Intelligent adaptation personalized education pushing system based on big data
Through big data analysis and natural language processing technology, combined with learning history and preferences, students are provided with a personalized education push system, which solves the problem of inaccurate content recommendations in the traditional education system and improves learning efficiency and satisfaction.
Patent Information
- Application Number
- CN202410977765.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-22
- Publication Date
- 2024-11-22
AI Technical Summary
The traditional personalized education push system lacks in-depth understanding and precise grasp of students' personalized needs, resulting in large differences between recommended content and learning needs and interests, affecting learning experience and efficiency.
Adopt intelligent adaptive personalized education push system based on big data, analyze learner data and resources through deep learning and natural language processing technology, combine learning history, preferences and current status, use recommendation algorithms to recommend precise content, and optimize through user feedback Push policy.
It realizes accurate content recommendations for learners, improves learning efficiency and satisfaction, enables students to quickly find content that meets their needs, and improves their learning experience.
Smart Images

Figure CN119003864A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of personalized education technology, and in particular to an intelligent adaptive personalized education push system based on big data. Background Art
[0002] Education is the whole process of training a new generation to prepare for social life, and it is also a key link in the inheritance and development of human social production experience. It mainly refers to the process of schools training children, teenagers and young people of school age, but it is not limited to school education, but also includes various forms such as family education and social education. With the rapid development of information technology, the field of education has also ushered in unprecedented changes. The traditional education model has gradually shifted to personalized education that pays more attention to individual development and learning effects. Personalized education emphasizes student-centeredness and provides customized educational resources and teaching methods based on the characteristics, interests, learning styles and needs of each student. This education model aims to stimulate students' learning motivation, improve learning efficiency and promote all-round development. The application of personalized education in education has a significant effect, which can not only improve students' academic performance, but also enhance students' learning motivation. Specifically, personalized education can maximize students' potential and advantages, help students overcome learning obstacles and improve learning effects. At the same time, personalized education can also reduce students' learning pressure, promote their mental health, and cultivate their emotional intelligence and psychological quality.
[0003] However, in the process of implementing personalized education, the traditional personalized education push system still has certain defects when used. The traditional personalized education push system often recommends content based on simple rules or manual settings, lacks in-depth understanding and accurate grasp of students' personalized needs, resulting in the push content may be quite different from students' actual learning needs and interests, affecting the learning experience and learning effect. The existence of a large number of educational resources makes it difficult for learners to choose appropriate learning content, making it difficult for students to quickly find content that meets their needs, reducing learning efficiency. Therefore, it is necessary to propose an intelligent adaptive personalized education push system based on big data to solve the problems in the existing technology. Summary of the invention
[0004] The purpose of this invention is to make up for the shortcomings of the existing technology and provide an intelligent adaptive personalized education push system based on big data. It can analyze learner data and learning resources through deep learning and natural language processing technology, achieve accurate content recommendation, and improve learners' learning efficiency and satisfaction.
[0005] In order to solve the above technical problems, the present invention provides the following technical solutions: a big data-based intelligent adaptive personalized education push system, the system comprising:
[0006] Data collection module, used to collect learners' learning history, preferences, learning behavior data, as well as the content, difficulty and applicable objects of learning resources;
[0007] Data analysis module, which uses deep learning and natural language processing technology to process and analyze the collected data to identify learners' learning needs and interests, as well as the characteristics of learning resources;
[0008] Content recommendation module: Based on the processing results of the data analysis module, this module uses a recommendation algorithm to recommend content that meets learners' learning needs and interests. The recommendation algorithm comprehensively considers learners' learning history, preferences, and current learning status;
[0009] The user feedback module is used to collect learners' feedback on the recommended content and use the feedback data as input to the data analysis module to improve the recommendation algorithm and optimize the push strategy.
[0010] Furthermore, the data analysis module includes a natural language processing submodule for parsing the content of learning resources and extracting key information such as topics, knowledge points and difficulty levels.
[0011] Furthermore, the data analysis module also includes a deep learning model training submodule, which uses the learner's historical learning data to train the model for identifying the learner's learning needs and interests.
[0012] Furthermore, the specific execution steps of the data collection module include:
[0013] Initialize the data collection environment, install and configure the necessary data collection tools and technology stack, and determine the sources of data collection;
[0014] Determine the data source. Select the appropriate data source based on the type and source of the data to be collected. For learner data, determine the specific methods and ways to collect learning history, preferences, and learning behavior data. For learning resource data, determine the specific methods and ways to collect content, difficulty, and applicable object information.
[0015] Connect to data source, connect to the selected data source, and establish a connection to communicate with it;
[0016] Data acquisition: Obtain required data from data sources. Learner data is collected through learners’ activities on the platform, questionnaires, and usage of learning tools. Learning resource data is obtained through web crawlers and API interfaces.
[0017] Furthermore, the specific execution steps of the data collection module also include:
[0018] Data preprocessing: cleaning, deduplication and formatting the collected data to ensure the accuracy, consistency and availability of the data;
[0019] Data storage: storing the preprocessed data in a specified database;
[0020] Data verification and monitoring: verify the stored data to ensure the integrity and accuracy of the data, monitor the data collection process, and ensure the stability and reliability of data collection.
[0021] Furthermore, the specific steps of data preprocessing by the data collection module include:
[0022] Missing value processing: check which fields and features in the data set have missing values, and use the mean, median, and mode of the feature column to fill the missing values. For the continuous variable X, its missing values are filled with the mean of the column: Use interpolation algorithms to estimate missing values. If the number of missing values is small, choose to delete rows and columns containing missing values.
[0023] Outlier processing: Use statistical indicators and clustering algorithms to detect outliers, based on the Z-score threshold method, where If |Z| exceeds a certain threshold, it is considered an outlier, and the K-means clustering algorithm is used to identify clusters and data points that may be abnormal;
[0024] Data format conversion, converting data formats into a format that can be understood by computers;
[0025] Data normalization converts data variables to the same value range to eliminate the influence between variables. The normalization formula is: Where X max and X min are the minimum and maximum values of feature X respectively;
[0026] Feature selection, selecting important features based on statistical indicators and model evaluation. Filtering methods, packaging methods and embedding methods are commonly used feature selection techniques.
[0027] Furthermore, the specific execution steps of the natural language processing submodule in the data analysis module include:
[0028] Text preprocessing: removing noise from the text and performing basic text processing such as word segmentation and part-of-speech tagging;
[0029] Feature extraction, using the Word2Vec word embedding algorithm to convert words in the text into dense vectors in a high-dimensional vector space. The objective function of Word2Vec is: Among them, w ois the word vector of the target word, h is the hidden layer representation of the center word, W neg is the set of words obtained by negative sampling, σ is the sigmoid function;
[0030] Key information extraction, named entity recognition: use conditional random fields, bidirectional long short-term memory networks combined with CRF algorithm to identify topics and author entities in the text, topic model: use latent Dirichlet allocation algorithm to identify the topic distribution of the text, topic distribution: θ d ~Dir(α), where θ is the topic distribution of document d, α is a hyperparameter, and the word distribution: β k ~Dir(η), where β is the word distribution of topic k and η is a hyperparameter;
[0031] Text classification: Use a recurrent neural network combined with a classification layer to determine the difficulty level and knowledge point classification of the text.
[0032] Furthermore, the specific execution steps of the deep learning model training submodule in the data analysis module include:
[0033] Data preparation: collating historical learning data of learners’ learning behaviors, grades, and feedback, constructing input features of learning behavior sequences, historical grades, and target variables of learning needs and interests;
[0034] Model construction, build a deep learning model of recurrent neural network according to the characteristics of the task, the hidden state update formula in the training of recurrent neural network model is: h t =σ(W xh x t +W hh h t-1 +b h ), where h t is the hidden state at time step t, x t is the input at time step t, W xh and W hh is the weight matrix, b h is the bias term, σ is the activation function;
[0035] Design the network structure, including the input layer, hidden layer and output layer. The output layer formula in the recurrent neural network is: t =softmax(W hy h t +b y ) where y t is the output probability distribution at time step t, W hy is the weight matrix, b y is the bias term, and the softmax function is used to convert the output into a probability distribution;
[0036] Model training, using the learner's historical learning data as the training set, trains the model through the back propagation algorithm and the gradient descent algorithm, the goal is to minimize the loss function to improve the prediction performance of the model;
[0037] Model evaluation and optimization: Use the validation set to evaluate the model, calculate the accuracy, recall, and F1 value indicators, adjust the model parameters, optimize the network structure, and try different models based on the evaluation results.
[0038] Furthermore, the specific execution steps of the content recommendation module include:
[0039] Data preparation: obtaining learners’ learning history, learning preferences and current learning status data from the data analysis module, and collating metadata on the content, subject and difficulty level of resources in the learning resource library;
[0040] Feature representation, which represents the behavior sequence, preference labels, current learning status, and content, topic, and difficulty of the learner and learning resource in the learning history as a feature vector;
[0041] Similarity calculation, using the recommendation algorithm to calculate the similarity between the learner feature vector and the resource feature vector, is implemented through the cosine similarity method. The cosine similarity formula is: Among them, A and B are learner feature vector and resource feature vector respectively. i and B i are the elements in the vector, and n is the dimension of the vector;
[0042] Recommendation list generation: sort learning resources according to similarity, generate a recommendation list, and select the resources with the highest similarity and matching degree as the recommendation results;
[0043] Real-time adjustment and optimization: adjust the recommendation algorithm in real time according to learners' feedback and learning behavior, and use machine learning methods to continuously optimize the parameters and structure of the recommendation model to improve the accuracy and satisfaction of the recommendation.
[0044] Furthermore, the specific execution steps of the user feedback module include:
[0045] Create an intuitive and easy-to-use interface that allows learners to rate their satisfaction with recommended content, submit learning achievement reports, leave comments and ratings, and capture and store users' feedback data as they interact with the feedback interface;
[0046] Remove invalid and duplicate feedback data, ensure data accuracy and consistency, and classify feedback data for subsequent analysis;
[0047] Statistics were collected on satisfaction evaluations, the average constant temperature satisfaction distribution was calculated, and natural language processing techniques were used to perform sentiment analysis and topic extraction on the reviews;
[0048] The processed feedback data is passed to the data analysis module as a basis for improving the recommendation algorithm and optimizing the push strategy.
[0049] Compared with the existing technology, this intelligent adaptive personalized education push system based on big data has the following beneficial effects:
[0050] The present invention analyzes and processes learner data and learning resources by using deep learning and natural language processing technology, and recommends content that meets the learners' learning needs and interests based on their learning history, preferences and current learning status, thereby achieving accurate content recommendation and greatly improving the learners' learning efficiency and satisfaction. The recommendation algorithm comprehensively considers the learners' learning history, preferences and current learning status to achieve personalized education push, enabling learners to move forward on the learning track that best suits them, enabling students to quickly find content that meets their needs and improve learning efficiency.
[0051] Other advantages, objectives and features of the present invention will be set forth in part in the following description and, in part, will be apparent to those skilled in the art based on an examination of the following or may be taught from the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS
[0052] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.
[0053] Figure 1 This is a structural diagram of an intelligent adaptive personalized education push system based on big data. DETAILED DESCRIPTION
[0054] The technical solutions in the embodiments of the present invention are described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0055] Embodiment 1
[0056] A big data-based intelligent adaptive personalized education push system, the system comprising:
[0057] Data collection module, used to collect learners' learning history, preferences, learning behavior data, as well as the content, difficulty and applicable objects of learning resources;
[0058] The data collection module also includes functions for preprocessing and standardizing the collected data to ensure data quality and consistency;
[0059] The specific execution steps of the data collection module include: initializing the data collection environment, installing and configuring the necessary data collection tools and technology stacks, and determining the source of data collection, such as databases, API interfaces, and web pages; determining the data source, selecting the appropriate data source according to the data type and source to be collected, and for learner data, determining the specific methods and ways to collect learning history, preferences, and learning behavior data; for learning resource data, determining the specific methods and ways to collect content, difficulty, and applicable object information;
[0060] Connecting to the data source, connecting to the selected data source and establishing a communication connection with it involves the use of corresponding technologies and protocols, such as database connection strings and API interface calls; data acquisition, obtaining the required data from the data source. For learner data, it is collected through learners' activities on the platform, questionnaires, and the use of learning tools. For learning resource data, it is obtained through web crawlers and API interfaces.
[0061] Data preprocessing: cleaning, deduplication and formatting the collected data to ensure the accuracy, consistency and availability of the data;
[0062] Missing value processing: check which fields and features in the data set have missing values, and use the mean, median, and mode of the feature column to fill the missing values. For the continuous variable X, its missing values are filled with the mean of the column: Use interpolation algorithms (such as linear interpolation and polynomial interpolation) to estimate missing values. If the number of missing values is small, select and delete rows and columns containing missing values.
[0063] Outlier processing: Use statistical indicators (such as mean, standard deviation) and clustering algorithms to detect outliers, based on the Z-score threshold method, where If |Z| exceeds a certain threshold, it is considered an outlier, and the K-means clustering algorithm is used to identify clusters and data points that may be abnormal; data format conversion, converting the data format into a format that can be understood by the computer, such as date format, text encoding; data normalization, converting data variables to the same value range to eliminate the influence between variables, normalization formula: Where X max and X minare the minimum and maximum values of feature X respectively;
[0064] Data dimensionality reduction, using algorithms such as principal component analysis (PCA) to reduce the dimensionality of the data. PCA usually involves linear transformation, mapping the original data into a new coordinate system, where the first new coordinate is the direction with the largest variance in the original data, the second new coordinate is the direction with the largest variance in the plane orthogonal to the first coordinate, and so on; Feature selection, selecting important features based on statistical indicators (such as variance, correlation coefficient) and model evaluation. Filtering methods, packaging methods, and embedding methods are commonly used feature selection techniques. Data encoding (if categorical variables are involved): Use One-Hot Encoding and Label Encoding methods to convert categorical features into a format suitable for model processing.
[0065] Data storage: store the preprocessed data in the designated database. The database design should take into account the data type, structure, and query efficiency factors. Data verification and monitoring: verify the stored data to ensure the integrity and accuracy of the data, monitor the data collection process, and ensure the stability and reliability of data collection.
[0066] Log records record key information in the data collection process, such as data source, collection time, and data volume. Log records are helpful for subsequent data analysis and troubleshooting. Data synchronization and updating ensure data synchronization and consistency for cross-platform and distributed data sources. Data is updated regularly and in real time to maintain data timeliness and accuracy. Security assurance ensures data security and privacy during the data collection process, complies with relevant laws, regulations and privacy policies, and protects user privacy and data security.
[0067] Embodiment 2
[0068] Data analysis module, which uses deep learning and natural language processing technology to process and analyze the collected data to identify learners' learning needs and interests, as well as the characteristics of learning resources;
[0069] The data analysis module includes a natural language processing submodule for parsing the content of learning resources and extracting key information such as topics, knowledge points and difficulty levels.
[0070] The specific execution steps of the natural language processing submodule in the data analysis module include: text preprocessing, removing noise in the text, such as HTML tags, special characters, stop words, and performing basic text processing such as word segmentation and part-of-speech tagging; feature extraction, using the Word2Vec word embedding algorithm to convert words in the text into dense vectors in a high-dimensional vector space. The objective function of Word2Vec is: Among them, w o is the word vector of the target word, h is the hidden layer representation of the center word, W neg is the set of words obtained by negative sampling, σ is the sigmoid function;
[0071] Key information extraction, named entity recognition (NER): use conditional random field (CRF), bidirectional long short-term memory network (BiLSTM) combined with CRF (BiLSTM-CRF) algorithm to identify the topic and author entities in the text, topic modeling: use the latent Dirichlet allocation algorithm to identify the topic distribution of the text, topic distribution: θ d ~Dir(α), where θ is the topic distribution of document d, α is a hyperparameter, and the word distribution: β k ~Dir(η), where β is the word distribution of topic k and η is a hyperparameter; Text classification: Use a recurrent neural network (RNN) combined with a classification layer (such as a softmax layer) to determine the difficulty level and knowledge point classification of the text.
[0072] The data analysis module also includes a deep learning model training submodule, which uses the learner's historical learning data to train a model for identifying the learner's learning needs and interests.
[0073] The specific execution steps of the deep learning model training submodule in the data analysis module include: data preparation, collating historical learning data of learners' learning behaviors (such as video viewing time, question-solving records), grades and feedback, constructing learning behavior sequences, input features of historical grades and target variables of learning needs and interests;
[0074] Model construction, build a deep learning model of recurrent neural network according to the characteristics of the task, the hidden state update formula in the training of recurrent neural network model is: h t =σ(W xh x t +W hh h t-1 +b h ), where h t is the hidden state at time step t, x t is the input at time step t, W xh and W hh is the weight matrix, b h is the bias term, σ is the activation function; design the network structure, including the input layer, hidden layer (which can contain multiple RNN layers, attention mechanism) and output layer (such as softmax layer for classification), the output layer formula in the recurrent neural network is: t =softmax(W hy h t +b y ) where yt is the output probability distribution at time step t, W hy is the weight matrix, b y is the bias term, and the softmax function is used to convert the output into a probability distribution;
[0075] Model training uses the learner's historical learning data as the training set, and trains the model through the back propagation algorithm and the gradient descent algorithm. The goal is to minimize the loss function (such as the cross entropy loss function) to improve the prediction performance of the model; model evaluation and optimization uses the validation set to evaluate the model, calculates the accuracy, recall rate, and F1 value indicators, adjusts the model parameters according to the evaluation results, optimizes the network structure, and tries different models.
[0076] Embodiment 3
[0077] Content recommendation module: Based on the processing results of the data analysis module, this module uses a recommendation algorithm to recommend content that meets learners' learning needs and interests. The recommendation algorithm comprehensively considers learners' learning history, preferences, and current learning status;
[0078] The recommendation algorithm comprehensively considers the learner's learning history, learning preferences and current learning status, and achieves accurate content push by calculating the similarity and matching between learning resources and learners. The content recommendation module also has the ability to adjust in real time, and can continuously optimize the recommendation effect based on learners' feedback and learning behavior;
[0079] The specific execution steps of the content recommendation module include: data preparation, obtaining the learner's learning history, learning preferences and current learning status data from the data analysis module, and collating the metadata of the content, theme and difficulty level information of the resources in the learning resource library; feature representation, representing the behavior sequence, preference label, current learning status and content, theme and difficulty of the learner and learning resource in the learning history as a feature vector;
[0080] Similarity calculation, using the recommendation algorithm to calculate the similarity between the learner feature vector and the resource feature vector, is implemented through the cosine similarity method. The cosine similarity formula is: Among them, A and B are learner feature vector and resource feature vector respectively. i and B i are the elements in the vector, and n is the dimension of the vector;
[0081] Recommendation list generation: sort learning resources according to similarity, generate a recommendation list, and select the resources with the highest similarity and matching as the recommendation results; real-time adjustment and optimization: adjust the recommendation algorithm in real time according to learners' feedback and learning behavior. For example, when a learner likes a recommended resource, the similarity between the resource and other resources can be increased; when a learner ignores a recommendation, the recommendation priority of the resource can be lowered. Use machine learning methods (such as reinforcement learning and online learning) to continuously optimize the parameters and structure of the recommendation model to improve the accuracy and satisfaction of the recommendations.
[0082] Embodiment 4
[0083] The user feedback module is used to collect learners' feedback on the recommended content and use the feedback data as input to the data analysis module to improve the recommendation algorithm and optimize the push strategy.
[0084] This module is configured to collect learners' feedback data on recommended content, including satisfaction evaluation and learning outcome reports. The user feedback module passes the collected data to the data analysis module as an important basis for improving the recommendation algorithm and optimizing the push strategy. In addition, the user feedback module also provides an interface for learners to interact with the system, such as comment and rating functions, so that learners can express their opinions and suggestions.
[0085] The specific execution steps of the user feedback module include: creating an intuitive and easy-to-use interface that allows learners to rate the satisfaction of recommended content (such as a five-star rating system), submit learning achievement reports, leave comments and ratings, and capture and store users' feedback data when they interact with the feedback interface; removing invalid and duplicate feedback data to ensure the accuracy and consistency of the data, and classifying the feedback data, such as satisfaction evaluation, comments, and learning achievements, for subsequent analysis;
[0086] Statistics are collected on satisfaction evaluations, the average satisfaction distribution is calculated, and natural language processing technology is used to perform sentiment analysis and topic extraction on comments; the processed feedback data is passed to the data analysis module as a basis for improving the recommendation algorithm and optimizing the push strategy;
[0087] Provide comment and rating interactive functions to ensure that learners can easily express their opinions and suggestions, and use deep learning-based sentiment analysis algorithms to judge the emotional tendency of comments (such as positive, negative, neutral).
[0088] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the same elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.
Claims
1. An intelligent adaptive personalized education push system based on big data, It is characterized in that The system includes: Data collection module, used to collect learners' learning history, preferences, learning behavior data, as well as the content, difficulty and applicable objects of learning resources; Data analysis module, which uses deep learning and natural language processing technology to process and analyze the collected data to identify learners' learning needs and interests, as well as the characteristics of learning resources; Content recommendation module: Based on the processing results of the data analysis module, this module uses a recommendation algorithm to recommend content that meets learners' learning needs and interests. The recommendation algorithm comprehensively considers learners' learning history, preferences, and current learning status; The user feedback module is used to collect learners' feedback on the recommended content and use the feedback data as input to the data analysis module to improve the recommendation algorithm and optimize the push strategy.
2. According to claim 1, a big data-based intelligent adaptive personalized education push system, It is characterized in that The data analysis module includes a natural language processing submodule for parsing the content of learning resources and extracting key information such as topics, knowledge points and difficulty levels.
3. According to the big data-based intelligent adaptive personalized education push system of claim 1, It is characterized in that The data analysis module also includes a deep learning model training submodule, which uses the learner's historical learning data to train a model for identifying the learner's learning needs and interests.
4. According to claim 1, a big data-based intelligent adaptive personalized education push system, It is characterized in that The specific execution steps of the data collection module include: Initialize the data collection environment, install and configure the necessary data collection tools and technology stack, and determine the sources of data collection; Determine the data source. Select the appropriate data source based on the type and source of the data to be collected. For learner data, determine the specific methods and ways to collect learning history, preferences, and learning behavior data. For learning resource data, determine the specific methods and ways to collect content, difficulty, and applicable object information. Connect to data source, connect to the selected data source, and establish a connection to communicate with it; Data acquisition: Obtain required data from data sources. Learner data is collected through learners’ activities on the platform, questionnaires, and usage of learning tools. Learning resource data is obtained through web crawlers and API interfaces.
5. According to claim 4, a big data-based intelligent adaptive personalized education push system, It is characterized in that The specific execution steps of the data collection module also include: Data preprocessing: cleaning, deduplication and formatting of collected data to ensure data accuracy, consistency and availability; Data storage: storing the preprocessed data in a specified database; Data verification and monitoring: verify the stored data to ensure the integrity and accuracy of the data, monitor the data collection process, and ensure the stability and reliability of data collection.
6. According to claim 5, a big data-based intelligent adaptive personalized education push system, It is characterized in that The specific steps of data preprocessing by the data collection module include: Missing value processing: check which fields and features in the data set have missing values, and use the mean, median, and mode of the feature column to fill the missing values. For the continuous variable X, its missing values are filled with the mean of the column: Use interpolation algorithms to estimate missing values. If the number of missing values is small, choose to delete rows and columns containing missing values. Outlier processing: Use statistical indicators and clustering algorithms to detect outliers, based on the Z-score threshold method, where If |Z| exceeds a certain threshold, it is considered an outlier, and the K-means clustering algorithm is used to identify clusters and data points that may be abnormal; Data format conversion, converting data formats into a format that can be understood by computers; Data normalization converts data variables to the same value range to eliminate the influence between variables. The normalization formula is: Where X max and X min are the minimum and maximum values of feature X respectively; Feature selection, selecting important features based on statistical indicators and model evaluation. Filtering methods, packaging methods and embedding methods are commonly used feature selection techniques.
7. According to claim 2, a big data-based intelligent adaptive personalized education push system, It is characterized in that The specific execution steps of the natural language processing submodule in the data analysis module include: Text preprocessing: removing noise from the text and performing basic text processing such as word segmentation and part-of-speech tagging; Feature extraction, using the Word2Vec word embedding algorithm to convert words in the text into dense vectors in a high-dimensional vector space. The objective function of Word2Vec is: Among them, w o is the word vector of the target word, h is the hidden layer representation of the center word, W neg is the set of words obtained by negative sampling, σ is the sigmoid function; Key information extraction, named entity recognition: use conditional random fields, bidirectional long short-term memory networks combined with CRF algorithm to identify topics and author entities in the text, topic model: use latent Dirichlet allocation algorithm to identify the topic distribution of the text, topic distribution: θ d ~Dir(α), where θ is the topic distribution of document d, α is a hyperparameter, and the word distribution: β k ~Dir(η), where β is the word distribution of topic k and η is a hyperparameter; Text classification: Use a recurrent neural network combined with a classification layer to determine the difficulty level and knowledge point classification of the text.
8. According to claim 3, a big data-based intelligent adaptive personalized education push system, It is characterized in that The specific execution steps of the deep learning model training submodule in the data analysis module include: Data preparation: collating historical learning data of learners’ learning behaviors, grades, and feedback, constructing input features of learning behavior sequences, historical grades, and target variables of learning needs and interests; Model construction, build a deep learning model of recurrent neural network according to the characteristics of the task, the hidden state update formula in the training of recurrent neural network model is: h t =σ(W xh x t +W hh h t-1 +b h ), where h t is the hidden state at time step t, x t is the input at time step t, W xh and W hh is the weight matrix, b h is the bias term, σ is the activation function; Design the network structure, including the input layer, hidden layer and output layer. The output layer formula in the recurrent neural network is: t =softmax(W hy h t +b y ) where y t is the output probability distribution at time step t, W hy is the weight matrix, b y is the bias term, and the softmax function is used to convert the output into a probability distribution; Model training, using the learner's historical learning data as the training set, trains the model through the back propagation algorithm and the gradient descent algorithm, the goal is to minimize the loss function to improve the prediction performance of the model; Model evaluation and optimization: Use the validation set to evaluate the model, calculate the accuracy, recall, and F1 value indicators, adjust the model parameters, optimize the network structure, and try different models based on the evaluation results.
9. According to claim 1, a big data-based intelligent adaptive personalized education push system, It is characterized in that The specific execution steps of the content recommendation module include: Data preparation: obtaining learners’ learning history, learning preferences and current learning status data from the data analysis module, and collating metadata on the content, subject and difficulty level of resources in the learning resource library; Feature representation, which represents the behavior sequence, preference labels, current learning status, and content, topic, and difficulty of the learner and learning resource in the learning history as a feature vector; Similarity calculation, using the recommendation algorithm to calculate the similarity between the learner feature vector and the resource feature vector, is implemented through the cosine similarity method. The cosine similarity formula is: Among them, A and B are learner feature vector and resource feature vector respectively. i and B i are the elements in the vector, and n is the dimension of the vector; Recommendation list generation: sort learning resources according to similarity, generate a recommendation list, and select the resources with the highest similarity and matching degree as the recommendation results; Real-time adjustment and optimization: adjust the recommendation algorithm in real time according to learners' feedback and learning behavior, and use machine learning methods to continuously optimize the parameters and structure of the recommendation model to improve the accuracy and satisfaction of the recommendation.
10. According to claim 1, a big data-based intelligent adaptive personalized education push system, It is characterized in that The specific execution steps of the user feedback module include: Create an intuitive and easy-to-use interface that allows learners to rate their satisfaction with recommended content, submit learning achievement reports, leave comments and ratings, and capture and store users' feedback data as they interact with the feedback interface; Remove invalid and duplicate feedback data, ensure data accuracy and consistency, and classify feedback data for subsequent analysis; Statistics were collected on satisfaction evaluations, the average constant temperature satisfaction distribution was calculated, and natural language processing techniques were used to perform sentiment analysis and topic extraction on the reviews; The processed feedback data is passed to the data analysis module as a basis for improving the recommendation algorithm and optimizing the push strategy.
Citation Information
Cited By
Intelligent learning behavior analysis method and system
CN119295275A
Learning resource pushing method and system based on artificial intelligence
CN121388261A
