python, is used to calculate correlations, perform regression analysis, and apply multi-criteria decision analysis methods, such as hierarchical analysis processes, to determine the weights of different factors in the evaluation of education effects. Through these analyses, the comprehensive evaluation result data of the education process is obtained. Next, data mining tools, such as the Scikit-learn library in Python, are used to process the comprehensive evaluation result data and extract education effect labels. These labels include "high effectiveness", "medium effectiveness", or "low effectiveness" classifications, which are obtained through cluster analysis or classification algorithms and can provide clear guidance for the optimization of the education process. Finally, a set of education effect labels is obtained.

[0147] Step S45: Perform supervised learning model training on the education effect full-cycle data set according to the education effect label set to obtain an education effect diagnosis model.

[0148] Specifically, a supervised learning model can be trained using an educational effect label set, with the goal of building a model that can diagnose educational effects. First, use data preprocessing techniques, such as feature scaling and normalization, to prepare a full-cycle dataset of educational effects. Then, select a suitable machine learning algorithm, such as random forest, gradient boosting machine, or deep neural network, to train the model. Use Python's Scikit-learn library or TensorFlow framework to build a model and use cross-validation methods to evaluate the performance of the model. During the model training process, the model parameters need to be continuously adjusted. Through training, an educational effect diagnostic model is finally obtained, which can predict the label of the educational effect based on the input educational process data.

[0149] The present invention provides a detailed data basis for the comprehensive evaluation of educational effects through full-cycle data collection, covering learning behaviors, outcomes, feedback, and teaching behaviors and evaluations. By accurately identifying the behavioral patterns of learners and teachers and extracting feature vectors, quantitative analysis is provided for educational interactions. By intelligently identifying and classifying teaching patterns, educators are helped to choose appropriate teaching methods. Through comprehensive effect evaluation, behavioral patterns, feedback, and evaluation data are used to provide a comprehensive analytical perspective for teaching effects. Through in-depth diagnosis of educational effects and supervised learning model training, in-depth insights are provided for educational improvement. The extraction of educational effect labels clearly guides educational decision-making and the formulation of personalized teaching strategies. Through continuous optimization of the educational process, combined with evaluation results and diagnostic information, educators are encouraged to continuously improve the quality of teaching. Personalized adaptive teaching methods meet the needs of different learners based on learner characteristics and feedback, and overall improve the personalization, efficiency, and quality of education.

[0150] Preferably, step S5 comprises the following steps:

[0151] Step S51: performing heterogeneous model alignment on the learner intelligent portrait recognition model, the teacher intelligent portrait recognition model, the teaching abnormality intelligent early warning model and the education effect diagnosis model to obtain a semantic unified framework of education multi-models;

[0152] Specifically, model management tools can be used to identify and integrate different types of intelligent portraits and early warning models. By defining a unified data exchange format, such as JSON or XML, ensure that the outputs and inputs of different models can be semantically aligned. Using data mapping technology and API interfaces, the data structures of learner intelligent portrait recognition models, teacher intelligent portrait recognition models, teaching abnormality intelligent early warning models, and educational effect diagnosis models are converted and mapped to achieve alignment of heterogeneous models. Next, an ontology approach is used to construct a semantically unified framework for educational multi-models, which defines the relationships between models and a common language for interaction.

[0153] Step S52: performing integrated learning on the educational multi-model semantic unified framework to obtain an educational intelligent fusion model;

[0154] Specifically, feature selection techniques can be used to identify the most representative features in the learner intelligent portrait recognition model, teacher intelligent portrait recognition model, teaching abnormality intelligent warning model, and education effect diagnosis model. Then, machine learning algorithms such as stacking or model fusion technology are used to combine these features and the prediction results of the model. Use Python's Scikit-learn library to achieve model stacking and fusion, and optimize the overall performance by adjusting the weights of each model. Through these steps, an educational intelligent fusion model that integrates the advantages of multiple models is obtained.

[0155] Step S53: Encapsulate the educational intelligence fusion model to obtain a big data model of the learning service platform.

[0156] Specifically, model encapsulation technology, such as Docker containerization, can be used to package the educational intelligent fusion model and its dependent environment to ensure the portability and consistency of the model on different platforms. Next, develop a model service interface (API) using a framework such as Flask or FastAPI to provide access and call functions to the model. Through this API, users can input data and receive the model's prediction results. Finally, the continuous integration / continuous deployment (CI / CD) process is used to automate the update and maintenance of the model. Through these steps, a packaged big data model for the learning service platform is finally obtained.

[0157] The present invention ensures the consistency and interoperability between different educational models by achieving multi-model semantic unification, and provides a basis for data integration and information flow. Through the use of integrated learning methods, the system performance is optimized, and the accuracy and robustness of the educational intelligent fusion model are improved. The constructed intelligent fusion model integrates the multi-dimensional characteristics of learners and teachers, and provides comprehensive support for educational decision-making. The use and maintenance of the model are simplified through model encapsulation technology, and the scalability and deployability of the model are enhanced. In addition, the intelligent fusion model enhances the personalized education experience, customizes education and teaching strategies according to personalized characteristics, and optimizes resource allocation through intelligent configuration of teaching resources, thereby improving teaching efficiency and education quality.

[0158] Preferably, the present invention also provides a learning service platform big data model construction system based on artificial intelligence, which is used to execute the learning service platform big data model construction method based on artificial intelligence as described above, and the learning service platform big data model construction system based on artificial intelligence includes:

[0159] The learner portrait recognition model construction module is used to collect multi-dimensional learner data of learners based on the preset learning service platform to obtain a multimodal data set of learners; based on the multimodal data set of learners, a learner portrait recognition model is constructed for learners to obtain an intelligent learner portrait recognition model;

[0160] The teacher portrait recognition model construction module is used to collect multi-dimensional teacher data on teachers based on the preset learning service platform to obtain a teacher multimodal data set; based on the teacher multimodal data set, a teacher portrait recognition model is constructed to obtain a teacher intelligent portrait recognition model;

[0161] The early warning model construction module is used to collect the teaching process time series data of learners and teachers based on the preset learning service platform to obtain the teaching process time series data set; based on the teaching process time series data set, the teaching early warning model is constructed for the learning-teaching interaction process to obtain the teaching abnormality intelligent early warning model;

[0162] The diagnostic model building module is used to collect the full-cycle data of educational effects of learners and teachers based on the preset learning service platform to obtain a full-cycle data set of educational effects; extract educational effect labels from the full-cycle data set of educational effects to obtain an educational effect label set; and construct an educational effect diagnostic model for the educational process based on the full-cycle data set of educational effects and according to the educational effect label set to obtain an educational effect diagnostic model;

[0163] The model integration module is used to integrate the learner intelligent portrait recognition model, the teacher intelligent portrait recognition model, the teaching abnormality intelligent early warning model and the education effect diagnosis model to obtain the big data model of the learning service platform.

[0164] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0165] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for constructing a big data model for a learning service platform based on artificial intelligence, characterized in that: The following steps are involved: Step S1: Based on the preset learning service platform, multi-dimensional learner data is collected for learners to obtain a multimodal data set of learners; based on the multimodal data set of learners, a learner portrait recognition model is constructed for learners to obtain a learner intelligent portrait recognition model; Step S2: Based on the preset learning service platform, multi-dimensional teacher data is collected for teachers to obtain a teacher multimodal data set; based on the teacher multimodal data set, a teacher portrait recognition model is constructed for teachers to obtain a teacher intelligent portrait recognition model; Step S3: Based on the preset learning service platform, the teaching process time series data of the learners and teachers are collected to obtain the teaching process time series data set; Based on the teaching process time series data set, a teaching early warning model is constructed for the learning-teaching interaction process to obtain an intelligent early warning model for teaching abnormalities; wherein step S3 includes the following steps: Step S31: Based on the preset learning service platform, the teaching process time series data of the learners and teachers are collected to obtain the teaching process time series data set, wherein the teaching process time series data set includes learner behavior data, teacher behavior data and learner-teacher interaction data, the learner behavior data includes learning progress data, learning time data, learning resource type data, learning resource access sequence data, learning interaction data, learning effectiveness feedback data, the teacher behavior data includes teacher teaching method data, classroom interaction frequency data and teaching resource usage data, and the learner-teacher interaction data includes teacher question-answer interaction data and classroom participation data; Step S32: performing learning trajectory recognition on the learning progress data and the learning duration data to obtain learning behavior trajectory data; Step S33: performing behavior pattern recognition on the learning interaction data to obtain learning interaction pattern data; Step S34: conducting a learning effectiveness evaluation on the learning effectiveness feedback data to obtain learning effectiveness indicator data; Step S35: performing multimodal learning behavior fusion on the learning behavior trajectory data, the learning interaction mode data and the learning effectiveness indicator data to obtain the learner's comprehensive learning behavior data; Step S36: tracking the interest preference of the learning resource type data to obtain the learning interest preference data, and identifying the learning style of the learning resource type data and the learning resource access sequence data to obtain the learner's learning style data; Step S37: Perform performance statistics on the learning performance feedback data to obtain learning performance data; Step S38: Based on the teacher behavior data and the learner-teacher interaction data and the learner's comprehensive learning behavior data, learning interest preference data, learner learning style data and learning performance effect data, a teaching warning model is constructed for the learning-teaching interaction process to obtain a teaching abnormality intelligent warning model; wherein, step S38 includes the following steps: Step S381: Identify the teaching characteristics of the teacher's teaching method data to obtain the teacher's teaching characteristics data, calculate the interactive activity of the classroom interaction frequency data to obtain the classroom interactive activity data, and calculate the teaching resource utilization efficiency of the teaching resource usage data to obtain the teaching resource utilization efficiency data; Step S382: performing multi-dimensional behavior pattern clustering on the teacher's teaching characteristics data, classroom interaction activity data, and teaching resource utilization efficiency data to obtain teaching behavior pattern label data; Step S383: Perform teaching style recognition on the teacher's question-answer interaction data to obtain the teacher's teaching style data; Step S384: performing a dialogue quality assessment on the teacher question-and-answer interaction data to obtain a teacher question-and-answer interaction quality score, and performing a student participation calculation on the classroom participation data to obtain student participation data; Step S385: Perform teaching quality efficiency evaluation on the teacher's question-and-answer interaction quality score and the student participation data to obtain teaching quality efficiency evaluation data; Step S386: Analyze the synchronization of learning behavior and teaching behavior on the learner's comprehensive learning behavior data and teaching behavior pattern label data to obtain the learning-teaching behavior consistency evaluation index, and calculate the semantic similarity and topic relevance of the learning interest preference data and the teacher's teaching feature data to obtain the interest-teaching feature matching matrix; Step S387: Performing style matching evaluation on the learner's learning style data and the teacher's teaching style data to obtain a style matching score, and performing effect-efficiency correlation difference analysis on the learning performance effect data and the teaching quality efficiency evaluation data to obtain learning-teaching effectiveness difference data; Step S388: extracting and fusing abnormal features of learner comprehensive learning behavior data, teaching behavior pattern label data, learning interest preference data, teacher teaching feature data, learner learning style data, teacher teaching style data, learning performance effect data and teaching quality efficiency evaluation data according to the learning-teaching behavior consistency evaluation index, interest-teaching feature matching matrix, style matching score and learning-teaching effectiveness difference data to obtain a learning-teaching abnormal feature vector set; Step S389: performing feature learning and pattern recognition on the learning-teaching abnormality feature vector set according to the learning-teaching behavior consistency evaluation index, the interest-teaching characteristic matching matrix, the style matching score and the learning-teaching effectiveness difference data to obtain a teaching abnormality intelligent early warning model; Step S4: Based on the preset learning service platform, the full-cycle data of educational effects of learners and teachers are collected to obtain a full-cycle data set of educational effects; educational effect labels are extracted from the full-cycle data set of educational effects to obtain an educational effect label set; based on the full-cycle data set of educational effects and according to the educational effect label set, an educational effect diagnosis model is constructed for the educational process to obtain an educational effect diagnosis model; Step S5: Integrate the learner intelligent portrait recognition model, the teacher intelligent portrait recognition model, the teaching abnormality intelligent early warning model and the education effect diagnosis model to obtain a big data model of the learning service platform.

2. The method for constructing a big data model for a learning service platform based on artificial intelligence according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: collecting the learning behavior data of the learners based on the preset learning service platform to obtain a learning behavior trajectory data set, wherein the learning behavior trajectory data set includes learning path data, learning time data, learning progress data and learning resource access frequency data; Step S12: Based on the preset learning service platform, the learner's learning resource type data is collected, and the course selection category is collected to obtain the learner's resource preference data and the learner's course category data; Step S13: clustering learners' interest preferences according to learners' resource preference data and learners' course category data to obtain learners' interest preference data; Step S14: Based on the preset learning service platform, learning effect data and learning achievement data of the learner are collected, and basic identity data is collected to obtain learning effect data, learning achievement data and basic identity data of the learner; Step S15: Record the learning behavior trajectory data set, learner resource preference data, learner course category data, learner interest preference data, learning effect data, learning achievement data and learner basic identity data as a learner multimodal data set; Step S16: construct a learner portrait recognition model for the learner based on the learner multimodal data set to obtain a learner intelligent portrait recognition model.

3. The method for constructing a big data model for a learning service platform based on artificial intelligence according to claim 2, characterized in that: Step S16 includes the following steps: Step S161: constructing a portrait framework for the learner to obtain a multi-dimensional portrait framework of the learner, wherein the multi-dimensional portrait framework of the learner includes a learner basic attribute dimension, a learner behavior pattern dimension, a learner learning style dimension, a learner interest tendency dimension, and a learner performance dimension; Step S162: performing sequence pattern mining on the learning path data to obtain learner path feature data, and performing time series analysis on the learning time data to obtain learning time distribution feature data; Step S163: Performing progress quantitative evaluation on the learning progress data to obtain learner progress quantitative data, and performing frequency pattern recognition on the learning resource access frequency data to obtain learning resource access pattern data; Step S164: performing behavior pattern fusion on the learner path feature data, the learning time distribution feature data, the learner progress quantification data and the learning resource access pattern data to obtain learner behavior pattern feature data; Step S165: clustering the learner resource preference data into learning styles to obtain learner learning style feature data; Step S166: performing learning topic identification on the learner's interest preference data to obtain learner's interest topic data, and performing dynamic interest tracking on the learner's interest topic data to obtain learner's interest evolution feature data; Step S167: performing effect quantification scoring on the learning effect data to obtain learner effect quantification data, and performing achievement quantification on the learning achievement data to obtain learner achievement quantification feature data; Step S168: Perform a comprehensive learning performance evaluation on the learner effect quantified data and the learner achievement quantified characteristic data to obtain learner performance characteristic data; Step S169: Based on the learner's basic identity data, the learner's behavior pattern feature data, the learner's learning style feature data, the learner's interest evolution feature data and the learner's performance feature data, the learner's feature fusion and portrait construction are performed according to the learner's multidimensional portrait framework to obtain the learner's personalized multidimensional portrait; the learner's personalized multidimensional portrait is subjected to learning feature mapping and model training to obtain the learner's intelligent portrait recognition model.

4. The method for constructing a big data model for a learning service platform based on artificial intelligence according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: collecting teacher basic attribute data based on a preset learning service platform to obtain a teacher static feature data set, wherein the teacher static feature data set includes teacher gender data, teacher age data, teacher professional background data and teacher teaching field data; Step S22: Based on the preset learning service platform, the teacher's teaching behavior time series data is collected, and the teaching content data is collected to obtain the teacher's teaching behavior data and the teacher's teaching content data; Step S23: performing content separation processing on the teacher's teaching content data to obtain teaching content video data and teaching content text data; Step S24: Based on the preset learning service platform, comprehensive data on the teaching quality and effectiveness of teachers is collected to obtain a teacher quality and effectiveness evaluation data set, wherein the teacher quality and effectiveness evaluation data set includes teacher achievement data, learning effect data, and teaching influence evaluation data; Step S25: Record the teacher static feature data set, teacher teaching behavior data, teaching content video data, teaching content text data and teacher quality and effectiveness evaluation data set as the teacher multimodal data set; Step S26: construct a teacher portrait recognition model for teachers based on the teacher multimodal data set to obtain a teacher intelligent portrait recognition model.

5. The method for constructing a big data model for a learning service platform based on artificial intelligence according to claim 4, characterized in that: Step S26 includes the following steps: Step S261: constructing a teacher portrait framework to obtain a teacher multi-dimensional portrait framework, wherein the teacher multi-dimensional portrait framework includes a teacher static attribute dimension, a teaching resource utilization dimension, a teaching style dimension, and a teaching effectiveness dimension; Step S262: extracting and encoding the teacher's gender data, teacher's age data, teacher's professional background data and teacher's teaching field data to obtain the teacher's basic information code; Step S263: performing usage statistics on the teaching content video data and the teaching content text data respectively to obtain video resource usage data and text resource usage data, and performing resource utilization efficiency calculation on the video resource usage data and the text resource usage data to obtain teaching resource utilization rate data; Step S264: Perform teaching style identification on the teaching content video data to obtain preliminary teaching style classification data; and annotate and verify the preliminary teaching style classification data to obtain teacher teaching style data, wherein the teacher teaching style data is any one or more combinations of directional, heuristic, personalized, interactive, and practical; Step S265: Evaluate the teacher achievement data for achievement influence, and obtain the teacher achievement influence score; Step S266: performing a learning effectiveness analysis on the learning effectiveness data to obtain a student learning effectiveness indicator set, and performing a learning effectiveness score on the student learning effectiveness indicator set to obtain a student comprehensive learning effectiveness score; Step S267: Perform multi-dimensional influence evaluation on the teaching influence evaluation data to obtain a comprehensive teacher influence score, wherein the multi-dimensional influence evaluation specifically includes student satisfaction dimension, peer evaluation dimension, social influence dimension, and scientific research contribution dimension; Step S268: Perform multi-dimensional effectiveness fusion on the teacher's achievement influence score, the student's comprehensive learning effect score and the teacher's comprehensive influence score to obtain the teacher's comprehensive effectiveness index; Step S269: Based on the teacher basic information coding, teaching resource utilization data, teacher teaching style data and teacher comprehensive effectiveness index, multi-dimensional features of teachers are fused according to the teacher multi-dimensional portrait framework to obtain the teacher comprehensive feature vector, and deep learning is performed on the teacher comprehensive feature vector to obtain the teacher intelligent portrait recognition model.

6. The method for constructing a big data model for a learning service platform based on artificial intelligence according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: Based on the preset learning service platform, full-cycle data collection of educational effects is performed on learners and teachers to obtain a full-cycle data set of educational effects, wherein the full-cycle data set of educational effects includes learning behavior data, learning achievement data, learning feedback data, teaching behavior data and teaching evaluation data; Step S42: extracting behavior pattern features from the learning behavior data and the teaching behavior data respectively to obtain a learner behavior pattern feature vector and a teacher behavior pattern feature vector; Step S43: performing teaching mode recognition on the teaching behavior data to obtain teaching mode classification label data; Step S44: performing a comprehensive effect evaluation on the education process according to the learner behavior pattern feature vector, the teacher behavior pattern feature vector, the learning feedback data, the teaching pattern classification label data and the teaching evaluation data to obtain the education process comprehensive evaluation result data, and extracting the education effect label from the education process comprehensive evaluation result data to obtain the education effect label set; Step S45: Perform supervised learning model training on the education effect full-cycle data set according to the education effect label set to obtain an education effect diagnosis model.

7. The method for constructing a big data model for a learning service platform based on artificial intelligence according to claim 1, characterized in that: Step S5 includes the following steps: Step S51: performing heterogeneous model alignment on the learner intelligent portrait recognition model, the teacher intelligent portrait recognition model, the teaching abnormality intelligent early warning model and the education effect diagnosis model to obtain a semantic unified framework of education multi-models; Step S52: performing integrated learning on the educational multi-model semantic unified framework to obtain an educational intelligent fusion model; Step S53: Encapsulate the educational intelligence fusion model to obtain a big data model of the learning service platform.

8. A big data model construction system for a learning service platform based on artificial intelligence, characterized in that: Used to execute the method for building a big data model of a learning service platform based on artificial intelligence as claimed in claim 1, the big data model building system of the learning service platform based on artificial intelligence includes: The learner portrait recognition model construction module is used to collect multi-dimensional learner data of learners based on the preset learning service platform to obtain a multimodal data set of learners; based on the multimodal data set of learners, a learner portrait recognition model is constructed for learners to obtain an intelligent learner portrait recognition model; The teacher portrait recognition model construction module is used to collect multi-dimensional teacher data on teachers based on the preset learning service platform to obtain a teacher multimodal data set; based on the teacher multimodal data set, a teacher portrait recognition model is constructed to obtain a teacher intelligent portrait recognition model; The early warning model construction module is used to collect the teaching process time series data of learners and teachers based on the preset learning service platform to obtain the teaching process time series data set; based on the teaching process time series data set, the teaching early warning model is constructed for the learning-teaching interaction process to obtain the teaching abnormality intelligent early warning model; The diagnostic model building module is used to collect the full-cycle data of educational effects of learners and teachers based on the preset learning service platform to obtain a full-cycle data set of educational effects; extract educational effect labels from the full-cycle data set of educational effects to obtain an educational effect label set; and construct an educational effect diagnostic model for the educational process based on the full-cycle data set of educational effects and according to the educational effect label set to obtain an educational effect diagnostic model; The model integration module is used to integrate the learner intelligent portrait recognition model, the teacher intelligent portrait recognition model, the teaching abnormality intelligent early warning model and the education effect diagnosis model to obtain the big data model of the learning service platform.

Images