CN113378076B - Online education-oriented learner collaborative learning social relationship construction method - Google Patents

Abstract

The invention discloses a construction method of a learner collaborative learning social relationship facing online education, which comprises the following steps: acquiring basic information data of a learner in an online education platform, and extracting attribute characteristics of the basic information data by adopting a sequence analysis method; performing behavior analysis on the basic behavior characteristics in the attribute characteristics to label the partnership; processing the partnership labels by using a k-Means clustering algorithm to obtain the learning community clustering discovery of the learner; constructing a general neural network according to the learning community clustering discovery of the learner, training the collaborative learning social relationship neural network of the learner by using the attribute characteristics to obtain the collaborative learning social relationship neural network of the learner, and further obtaining a current buddy recommendation list; and (3) converging the learner partner network and rearranging the partner recommendation list based on the scale-free network theory and the social triangle theory. The method can recommend a proper learning partner for the learner, and further enhance the learning effect of the learner on the platform.

Description

Online education-oriented learners' collaborative learning social relationship construction method

technical field

The invention relates to the technical field of online education platforms, in particular to a construction method for online education-oriented learners to collaboratively learn social relationships.

Background technique

With the rapid development of information technology, people's production, living and working methods are changing, and the world we live in is becoming increasingly materialized, interconnected and intelligent. The educational goals and teaching models under the background of the traditional industrial age are no longer suitable for the needs of today's information age and big data age. Various institutions and researchers have put forward many new innovations for the realization of smart education, such as the application and practice of online classrooms and flipped classrooms, changing the previous teaching mode, so that learners can get more independent learning time. and corresponding learning resources to make the learning process more efficient. With the introduction and gradual popularization of new education model e-Learning and MOOC (Massive Open Online Courses, Massive Open Online Courses), more learners have the opportunity to visit and learn on the online platform. The platform has accumulated a large amount of teaching behavior data and knowledge resources, which provides a good foundation for the updating and improvement of the platform itself. However, compared with traditional school education, online education not only provides richer learning resources, but also limits the learner's learning platform to the Internet. One of the most important learners in traditional school learning and life is Intercommunication learning methods are fragmented due to the mutual isolation of learners in physical space. Each learner is basically learning alone in the process of learning. However, in the face of difficult courses, learners cannot solve their doubts through effective communication, which greatly affects the learning effect and enthusiasm of learners on the online education platform. These problems usually cause learners to lose their way. , The frustration of learning halfway. In order to enable learners to better understand the learning content of online education platforms, improve learning efficiency, communicate more fully with other learners, establish good partnerships, and improve learners' platform experience and enthusiasm, it is necessary to provide online education for online education. Platform learners are innovative in their approach to providing learning partners.

SUMMARY OF THE INVENTION

The present invention aims to solve one of the technical problems in the related art at least to a certain extent.

Therefore, the purpose of the present invention is to propose a construction method for online education-oriented learners' collaborative learning social relationship, which can recommend suitable learning partners for learners, thereby enhancing the learning effect of learners on the platform.

In order to achieve the above object, an embodiment of the present invention proposes a method for constructing online education-oriented learners' collaborative learning of social relationships, which includes the following steps: Step S1, acquiring basic information data of learners in an online education platform, and extracting all the data by using sequence analysis. The attribute characteristics of the basic information data, wherein the attribute characteristics include the classified learner behavior characteristics, the basic distance between various disciplines and the positive interaction data; Step S2, conduct behavior analysis on the basic behavior characteristics in the attribute characteristics , to carry out partnership labeling; step S3, use the k-Means clustering algorithm to process the partnership labeling, and obtain a learner learning community clustering discovery; step S4, construct a population based on the learner learning community clustering discovery neural network, and use the attribute feature to train the learner's collaborative learning social relationship neural network to obtain the learner's collaborative learning social relationship neural network, and then obtain the current partner recommendation list; step S5, based on the scale-free network theory and The social triangle theory basically sorts the current learner's partners in a convergent state in the learner's collaborative learning social relationship neural network, and supplements the current partner recommendation list according to the long-tail theory of the recommendation system.

In addition, the method for constructing online education-oriented learners to collaboratively learn social relationships according to the foregoing embodiments of the present invention may also have the following additional technical features:

Further, in an embodiment of the present invention, the step S1 specifically includes: step S101 , acquiring the initial information data of the learner in the online education platform, and performing null value filling and basic heat quantization processing on the basic information data , obtain basic information data, wherein the basic information data includes behavior sequence, community interaction behavior record, subject preference, study time, study location and system login method; step S102, using sequence analysis method to extract the basic information data , obtain the basic behavioral characteristics and the corresponding vectorized results; Step S103, process the basic behavioral characteristics, the corresponding vectorized results and the community interaction behavior records according to the psychological personality division and scoring mechanism to obtain the learner's personality characteristics Numerical value; Step S104, carry out the weighted average numerical calculation of the basic behavioral characteristics and the learner's character characteristic numerical value, obtain the first-level numerical value to calculate the user's average social activity and platform activity as the learner's behavioral characteristics, and according to the crowd The characteristics of the behavior attribute Gaussian distribution are Gaussian distribution and classification of the learner behavior characteristics to obtain the classified learner behavior characteristics; step S105, using word2vec to quantify the learner's subject preference in the basic information data , to obtain the basic distance between various disciplines; step S106, use a word segmentation tool to perform word segmentation analysis on the community interaction behavior records in the basic information data, and numerically quantify whether the current interaction is active according to the sentiment value of the Chinese word, and obtain the interaction Positive data.

Further, in an embodiment of the present invention, the step S2 specifically includes: step S201, dividing and filtering the basic behavior features in the attribute features to obtain the learner's real interactive behavior and the learner's non-interactive behavior; Step S202, analyzing the learner's real interactive behavior and the learner's non-interactive behavior to derive a partnership label.

Further, in an embodiment of the present invention, the step S3 specifically includes: step S301, defining the learner-partner interaction behavior in the partnership annotation; step S302, according to the defined learner-partner interaction behavior Calculate the Jaccard Distance between the learners; Step S303, use the k-Means clustering algorithm to process the Jaccard Distance between the learners to obtain a cluster discovery of the learner learning community.

Further, in an embodiment of the present invention, the step S4 specifically includes: step S401, obtaining a time series relationship through the learner learning community clustering discovery, and dividing the learner behavior sequence according to the time series relationship , the recommended partner list is obtained by calculating the sequence of learners with later time and the sequence of learners with earlier time; step S402, the interactive behaviors between learners and other users are obtained through the learner learning community clustering discovery, to Labeling the partnership between the two; Step S403, based on deep learning, build an overall neural network according to the recommended partner list and the labeled partnership between the two, and use the attribute feature to train the overall neural network, Get the learner to collaboratively learn the social relationship neural network, and then obtain the current partner recommendation list.

Further, in an embodiment of the present invention, the step S5 specifically includes: step S501, inspecting the current social network behavior in the learner's collaborative learning social relationship neural network, and confirming whether it conforms to the basic scale-free network characteristics and evolution If it matches, go to step S502; step S502, carry out the evolution of the current partner network according to the probability connection and enhancement method of the scale-free network until it reaches a convergent state; step S503, based on the social triangle stability principle, according to the current learner's convergence For the partners in the state, the partners' partners are basically sorted, and the current partner recommendation list is supplemented according to the long tail theory of the recommendation system.

Further, in an embodiment of the present invention, it also includes: step S6, tracking the basic information data of the learner to periodically update the learner's characteristic attribute and partner network structure, and conduct new learning according to the new learner's personalized characteristic attribute of learners recommend partnership calculations.

The method for constructing online education-oriented learners' collaborative learning social relationship according to the embodiment of the present invention has at least the following advantages:

(1) Most of the learners of the existing online education platform watch video learning by themselves, and hardly participate in the discussion area to conduct valuable discussion and learning. The embodiment of the present invention makes up for the large number of online education learners participating, anonymous identities and non-discriminatory identities. In the case of contact, it is difficult for learners to carry out interactive learning, which emphasizes the matching of learning partners and learners in the recommendation process and the purpose of improving learning effect and efficiency, so that the learning process of learners is no longer a person. Learning alone makes the learning process easier and more enjoyable;

(2) The embodiment of the present invention emphasizes establishing a personalized model for the learner by means of the basic inherent attributes of the learner, such as age, gender, educational background, etc., and the accumulated behavioral and character characteristic data, which can screen out the learner for the learner. The learning partner that best fits his personal habits and personality, provides personalized recommendation of learning partners, which is more helpful for learners to answer questions in the communication with other users, improve learning efficiency, and make the learning process no longer boring. And monotony, all problems can be solved in time through communication, do not eat "cooked rice";

(3) The embodiment of the present invention includes not only the extraction of learner feature vectors and algorithm design, but also the system design of the learner interaction platform. The recommended partners of the online education platform and the friends of the interaction platform are directly mapped, which is better. provide services to learners;

(4) In the embodiment of the present invention, by continuously collecting the behavior data of platform users, synchronously updating the personalized characteristic data corresponding to the learners, and dynamically updating the list of learning partners recommended by the learners, it can be updated over time and continuously refined For the characterization of learners, the recommendation of learning partners is becoming more and more targeted.

Additional aspects and advantages of the present invention will be set forth, in part, from the following description, and in part will be apparent from the following description, or may be learned by practice of the invention.

Description of drawings

The above and/or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of embodiments taken in conjunction with the accompanying drawings, wherein:

1 is a flowchart of a method for constructing online education-oriented learners to collaboratively learn social relationships according to an embodiment of the present invention;

2 is a schematic diagram of a learner characteristic attribute according to an embodiment of the present invention;

Fig. 3 is the partnership algorithm labeling for the online education platform in the present invention;

Fig. 4 is the specific execution flow chart of step S3 in one embodiment of the present invention;

5 is a schematic structural diagram of the neural network structure of the learner collaborative learning social relationship constructed in step S4 in an embodiment of the present invention;

6 is a schematic diagram of the execution of a learner recommendation partner list interaction network simulation convergence algorithm in an embodiment of the present invention;

7 is a schematic diagram of the execution of a convergent network partner rearrangement algorithm in an embodiment of the present invention;

FIG. 8 is an architecture diagram of a learner interaction system of an online education platform in an embodiment of the present invention.

Detailed ways

The following describes in detail the embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals refer to the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary, and are intended to explain the present invention and should not be construed as limiting the present invention.

First of all, in the learning process of the online education platform, the courses usually include basic video learning, courseware quizzes, after-school learning, discussion in discussion forums, final homework and final exams. By recording the learner's action category, duration and other content on the platform, the learner's behavior sequence with time tags on the current platform can be constructed. However, these data are often only in the state of collection, and have not been taken out to analyze the learning status, abilities, characteristics and other personalized attributes of learners, and have not been fully utilized. The embodiment of the present invention proposes to make full use of a large amount of collected learner behavior data and basic characteristics, and specifically proposes to analyze the learner's behavior sequence and basic attributes, so as to recommend suitable learning partners for the learner to enhance the The learning effect on the platform reduces the trouble caused by individual learning alone. The following is a detailed description of the method for constructing the online education-oriented learners' collaborative learning social relationship proposed by the embodiments of the present invention.

The following describes the construction method for online education-oriented learners' collaborative learning social relationship according to the embodiments of the present invention with reference to the accompanying drawings.

FIG. 1 is a flowchart of a method for constructing online education-oriented learners to collaboratively learn social relationships according to an embodiment of the present invention.

As shown in Figure 1, the method for constructing the online education-oriented learners' collaborative learning social relationship includes the following steps:

In step S1, the basic information data of the learners in the online education platform is obtained, and the attribute features of the basic information data are extracted by using sequence analysis method, wherein the attribute features include the behavioral features of the learners after classification, the basic distance and interaction between subjects Positive data.

That is to say, in step S1, feature attribute extraction is performed by analyzing the learner's behavior sequence.

It should be noted that, before proceeding to step S1, learners are required to fill in basic attributes on the online education platform based on their personal interests, and then the online education platform filters the course learners, and learners who simply understand or browse the course content will be rejected. The filtering does not carry out in-depth analysis and partner recommendation, that is, it is not used as the basic information data of the learner in step S1.

Further, step S1 specifically includes:

Step S101: Acquire the initial information data of the learners in the online education platform, perform null value filling and basic heat treatment on the basic information data, and obtain basic information data, wherein the basic information data includes behavior sequences, community interaction behavior records, disciplines. Preferences, study time, study location and system login method.

It should be noted that the corresponding learner's behavior sequence on the current platform is generated according to the time series through the browsing records, learning records, exercise records and discussion records of the corresponding learners in the database (ie, initial information data).

Step S102 , extracting the basic information data by using the sequence analysis method to obtain basic behavior characteristics and corresponding vectorization results.

That is to say, the basic information data is extracted by means of sequence analysis, and the basic behavioral characteristics and corresponding vectorized results of learners in terms of common learning time, learning location and common learning system are obtained.

In step S103, the basic behavioral characteristics, the corresponding vectorization results and the community interaction behavior records are processed according to the psychological personality classification and scoring mechanism to obtain the learner's personality characteristic value, and at the same time, the database is stored locally.

Step S104: Calculate the weighted average value of the basic behavioral characteristics and the learner's character characteristic value, and obtain the first-level value to calculate the user's average social activity and platform activity as the learner's behavioral characteristics, and according to the characteristics of the Gaussian distribution of behavioral attributes in the crowd. Gaussian distribution and classification are performed on the learner's behavioral characteristics, and the classified learner's behavioral characteristics are obtained.

Specifically, the basic behavioral characteristics of learners, including video viewing rate, video viewing speed, average number of watched videos, average homework completion rate, average number of posts, average number of replies, and average number of likes, are weighted by numerical weighting. The average value is calculated, and the first-level value is obtained to calculate the second-level value such as the user's average social activity and platform activity as the learner's behavioral characteristics, and the learner's behavioral characteristics are Gaussian distribution and classification according to the characteristics of the Gaussian distribution of behavioral attributes in the crowd. , to obtain the learned behavioral characteristics after classification.

Step S105, using word2vec to quantify the learner's subject preference in the basic information data to obtain the basic distance between subjects.

Step S106, using a word segmentation tool to perform word segmentation analysis on the community interaction behavior records in the basic information data, and numerically quantify whether the current interaction is active according to the sentiment value of the Chinese word, to obtain interaction positive data.

For example, as shown in FIG. 2 , because learners cannot directly communicate with their partners face-to-face as in a traditional teaching environment, the embodiment of the present invention proposes a learner’s partner recommendation activity. The extracted content and the corresponding calculation method can be as follows: Show:

The embodiment of the present invention divides the attributes of learners to be investigated into the following types in combination with the basic content of educational psychology:

Similarity factor: Students with high degree of similarity in beliefs, attitudes, values, or in terms of identity, cultural level, and social background are easy to approach and form positive interpersonal relationships with each other; relation. The learner's basic attributes and personality characteristics (operation data obtained by other algorithms) are inherent characteristics of the learner, which will not change too much in a short period of time. The one-hot encoding after marking is performed by the method of category division.

Compensatory factor: Compensatory factor is a relatively important factor, that is, interpersonal relationships are conditioned by mutual expected compensation. Students are eager to achieve certain goals and meet certain needs in communication and interpersonal relationships. This conscious or unconscious social motivation is a natural psychological tendency of individuals. Such compensation includes receiving affirmation, sympathy, encouragement, support and assistance from others. On the online education platform, the recognition and agreement of other learners on their own learning habits and learning intentions is an important element of partnership building. Therefore, in the embodiment of the present invention, the subject preference attribute corresponding to the learner and the corresponding course learning intention attribute are taken out as compensatory factor feature extraction. Among them, the subject preference is based on the course selection tendency. There are a total of 15 kinds of computer, foreign language, management, philosophy, economics, law, education and teaching, literature and culture, history, science, engineering, agriculture, forestry and horticulture, medicine and health, art and design, and others. optional. The corresponding learning intent is the basic result data obtained by other algorithms, which can be used directly. Contains both course-level intents and chapter-level intents. Course-level intentions include {audio courses, browse references, check vacancies, obtain credits, and obtain certificates}, and chapter-level intentions include {browse references, learn knowledge, and obtain credits}, and the intentions are gradually strengthened.

The learner's behavior characteristics include two parts: the learner's basic learning habits, and the learner's historical interaction fit. The basic learning habits are divided into different action categories and durations based on the learner’s basic behavior sequence constructed before, and weighted and post-processed to obtain specific values. After that, the normal distribution is normalized, and the difference is made by the upper quantile on the right of 0.3 and 0.1. accustomed to the division of people. The learner's historical interaction fit is processed for the interaction records between the current study learner and a certain platform user. In the previous discussion, the Chinese word sentiment dictionary is used to score the overall sentiment after stammering, to determine whether there has been a quarrel or a good conversation between the current user and the learner, and the corresponding positive interaction data or fit score is obtained.

In step S2, behavioral analysis is performed on the basic behavioral features in the attribute features to obtain a partnership label.

In other words, the interaction behavior of learners in the online education platform is analyzed for partnership annotation.

Further, step S2 specifically includes:

Step S201, dividing and filtering the basic behavior features in the attribute features to obtain the learner's real interactive behavior and the learner's non-interactive behavior;

Step S202, analyzing the learner's real interactive behavior and the learner's non-interactive behavior to derive a partnership label.

Specifically, as shown in FIG. 3 , in the traditional educational environment, the partnership between students is usually marked by means of questionnaires, direct inquiry or indirect observation. However, unlike the traditional school education environment, the communication content between learners on the online education platform is too sparse. According to statistics, the learners who actively participate in the discussion in the discussion area are less than 1% of the overall learners. Effective discussion of learning aspects Content is less than 2%. Therefore, the embodiment of the present invention invokes the attribute feature obtained in step S1 to perform vector distance calculation, then uses a basic exponential function to perform an exponential operation on each distance value, and then performs MIN-MAX normalization for each user's optional partner set deal with. Due to the lack of direct interactive content, there is a logical error in the judgment after direct sorting. Therefore, in the embodiment of the present invention, a random number judgment method is used to mark the corresponding partnership. Understandably, learners with closer vector distances are more likely to be judged as learning partners. For a small number of learner pairs who have had direct communication, the user's social fit obtained in step S1 is used to judge whether it is an active conversation, and there are no quarrels and abuse. If not, mark it as a learning partner.

In step S3, the partnership labeling is processed using the k-Means clustering algorithm, and the clustering discovery of the learner learning community is obtained.

That is to say, the basic partner network is constructed based on the k-Means clustering algorithm, that is, the learner learns the community clustering discovery.

Further, in an embodiment of the present invention, step S3 specifically includes:

Step S301, define the learner-partner interaction behavior in the partnership labeling;

Step S302, calculating the JaccardDistance distance between learners according to the defined interaction behavior of the learner's partner;

Step S303 , using the k-Means clustering algorithm to process the Jaccard Distance between learners, to obtain the clustering discovery of the learner's learning community.

Specifically, the interactive behaviors between learners include the following: learning a course together for more than a period of time, having discussions in the discussion board, jointly completing a project or a large assignment, in the background information from the same Schools and ages are similar to these four categories. If there is one or more of the above connections between the learner and a user, an interactive edge is constructed between the two, and the value is 1. Different from the traditional vector clustering method based on k-Means, this The embodiment of the invention adopts the Jaccard distance for clustering and community discovery. In pedagogy, the influence of various factors on the weight of vector distance cannot be completely determined. Using Jaccard distance to judge the distance between different learners can more effectively solve the influence of subjective human factors in community discovery. After completing the community classification, users within the same community will enter into the next step of neural network more refined calculation and recommendation as each other's optional recommendation learning partners.

In step S4, an overall neural network is constructed according to the clustering discovery of the learner learning community, and the attribute features are used to train the learner's collaborative learning social relationship neural network to obtain the learner's collaborative learning social relationship neural network, and then the current partner recommendation list is obtained. . That is, the overall neural network is constructed and trained based on the learning community clustering discovery of learners.

Further, in an embodiment of the present invention, step S4 specifically includes:

Step S401, obtaining a time-series relationship through learner learning community clustering discovery, dividing the learner's behavior sequence according to the time-series relationship, and calculating a recommended partner list by calculating the learner sequence later in time and the learner sequence earlier in time;

Step S402, discovering and obtaining the interactive behavior between the learner and other users through the learner learning community clustering, so as to mark the partnership between the two;

Step S403, based on deep learning LSTM, embedding and other methods, and construct an overall neural network according to the recommended partner list and the annotated partnership between the two, and use attribute features to train the overall neural network to obtain a learner collaborative learning social relationship neural network. , and then obtain the current partner recommendation list.

Specifically, as shown in Fig. 5, the input end of the neural network for learners' collaborative learning of social relations includes the basic data of the user, the learner's learning habits, the learner's subject preference, and the learner's experience in the course extracted according to the learner's previous learning experience. of learning intent. The output terminal is the 0 or 1 classification result, that is, whether the user currently under investigation can become a good learning partner of the learner in the course. The design of the neural network adopts the following basic structures from front to back:

1) One-hot encoding layer:

Most of the basic information and learning habits of learners are classified into discrete categories. Using one-hot encoding can make data transformation into a method that is clear in meaning and easy to calculate by machine learning neural networks.

2)Embedding/Autoencoder layer:

In the current neural network operation process, the input vector dimension is basically required to be an integer power of 2 or a combination. However, it is obvious that many of the neural network input features extracted by this topic are not integer powers of 2 after one-hotization. Therefore, use the autoencoder to compress it into the nearest power-of-2 dimension vector. The calculation is more convenient and fast, and at the same time, in the case of unsupervised learning of autoencoder, the influence of human factors caused by annotation can be reduced.

3) Word2vec layer:

The learner's preferred subject is the content that the learner focuses on on the platform. However, simply comparing the number of similarities or differences between the preferred subject vectors among learners obviously does not meet the similarity factor requirements. The degree of similarity between two learners who like physics and mathematics with different preferences must be greater than that between two learners who like physics and literature. The degree of such relevance is determined by the characteristics of the preferred discipline itself. Therefore, the embodiment of the present invention decides to use the word2vec method to calculate the preferred subject distance between learners as a similarity factor for investigation.

In step S5, based on the scale-free network theory and the social triangle theory, the current learner's partners in the convergent state in the learner's collaborative learning social relationship neural network are basically sorted, and the current partner is recommended according to the long-tail theory of the recommendation system. list to be supplemented.

Further, in an embodiment of the present invention, step S5 specifically includes:

Step S501, examine the current social network behavior in the learner's collaborative learning social relationship neural network, and confirm whether it conforms to the basic scale-free network characteristics and evolution mode, and if so, go to step S502;

Step S502, the evolution of the current partner network is carried out according to the probability connection and enhancement mode of the scale-free network until the convergence state is reached;

Step S503 , based on the social triangle stability principle, according to the current learner's partners in a convergent state, basically sort the partners' partners, and supplement the current partner recommendation list according to the long-tail theory of the recommendation system.

Embodiments of the present invention also include:

In step S6, the basic information data of the learner is tracked to periodically update the learner's characteristic attribute and the partner network structure, and a new learner's recommended partnership calculation is performed according to the new learner's personalized characteristic attribute.

Specifically, as shown in FIG. 6 , according to the scale-free-small-world basic social network theory, the embodiment of the present invention uses a more stable Jaccard Distance to determine the basic distance between pairs of users. Those who are closer are more likely to interact in the next process. Use random numbers to simulate the next interactive process. If there is interaction, the edge weight is increased by 1, and the interaction probability in the next round is further increased according to the social aggregation coefficient. Otherwise, the edge weights are reduced according to the social relationship elimination rate. When the edge weight is below 0.5, it is judged that the current user basically does not interact, and the user is deleted. Convergence is reached until the number of edges in the whole graph is almost unchanged. At this time, the previously recommended partnerships have converged over time. In this embodiment of the present invention, in order to prevent learners from being limited to a small circle of learning, combined with the social triangle theory, additional users are re-considered into the list of recommended partners. The general process is shown in the figure. 7 is shown. After inverting the partners of the convergent partners according to the cosine distance, the embodiment of the present invention selects two users to be added to the current learner's recommended partner list according to the classic rebeat coefficient of the recommendation method of 0.2.

Finally, the embodiment of the present invention finally adopts the above algorithm combined with the online education platform to realize the online communication and learning platform for learners based on the system learning social relationship recommendation. As shown in FIG. 8 , the system is divided into a data collection module, an algorithm execution module, and a data persistence module. Modules and basic frontend presentation modules.

To sum up, the construction method for online education-oriented learners to collaboratively learn social relations proposed by the embodiments of the present invention introduces the logical structure and technical process of feature engineering and social network analysis, and adopts multi-dimensional quantification, graph structure convergence and neural network refinement. The method of evaluation is based on the behavior sequence on the platform and the speech records in the discussion area, etc. to generate the basic characteristic attributes of learners, and use attribute calculation and matching to construct the social relationship of learners' collaborative learning, so that the learners in the platform can learn The process is not carried out alone, and more efficient learning is completed through the mutual communication between different learners, thus solving the problem of the low level of interaction between learners in the current online education platform, and the discovery of good learning partners for learners. By analyzing the specific behavior sequences and specific characteristics of learners on the platform, it recommends suitable learning partners for learners to enhance the learning effect of learners on the platform.

In addition, the terms "first" and "second" are only used for descriptive purposes, and should not be construed as indicating or implying relative importance or implying the number of indicated technical features. Thus, a feature delimited with "first", "second" may expressly or implicitly include at least one of that feature. In the description of the present invention, "plurality" means at least two, such as two, three, etc., unless otherwise expressly and specifically defined.

In the description of this specification, description with reference to the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples", etc., mean specific features described in connection with the embodiment or example , structure, material or feature is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms are not necessarily directed to the same embodiment or example. Furthermore, the particular features, structures, materials or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, those skilled in the art may combine and combine the different embodiments or examples described in this specification, as well as the features of the different embodiments or examples, without conflicting each other.

Although the embodiments of the present invention have been shown and described above, it should be understood that the above-mentioned embodiments are exemplary and should not be construed as limiting the present invention. Embodiments are subject to variations, modifications, substitutions and variations.

Claims (7)

1. a construction method for online education-oriented learners to collaboratively learn social relations, is characterized in that, comprises the following steps: Step S1: Obtain the basic information data of learners in the online education platform, and extract attribute features of the basic information data by using a sequence analysis method, wherein the attribute features include classified learner behavior characteristics, basic information between subjects. distance and interaction positive data; Step S2, conduct behavior analysis on the basic behavior characteristics in the attribute characteristics, so as to mark the partnership; Step S3, using the k-Means clustering algorithm to process the labeling of the partnership, to obtain the clustering discovery of the learner learning community; Step S4, constructing an overall neural network according to the learner learning community clustering discovery, and using the attribute features to train the learner collaborative learning social relationship neural network to obtain a learner collaborative learning social relationship neural network, and then obtaining: Current partner recommendation list; Step S5, based on the scale-free network theory and social triangle theory, basically sort the current learner's partners in a convergent state in the learner's collaborative learning social relationship neural network, and rank the current learner's partners according to the recommendation system long-tail theory. The list of partner referrals is supplemented. 2. The method for constructing online education-oriented learners to collaboratively learn social relations according to claim 1, wherein the step S1 specifically comprises: Step S101: Acquire initial information data of learners in the online education platform, perform null value filling and basic heat treatment on the basic information data, and obtain basic information data, wherein the basic information data includes behavior sequences, community interactions Behavior records, subject preferences, study time, study location and system login method; Step S102, using sequence analysis method to extract the basic information data to obtain basic behavior characteristics and corresponding vectorization results; Step S103, processing the basic behavioral characteristics, the corresponding vectorization results and the community interaction behavior records according to the psychological personality classification and scoring mechanism, to obtain the learner's personality characteristic value; Step S104, performing a weighted average numerical calculation on the basic behavioral characteristics and the learner's character characteristic values to obtain a first-level value to calculate the average social activity of the user and the activity of the platform as the learner's behavioral characteristics, and according to the behavioral attributes in the crowd. The characteristics of the Gaussian distribution perform Gaussian distribution and classification on the learner behavior characteristics to obtain the classified learner behavior characteristics; Step S105, using word2vec to quantify the learner's subject preference in the basic information data to obtain the basic distance between subjects; Step S106, using a word segmentation tool to perform word segmentation analysis on the community interaction behavior records in the basic information data, and numerically quantify whether the current interaction is positive according to the sentiment value of the Chinese word, to obtain interaction positive data. 3. The method for constructing online education-oriented learners to collaboratively learn social relations according to claim 1, wherein the step S2 specifically comprises: Step S201, dividing and filtering the basic behavior features in the attribute features to obtain the learner's real interactive behavior and the learner's non-interactive behavior; Step S202, analyzing the learner's real interactive behavior and the learner's non-interactive behavior to derive a partnership label. 4. The method for constructing online education-oriented learners to collaboratively learn social relations according to claim 1, wherein the step S3 specifically comprises: Step S301, define the learner-partner interaction behavior in the partnership labeling; Step S302, calculating the Jaccard Distance between learners according to the defined interaction behavior of the learner's partner; Step S303 , using the k-Means clustering algorithm to process the Jaccard Distance between the learners to obtain the clustering discovery of the learner learning community. 5. The method for constructing online education-oriented learners to collaboratively learn social relations according to claim 1, wherein the step S4 specifically comprises: Step S401, obtaining a time series relationship through the learner learning community clustering discovery, dividing the learner behavior sequence according to the time series relationship, and calculating the learner sequence with a later time and a learner sequence with an earlier time. list of recommended partners; Step S402, obtaining the interactive behavior between the learner and other users through the learner learning community clustering discovery, so as to mark the partnership between the two; Step S403, based on deep learning, construct an overall neural network according to the recommended partner list and the marked partnership, and use the attribute features to train the overall neural network, to obtain the learners' collaborative learning social relationship neural network. network, and then obtain the current partner recommendation list. 6. The method for constructing online education-oriented learners to collaboratively learn social relations according to claim 1, wherein the step S5 specifically comprises: Step S501, examine the current social network behavior in the learner's collaborative learning social relationship neural network, and confirm whether it conforms to the basic scale-free network characteristics and evolution mode, and if so, execute step S502; Step S502, the evolution of the current partner network is carried out according to the probability connection and enhancement mode of the scale-free network until the convergence state is reached; Step S503 , based on the social triangle stability principle, according to the current learner's partners in a convergent state, basically sort the partners' partners, and supplement the current partner recommendation list according to the long-tail theory of the recommendation system. 7. The construction method of online education-oriented learners' collaborative learning social relationship according to claim 1, further comprising: Step S6, track the learner's basic information data to periodically update the learner's characteristic attribute and partner network structure, and perform a new learner recommendation partnership calculation according to the new learner's personalized characteristic attribute.

Images