Intelligent push system for lifelong learning resources based on credit bank and big data analysis

Abstract

The present invention provides a lifelong learning resource intelligent push system based on credit bank and big data analysis, including a knowledge construction module, a resource fragmentation module, a feature association module, and an intelligent push module. The present invention constructs a learning tree model based on the keywords of learning knowledge points, and then divides and extracts keywords from learning materials of different formats based on the tree model, so that books, teaching videos, audios and other information that require a unified longer learning time period can be fragmented, indexed by keywords, and reliability matching is performed based on big data, so that the student's time can be more fully used. The present invention matches accurate learning resources for students, effectively improves the students' learning interest and learning efficiency, and solves the problem that the existing learning matching method has a single function, resulting in low learning interest and learning efficiency for students.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent data push, and in particular to an intelligent push system for lifelong learning resources based on a credit bank and big data analysis. Background Art

[0002] Due to the continuous development of the Internet, the traditional education method adopts face-to-face teaching. Now more and more training education is conducted through online education. Online education can adopt one-on-one teacher video teaching, and students can freely choose courses according to their own preferences, which improves the convenience of learning and the diversity of learning methods.

[0003] However, in the existing online learning process, due to the proliferation of learning resources, it is difficult for self-taught students to find learning methods and courses that match their preferences. When students are learning in learning methods they do not like, with the popularization of fragmented learning and online learning, students can make better use of fragmented time to learn, but the current learning resources are generally long and time-consuming, so it is impossible to truly provide a reliable technical support environment for fragmented learning. The problem is low learning efficiency and insufficient learning effectiveness. In the existing online course matching methods, usually a model selected by students is adopted, and then the model is searched according to the keywords to obtain the corresponding learning courses. This matching mode has a single function and also requires course selection from the student side, so it is difficult to accurately match the learning needs of students. Summary of the invention

[0004] In view of the shortcomings of the prior art, the purpose of the present invention is to provide an intelligent push system for lifelong learning resources based on credit banks and big data analysis. By acquiring the learning habits of students and processing the data, accurate learning resources are matched for students, effectively improving students' learning interest and learning efficiency, so as to solve the problem that the existing learning matching methods have a single function, resulting in low learning interest and learning efficiency of students.

[0005] In order to achieve the above-mentioned purpose, the present invention is implemented through the following technical solutions: a lifelong learning resource intelligent push system based on credit bank and big data analysis, including a knowledge construction module, a resource fragmentation module, a feature association module, and an intelligent push module.

[0006] The knowledge construction module includes a knowledge construction strategy, which is used to construct a subject knowledge model. The knowledge construction strategy has the following steps:

[0007] Step S11, classifying knowledge information according to disciplines to generate knowledge keywords;

[0008] Step S12, generating a subject knowledge model tree according to the knowledge keywords corresponding to the relationship tags of the knowledge keywords;

[0009] Step S13, storing the knowledge keywords in the knowledge model database according to the positions of the knowledge keywords in the corresponding subject model tree;

[0010] The resource fragmentation module includes a resource processing strategy, and the resource processing strategy includes the following steps:

[0011] Step S21, identifying the type of learning resource. If the learning resource is a video resource, proceed to step S22-1; if the learning resource is an audio resource, proceed to step S22-2; if the learning resource is a question type resource, proceed to step S22-3;

[0012] Step S22-1, respectively obtain the picture information stream and audio information stream of the video resource, determine the corresponding picture features in real time from the picture information stream by a preset picture recognition algorithm, determine the corresponding audio features in real time from the audio information stream by a preset audio recognition algorithm, determine the video division basis according to the picture features and audio features, and divide the video resource according to the determined video division basis to obtain multiple new learning resources; Step S22-2, obtain the audio information stream of the audio resource, determine the corresponding audio features in real time from the audio information stream by a preset audio recognition algorithm, determine the audio division basis according to the audio features, and divide the audio resource according to the determined audio division basis to obtain multiple new learning resources; Step S22-3, divide the question type resources according to the preset question type division basis to obtain multiple new learning resources;

[0013] The feature association module includes a feature association strategy, and the feature association strategy includes the following steps:

[0014] Step S31, extracting keywords from the learning resources to generate keyword features;

[0015] Step S32, locating the corresponding keyword location from the knowledge model database according to the obtained keyword features to generate knowledge matching information, wherein the knowledge matching information reflects the position relationship between the learning resource and the corresponding keyword in the subject knowledge model tree;

[0016] Step S33, storing the knowledge matching information in the knowledge learning library;

[0017] The smart push module includes a smart push strategy, and the smart push strategy includes the following steps:

[0018] Step S41, recording a number of learning parameters of each student during the learning process, wherein the learning parameters include learning time, learning course type, and learning method;

[0019] Step S42, quantifying the recorded learning parameters, obtaining a learning bias value corresponding to the student according to the learning parameters, and establishing a learning bias model of the student according to the learning bias value;

[0020] Step S43, determining the location of the learning target from the knowledge model tree according to the learning list, retrieving the corresponding learning resource from the knowledge learning library, and calculating the learning reference value according to the learning content included in the learning resource;

[0021] Step S44, comparing the student's learning bias value with the learning reference value to match the best learning resources for the student.

[0022] Furthermore, the picture recognition algorithm includes a configured picture algorithm library, the picture algorithm library stores a number of picture extraction sub-algorithms for extracting picture features, each of the picture extraction sub-algorithms corresponds to a picture index condition, and the picture recognition algorithm determines the picture extraction sub-algorithm corresponding to the learning resource according to the picture index condition; the picture index condition includes font relationship condition, picture switching relationship condition, content sequence number relationship condition, and directory index relationship condition.

[0023] Furthermore, the audio recognition algorithm is configured with a conversion keyword library, which stores a number of conversion words, and the audio recognition algorithm recognizes the corresponding conversion words in the audio information flow as the audio features.

[0024] Furthermore, the audio recognition algorithm also includes calculating the audio conversion interval based on the audio change frequency in the audio information stream. When the audio in the audio information stream does not change for more than the audio conversion interval, the corresponding time period is used as the audio feature.

[0025] Further, the step S41 further includes step A1 and step A2, wherein step A1 includes: classifying the learning course type of the student into a plurality of learning courses, and labeling the plurality of learning courses as Kc1 to Kcn, respectively, wherein Kc1 is the first type of learning course, Kcn is the nth type of learning course, Kc is the code of the learning course, and 1 to n are the labeling sequences corresponding to the plurality of learning courses;

[0026] The step A2 includes: dividing the student's learning methods into video learning, audio learning, problem-solving learning and material-reading learning, and recording the learning time spent by the student in different learning courses respectively, and recording them respectively as the video learning time, audio learning time, problem-solving learning time and material-reading learning time under the learning course.

[0027] Further, the step S42 also includes a step B1, which includes: calculating the video learning ratio value of the student through a first bias algorithm, calculating the audio learning ratio value of the student through a second bias algorithm, calculating the question-solving learning ratio value of the student through a third bias algorithm, and calculating the material-reading learning ratio value of the student through a fourth bias algorithm;

[0028] Then, the video learning ratio, audio learning ratio, exercise learning ratio, and material learning ratio are brought into the learning bias algorithm to obtain the learning bias value;

[0029] The first bias algorithm is configured as follows: ; The second bias algorithm is configured as: ; The third bias algorithm is configured as: ; The fourth bias algorithm is configured as: ; Among them, Psp is the video learning ratio value, Kc1sp is the video learning time spent on the first learning course, Kcnsp is the video learning time spent on the nth learning course, Zsc is the total time spent by the student, Pyp is the audio learning ratio value, Kc1yp is the audio learning time spent on the first learning course, Kcnyp is the audio learning time spent on the nth learning course, Pzt is the problem-solving learning ratio value, Kc1zt is the problem-solving learning time spent on the first learning course, Kcnzt is the problem-solving learning time spent on the nth learning course, Pcl is the material-reading learning ratio value, Kc1cl is the material-reading learning time spent on the first learning course, Kcncl is the material-reading learning time spent on the nth learning course, and α is the ratio conversion coefficient.

[0030] Furthermore, the learning bias algorithm is configured as follows: ; Among them, Pxxp is the learning bias value, k1 is the video bias ratio, k2 is the audio bias ratio, k3 is the question-solving bias ratio, k4 is the material-reading bias ratio, b1 is the video reduction ratio, b2 is the audio reduction ratio, b3 is the question-solving reduction ratio, b4 is the material-reading reduction ratio, and k1, k2, k3, k4, b1, b2, b3 and b4 are all greater than zero.

[0031] Further, the step S43 also includes a step C1, which includes: classifying the learning resources according to the learning courses, and then dividing each learning resource in the classified learning courses according to the video proportion time, audio proportion time, number of questions and material proportion time, and calculating the learning reference value of each learning resource by the learning reference algorithm;

[0032] The learning reference algorithm is configured as follows: ; Among them, Pxck is the learning reference value, Sspc is the video proportion time, Sypc is the audio proportion time, Sszt is the number of questions, Sclc is the material proportion time, k5 is the video proportion ratio, k6 is the audio proportion ratio, k7 is the number of questions proportion, k8 is the material proportion ratio, b5 is the video proportion reduction value, b6 is the audio proportion reduction value, b7 is the number of questions reduction value, b8 is the material proportion reduction value, and k5, k6, k7, k8, b5, b6, b7 and b8 are all greater than zero.

[0033] Further, the step S44 also includes step D1, which includes: comparing the duration of individual learning courses of the plurality of learning courses, obtaining the top three individual learning courses with the longest duration, and sorting them from the longest duration to the shortest duration and marking them as the first biased course, the second biased course and the third biased course in sequence;

[0034] According to the recommended order of first-preference courses, second-preference courses and third-preference courses, three learning resources under this learning course type are recommended in turn.

[0035] Furthermore, the step S44 also includes a step D2, which includes: comparing the learning bias value with the learning reference value of the learning resources in the first bias course in sequence, arranging them in order from small to large according to the absolute value of the comparison difference, and selecting the top three learning resources as recommended learning resources for the first bias course;

[0036] Then the learning bias values ​​are compared with the learning reference values ​​of the learning resources in the second bias course in turn, and the learning resources are arranged in order from small to large according to the absolute value of the comparison difference, and the top three learning resources are selected as the recommended learning resources for the second bias course;

[0037] Finally, the learning bias values ​​are compared with the learning reference values ​​of the learning resources in the third bias course in turn, and they are arranged in order from small to large according to the absolute value of the comparison difference. The top three learning resources are selected as recommended learning resources for the third bias course.

[0038] Beneficial effects of the present invention: The present invention constructs a learning tree model with keywords of learning knowledge points, and then divides and extracts keywords from learning materials of different formats based on the tree model, so that books, teaching videos, audios and other information that require a unified longer learning period can be fragmented, indexed by keywords, and matched reliably based on big data, so that the students' time can be more fully utilized. By recording a number of learning parameters of each student in the learning process, and then quantifying the recorded learning parameters, the learning bias value corresponding to the student is obtained according to the learning parameters, and the learning bias model of the student is established according to the learning bias value, and then the learning resources in the knowledge database are classified, and the learning reference value is calculated according to the learning content included in the learning resources, and finally the learning bias value of the student is compared with the learning reference value, which can match the best learning resources for the student, thereby improving the learning interest and learning efficiency of the student, and improving the effectiveness and accuracy of learning resource matching. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0040] Figure 1 This is the schematic diagram of the system architecture;

[0041] Figure 2 It is a flow chart of the overall processing strategy of the present invention;

[0042] Figure 3 It is a principle block diagram of the comprehensive analysis system of the present invention.

[0043] Attached figure numbers: 100, knowledge construction module; 200, resource fragmentation module; 300, feature association module; 400, intelligent push module; V1, knowledge construction strategy; V2, resource processing strategy; V3, feature association strategy; V4, intelligent push strategy; 1. Knowledge model database; 2. Knowledge learning library; 11. Credit bank database module; 12. Intelligent analysis module; 13. Intelligent question-answering module; 14. Comprehensive evaluation module; 141. Intelligent conversion unit. DETAILED DESCRIPTION

[0044] In order to make the technical means, creative features, objectives and effects achieved by the present invention easy to understand, the present invention is further explained below in conjunction with specific implementation methods.

[0045] (1) A big data governance platform for multiple data contents. In the traditional model, only structured data is stored, and there is no macro-level definition and restriction on data association and data standards. When building the data platform, this project needs to consider the diversity of credit bank data storage (mainly structured and semi-structured), the requirements of global data governance, especially the global definition of data standards, and the lineage of data.

[0046] (2) Construction of an open data platform, which includes data access, data cleaning, and data standard access requirements. For data access and data cleaning, the main method is quick definition, and data access will be monitored and warned. In addition, technical standards and model standards for different types of data access will be provided to facilitate data access in the ecosystem.

[0047] (3) Presentation of data value. Different data visualization contents will be provided for different user groups at different data levels, so as to provide data basis for operation and management decision-making, so that data can be stored, managed and used. The core of this invention is to use cloud data in two aspects. The first is to push the best learning resources through the analysis of user habits. The second is to build a knowledge base based on cloud data and overcome the problem of inefficient learning resources through the keyword network. Cloud resources provide reliable support for the improvement and feasibility of keyword network models.

[0048] See also Figure 1 , including a knowledge construction module 100, a resource fragmentation module 200, a feature association module 300, and an intelligent push module 400,

[0049] The knowledge construction module 100 includes a knowledge construction strategy V1, which is used to construct a subject knowledge model. The knowledge construction strategy V1 has the following steps:

[0050] Step S11, classify the knowledge information according to the subject to generate knowledge keywords; this is different from general keywords. Due to the characteristics of the subject, the technical terms are standardized and characterized. The specificity of technical terms is strong, which provides a good basis for the classification of knowledge keywords.

[0051] Step S12, generating a subject knowledge model tree according to the knowledge keywords corresponding to the relationship tags of the knowledge keywords; if only the knowledge keywords are used as the basis for dividing the learning materials, it is inevitable that the knowledge keywords will be mixed, such as the use of mathematical knowledge keywords in the physics subject, so the knowledge keywords should be associated with each other. For example, a knowledge keyword such as "Newton's First Law" has keywords such as "mechanics" and "physics" upward, "maintaining stillness" and "external force" downward, and "ideal inclined plane experiment" and "Galileo" horizontally. The relationship tree can not only divide the features according to the direction of the relationship, but also be refined on this basis, such as the classification of the upper keywords, the inclusion relationship of the keywords included downward, the arguments and relationships of the horizontal keywords, etc., to establish a subject knowledge model tree of keywords from the knowledge system.

[0052] Step S13, store the knowledge keywords in the knowledge model database 1 according to their positions in the corresponding subject model tree; and according to the analysis relationship, the corresponding knowledge system can be stored in the knowledge model database 1, thus constructing a logical tree with keywords as the framework of the entire knowledge learning.

[0053] After constructing the logic tree, the resource fragmentation module 200 includes a resource processing strategy V2, and the resource processing strategy V2 includes the following steps:

[0054] Step S21, identify the type of learning resources. If the learning resource is a video resource, go to step S22-1; if the learning resource is an audio resource, go to step S22-2; if the learning resource is a question type resource, go to step S22-3. First, although the learning resources are stored in electronic format, there are great differences due to their different types. Therefore, the types of learning resources are first classified, and a preliminary division can be made according to the format.

[0055] Step S22-1, respectively obtain the picture information stream and audio information stream of the video resource, determine the corresponding picture features in real time from the picture information stream by a preset picture recognition algorithm, determine the corresponding audio features in real time from the audio information stream by a preset audio recognition algorithm, determine the video division basis according to the picture features and audio features, and divide the video resource according to the determined video division basis to obtain multiple new learning resources; Step S22-2, obtain the audio information stream of the audio resource, determine the corresponding audio features in real time from the audio information stream by a preset audio recognition algorithm, determine the audio division basis according to the audio features, and According to the determined audio division basis, the audio resources are divided to obtain multiple new learning resources; step S22-3, the question type resources are divided according to the preset question type division basis to obtain multiple new learning resources; the picture recognition algorithm includes a configured picture algorithm library, the picture algorithm library stores a number of picture extraction sub-algorithms for extracting picture features, each of the picture extraction sub-algorithms corresponds to a picture index condition, and the picture recognition algorithm determines the picture extraction sub-algorithm corresponding to the learning resource according to the picture index condition; the picture index condition includes a font relationship condition, a picture switching relationship condition, a content sequence number relationship condition, and a directory index relationship condition. The audio recognition algorithm is configured with a conversion keyword library, the conversion keyword library stores a number of conversion words, and the audio recognition algorithm identifies the corresponding conversion words in the audio information stream as the audio feature. The audio recognition algorithm also includes calculating the audio conversion interval time according to the audio change frequency in the audio information stream, and when the audio in the audio information stream does not change for more than the audio conversion interval time, the corresponding time period is used as the audio feature. Since learning resources need to be fragmented, there needs to be a certain basis for division. For example, if it is video information, the picture information stream can identify the text information through image recognition technology, and the audio information stream can extract features by identifying the information corresponding to the voice. If it is a picture feature, it is generally presented in the form of a training PPT combined with the teacher. Therefore, if the PPT title can be identified, whether the PPT screen is switched, the changes required for the picture title, whether the keywords of the directory index appear, etc., it can be used as the basis for whether it is a knowledge point, so that the video can be divided according to the picture index conditions. Dividing the video into multiple fragmented knowledge information is conducive to the use of fragmented learning time. Similarly, the conversion of the audio recognition algorithm may be based on some key sentences "Let's look at the next knowledge point" or "Enter the next chapter", etc., pre-store these keywords, and if there is a match, then a complete audio information can be divided, or the appearance of a piece of music and a long pause can be used as the basis for division, so that different types of learning resources are divided differently.

[0056] After the learning resources are fragmented, the feature association module 300 includes a feature association strategy V3 by processing the learning resources. The feature association strategy V3 includes the following steps:

[0057] Step S31, extracting keywords from the learning resources to generate keyword features; first, re-extracting keyword features from each learning word, which can be achieved through audio recognition, image recognition, text information recognition and other technologies.

[0058] Step S32, locate the corresponding keyword from the knowledge model database according to the obtained keyword features to generate knowledge matching information, and the knowledge matching information reflects the positional relationship between the learning resources and the corresponding keywords in the subject knowledge model tree; and for example, after the keywords are extracted, the position of each extracted keyword in the corresponding knowledge model tree is determined by comparison, and the knowledge model tree to which the keyword belongs is determined according to the frequency of keyword appearance, the position of appearance (title, directory, content) and other information. It should be noted that a keyword may have more than two positions in the knowledge model tree, because for example, "trigonometric function" may appear in the learning materials for calculating spatial position and plane position relationship, but if the keywords "cube" and "Z axis" appear, then the positional relationship in the model tree can be determined.

[0059] Step S33, storing the knowledge matching information in the knowledge learning library 2; then by constructing the knowledge learning library 2, the learning resources can be stored in the form of the location of the keywords, and can be matched according to the learning needs.

[0060] The smart push module 400 includes a smart push strategy, which includes the following steps:

[0061] The push method comprises the following sub-steps:

[0062] Step A1, said step A1 comprises: dividing the learning course type of the student into several learning courses, and labeling the several learning courses, respectively marked as Kc1 to Kcn, wherein Kc1 is the first type of learning course, Kcn is the nth type of learning course, Kc is the code of the learning course, 1 to n are the labeling sequences corresponding to the several learning courses, after the learning courses are labeled, subsequent classification correspondence management can be facilitated, and when the learning reference value of the learning resources in each learning course and the learning bias value of the student are matched, it can be more accurate, thereby improving the accuracy and efficiency of learning resource selection.

[0063] Step A2, said step A2 includes: dividing the student's learning methods into video learning, audio learning, problem-solving learning and reading material learning, recording the learning time spent by the student in different learning courses, and recording it as the video learning time, audio learning time, problem-solving learning time and reading material learning time under the learning course. Generally speaking, learning methods include video learning, audio learning, problem-solving learning and reading material learning. The learning efficiency of other types of learning methods is lower, so they are not used as a reference.

[0064] Step B1, the step B1 includes: calculating the video learning ratio value of the student through a first bias algorithm, calculating the audio learning ratio value of the student through a second bias algorithm, calculating the question-solving learning ratio value of the student through a third bias algorithm, and calculating the material-reading learning ratio value of the student through a fourth bias algorithm;

[0065] The first bias algorithm is configured as follows: ; The second bias algorithm is configured as: ; The third bias algorithm is configured as: ; The fourth bias algorithm is configured as: ; By calculating the proportion of time spent on video learning, audio learning, problem-solving learning, and reading material learning, we can find out which learning method the student likes to use most, where Psp is the video learning proportion value, Kc1sp is the video learning time spent on the first learning course, Kcnsp is the video learning time spent on the nth learning course, Zsc is the total time spent by the student, Pyp is the audio learning proportion value, Kc1yp is the audio learning time spent on the first learning course, Kcnyp is the audio learning time spent on the nth learning course, Pzt is the problem-solving learning proportion value, Kc1zt is the problem-solving learning time spent on the first learning course, Kcnzt is the problem-solving learning time spent on the nth learning course, Pcl is the reading material learning proportion value, Kc1cl is the reading material learning time spent on the first learning course, Kcncl is the reading material learning time spent on the nth learning course, and α is the proportion value conversion coefficient.

[0066] The video learning ratio, audio learning ratio, exercise learning ratio, and material reading learning ratio are brought into the learning bias algorithm to obtain the learning bias value;

[0067] The learning bias algorithm is configured as: ; Among them, Pxxp is the learning bias value, k1 is the video bias ratio, k2 is the audio bias ratio, k3 is the question-solving bias ratio, k4 is the material-reading bias ratio, b1 is the video reduction ratio, b2 is the audio reduction ratio, b3 is the question-solving reduction ratio, b4 is the material-reading reduction ratio, and k1, k2, k3, k4, b1, b2, b3 and b4 are all greater than zero. By uniformly calculating the ratio values ​​of video learning, audio learning, question-solving learning and material-reading learning, it is convenient for subsequent unified comparison processing.

[0068] Step C1, said step C1 includes: classifying the learning resources according to the learning courses, and then dividing each learning resource in the classified learning courses according to the proportion of video time, the proportion of audio time, the number of questions and the proportion of material time, and calculating the learning reference value of each learning resource through the learning reference algorithm.

[0069] The learning reference algorithm is configured as follows: ; Among them, Pxck is the learning reference value, Sspc is the video proportion time, Sypc is the audio proportion time, Sszt is the number of questions, Sclc is the material proportion time, k5 is the video proportion ratio, k6 is the audio proportion ratio, k7 is the number of questions proportion, k8 is the material proportion ratio, b5 is the video proportion reduction value, b6 is the audio proportion reduction value, b7 is the number of questions reduction value, b8 is the material proportion reduction value, and k5, k6, k7, k8, b5, b6, b7 and b8 are all greater than zero. By calculating the learning reference value of the learning resources, it is convenient to compare with the student's learning bias value, so as to recommend designated learning resources to the students.

[0070] Step D1, the step D1 comprises: comparing the duration of individual learning courses of a plurality of learning courses, obtaining the top three individual learning courses with the longest duration, and sorting them from the longest duration to the shortest duration and marking them as the first bias course, the second bias course and the third bias course in sequence;

[0071] According to the recommended order of first-preference courses, second-preference courses and third-preference courses, three learning resources under this learning course type are recommended in turn.

[0072] In the process of making recommendations, we cannot simply compare the learning bias value and the learning reference value. We should make recommendations based on the type of learning courses that the student likes, so as to further improve the effectiveness of learning resource recommendations.

[0073] Step D2, the step D2 comprises: comparing the learning bias value with the learning reference value of the learning resources in the first bias course in sequence, arranging them in order from small to large according to the absolute value of the comparison difference, and selecting the top three learning resources as recommended learning resources for the first bias course;

[0074] Then the learning bias values ​​are compared with the learning reference values ​​of the learning resources in the second bias course in turn, and the learning resources are arranged in order from small to large according to the absolute value of the comparison difference, and the top three learning resources are selected as the recommended learning resources for the second bias course;

[0075] Finally, the learning bias values ​​are compared with the learning reference values ​​of the learning resources in the third bias course, and the comparison differences are arranged in order from small to large. The top three learning resources are selected as recommended learning resources for the third bias course. Based on the learning course, by comparing the learning reference value and the learning bias value, students can be accurately matched with better learning resources, thereby improving the efficiency and effectiveness of students' course selection.

[0076] See also Figure 2 , the push method is mainly summarized as follows: Step S41, recording a number of learning parameters of each student during the learning process, the number of learning parameters including learning time, learning course type and learning method;

[0077] Step S42, quantifying the recorded learning parameters, obtaining a learning bias value corresponding to the student according to the learning parameters, and establishing a learning bias model of the student according to the learning bias value;

[0078] Step S43, determine the position of the learning goal from the knowledge model tree according to the learning list, and call the corresponding learning resources from the knowledge learning library 2, and calculate the learning reference value according to the learning content included in the learning resources; through such a setting, the corresponding data can be called from the student's learning list, and the corresponding learning tasks can be matched in fragmented time, which provides the possibility for intelligent push. After the electronic data is fragmented, it is indexed by the position of the knowledge tree, and then the corresponding learning resources are obtained with the same calling logic, and matched according to the specific learning situation of the learner.

[0079] Step S44, comparing the student's learning bias value with the learning reference value to match the best learning resources for the student.

[0080] See also Figure 3The application of the push method can be expanded by using the knowledge model database. The push method is applied to a credit bank comprehensive analysis system. The comprehensive analysis system 1 includes a credit bank database module 11, an intelligent analysis module 12, an intelligent question-answering module 13 and a comprehensive evaluation module 14.

[0081] Among them, the information stored in the credit bank database module 11 comes from a wide range of sources. The credit bank database module 11 includes ministry data, provincial agency data, public service data, school data, basic education data, ecological data, etc., and can use a larger sample as a basis to improve the accuracy of intelligent analysis results.

[0082] The intelligent analysis module 12 is used to intelligently analyze students' learning data and track students' learning status in real time; by intelligently monitoring and analyzing students' learning status, it can further provide a basis for teachers to adjust their teaching methods, ultimately achieving a good state of synchronous unity of learning and teaching, and serving as an important part of the performance evaluation system.

[0083] The intelligent question-answering module 13 is used to give intelligent replies to questions raised by students. In open education, most student users are on-the-job education types who study while working. They only have time to study after work, such as on weekends and evenings. However, this time period is generally the teacher's rest time after get off work or on weekends, which leads to the isolation of teachers and students in time and space and makes it difficult to synchronize with each other. Therefore, setting up an intelligent question-answering module can intelligently answer some questions of student users in the teaching platform, and can provide automatic question-answering 24 hours a day, thereby solving the problem that teachers and students are not online at the same time.

[0084] The comprehensive evaluation module 14 is used to evaluate the learning situation of the students. The comprehensive evaluation module configuration 14 has the intelligent push method for lifelong learning resources based on the credit bank and big data analysis in the first embodiment; the push method includes: recording a number of learning parameters of each student in the learning process, the learning parameters include learning time, learning course type and learning method; then quantifying the recorded learning parameters, obtaining the learning bias value corresponding to the student according to the learning parameters, and establishing the learning bias model of the student according to the learning bias value; then classifying the learning resources in the knowledge database, and calculating the learning reference value according to the learning content included in the learning resources; finally, comparing the learning bias value of the student with the learning reference value, and matching the best learning resources for the student. Through the push method, the user's learning bias can be analyzed, and the learning evaluation can be performed based on the student's learning bias, and finally the matching learning resources can be pushed to the student user, thereby improving the actual application effect of the analysis method of the credit bank comprehensive analysis system.

[0085] The comprehensive evaluation module 14 also includes an intelligent conversion unit 141, which is used to process and convert student data and credits in the credit bank, thereby improving the efficiency of data replacement and realizing intelligent data docking and conversion in the platform.

[0086] It should be noted that the above content is all collected, processed, calculated, and stored through cloud servers, in order to combine big data to improve data reliability and accuracy. Through the cloud computing architecture, a large amount of unformatted data can be uniformly processed, quantified, and calculated, simplifying complex models. The model tends to be reliable under continuous training, achieving the effect of accurate push.

[0087] Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention is described in detail with reference to the above-described embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-described embodiments within the technical scope disclosed by the present invention, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. An intelligent push system for lifelong learning resources based on credit bank and big data analysis, characterized by: Including knowledge construction module, resource fragmentation module, feature association module, and intelligent push module. The knowledge construction module includes a knowledge construction strategy, which is used to construct a subject knowledge model. The knowledge construction strategy has the following steps: Step S11, classifying knowledge information according to disciplines to generate knowledge keywords; Step S12, generating a subject knowledge model tree according to the knowledge keywords corresponding to the relationship tags of the knowledge keywords; Step S13, storing the knowledge keywords in the knowledge model database according to the positions of the knowledge keywords in the corresponding subject model tree; The resource fragmentation module includes a resource processing strategy, and the resource processing strategy includes the following steps: Step S21, identifying the type of learning resource. If the learning resource is a video resource, proceed to step S22-1; if the learning resource is an audio resource, proceed to step S22-2; if the learning resource is a question type resource, proceed to step S22-3; Step S22-1, respectively obtain the picture information stream and audio information stream of the video resource, determine the corresponding picture features in real time from the picture information stream by a preset picture recognition algorithm, determine the corresponding audio features in real time from the audio information stream by a preset audio recognition algorithm, determine the video division basis according to the picture features and audio features, and divide the video resource according to the determined video division basis to obtain multiple new learning resources; Step S22-2, obtain the audio information stream of the audio resource, determine the corresponding audio features in real time from the audio information stream by a preset audio recognition algorithm, determine the audio division basis according to the audio features, and divide the audio resource according to the determined audio division basis to obtain multiple new learning resources; Step S22-3, divide the question type resources according to the preset question type division basis to obtain multiple new learning resources; The feature association module includes a feature association strategy, and the feature association strategy includes the following steps: Step S31, extracting keywords from the learning resources to generate keyword features; Step S32, locating the corresponding keyword from the knowledge model database according to the obtained keyword features to generate knowledge matching information, wherein the knowledge matching information reflects the positional relationship between the learning resource and the corresponding keyword in the subject knowledge model tree; Step S33, storing the knowledge matching information in the knowledge learning library; The smart push module includes a smart push strategy, and the smart push strategy includes the following steps: Step S41, recording a number of learning parameters of each student during the learning process, wherein the learning parameters include learning time, learning course type, and learning method; Step S42, quantifying the recorded learning parameters, obtaining a learning bias value corresponding to the student according to the learning parameters, and establishing a learning bias model of the student according to the learning bias value; Step S43, determining the location of the learning target from the knowledge model tree according to the learning list, and retrieving the corresponding learning resource from the knowledge learning library, and calculating the learning reference value according to the learning content included in the learning resource; Step S44, comparing the student's learning bias value with the learning reference value to match the best learning resources for the student.

2. The intelligent push system for lifelong learning resources based on credit bank and big data analysis according to claim 1 is characterized in that: The picture recognition algorithm includes a configured picture algorithm library, which stores a number of picture extraction sub-algorithms for extracting picture features, each of which corresponds to a picture index condition, and the picture recognition algorithm determines the picture extraction sub-algorithm corresponding to the learning resource based on the picture index condition; the picture index condition includes font relationship condition, picture switching relationship condition, content sequence number relationship condition, and directory index relationship condition.

3. The intelligent push system for lifelong learning resources based on credit bank and big data analysis according to claim 1 is characterized in that: The audio recognition algorithm is configured with a conversion keyword library, which stores a number of conversion words. The audio recognition algorithm recognizes the corresponding conversion words in the audio information flow as the audio features.

4. The intelligent push system for lifelong learning resources based on credit bank and big data analysis according to claim 3 is characterized in that: The audio recognition algorithm also includes calculating the audio conversion interval based on the audio change frequency in the audio information stream. When the audio in the audio information stream does not change for more than the audio conversion interval, the corresponding time period is used as the audio feature.

5. The intelligent push system for lifelong learning resources based on credit bank and big data analysis according to claim 1 is characterized in that: The step S41 further includes step A1 and step A2, wherein step A1 includes: classifying the learning course type of the student into a plurality of learning courses, and labeling the plurality of learning courses, respectively labeling them as Kc1 to Kcn, wherein Kc1 is the first type of learning course, Kcn is the nth type of learning course, Kc is the code of the learning course, and 1 to n are the labeling sequences corresponding to the plurality of learning courses; The step A2 includes: dividing the student's learning methods into video learning, audio learning, problem-solving learning and material-reading learning, and recording the learning time spent by the student in different learning courses respectively, and recording them respectively as the video learning time, audio learning time, problem-solving learning time and material-reading learning time under the learning course.

6. The intelligent push system for lifelong learning resources based on credit bank and big data analysis according to claim 5 is characterized in that: The step S42 also includes a step B1, wherein the step B1 includes: calculating the video learning ratio value of the student through a first bias algorithm, calculating the audio learning ratio value of the student through a second bias algorithm, calculating the question-solving learning ratio value of the student through a third bias algorithm, and calculating the material-reading learning ratio value of the student through a fourth bias algorithm; Then, the video learning ratio, audio learning ratio, exercise learning ratio, and material reading learning ratio are brought into the learning bias algorithm to obtain the learning bias value; The first bias algorithm is configured as follows: ; The second bias algorithm is configured as: ; The third bias algorithm is configured as: ; The fourth bias algorithm is configured as: ; Among them, Psp is the video learning ratio value, Kc1sp is the video learning time spent on the first learning course, Kcnsp is the video learning time spent on the nth learning course, Zsc is the total time spent by the student, Pyp is the audio learning ratio value, Kc1yp is the audio learning time spent on the first learning course, Kcnyp is the audio learning time spent on the nth learning course, Pzt is the problem-solving learning ratio value, Kc1zt is the problem-solving learning time spent on the first learning course, Kcnzt is the problem-solving learning time spent on the nth learning course, Pcl is the material-reading learning ratio value, Kc1cl is the material-reading learning time spent on the first learning course, Kcncl is the material-reading learning time spent on the nth learning course, and α is the ratio conversion coefficient.

7. The intelligent push system for lifelong learning resources based on credit bank and big data analysis according to claim 6 is characterized in that: The learning bias algorithm is configured as: ; Among them, Pxxp is the learning bias value, k1 is the video bias ratio, k2 is the audio bias ratio, k3 is the question-solving bias ratio, k4 is the material-reading bias ratio, b1 is the video reduction ratio, b2 is the audio reduction ratio, b3 is the question-solving reduction ratio, b4 is the material-reading reduction ratio, and k1, k2, k3, k4, b1, b2, b3 and b4 are all greater than zero.

8. The intelligent push system for lifelong learning resources based on credit bank and big data analysis according to claim 7 is characterized in that: The step S43 also includes a step C1, which includes: classifying the learning resources according to the learning courses, and then dividing each learning resource in the classified learning courses according to the video proportion time, audio proportion time, number of questions and material proportion time, and calculating the learning reference value of each learning resource by a learning reference algorithm; The learning reference algorithm is configured as follows: ; Among them, Pxck is the learning reference value, Sspc is the video proportion time, Sypc is the audio proportion time, Sszt is the number of questions, Sclc is the material proportion time, k5 is the video proportion ratio, k6 is the audio proportion ratio, k7 is the number of questions proportion, k8 is the material proportion ratio, b5 is the video proportion reduction value, b6 is the audio proportion reduction value, b7 is the number of questions reduction value, b8 is the material proportion reduction value, and k5, k6, k7, k8, b5, b6, b7 and b8 are all greater than zero.

9. The intelligent push system for lifelong learning resources based on credit bank and big data analysis according to claim 8 is characterized in that: The step S44 also includes a step D1, which includes: comparing the duration of individual learning courses of the plurality of learning courses, obtaining the top three individual learning courses with the longest duration, and sorting them from the longest duration to the shortest duration and marking them as the first biased course, the second biased course, and the third biased course in sequence; According to the recommended order of first-preference courses, second-preference courses and third-preference courses, three learning resources under this learning course type are recommended in turn.

10. The intelligent push system for lifelong learning resources based on credit bank and big data analysis according to claim 9 is characterized in that: The step S44 also includes a step D2, which includes: comparing the learning bias value with the learning reference value of the learning resources in the first bias course in sequence, arranging them in order from small to large according to the absolute value of the comparison difference, and selecting the top three learning resources as recommended learning resources for the first bias course; Then the learning bias values ​​are compared with the learning reference values ​​of the learning resources in the second bias course in turn, and the learning resources are arranged in order from small to large according to the absolute value of the comparison difference, and the top three learning resources are selected as the recommended learning resources for the second bias course; Finally, the learning bias values ​​are compared with the learning reference values ​​of the learning resources in the third bias course in turn, and they are arranged in order from small to large according to the absolute value of the comparison difference. The top three learning resources are selected as recommended learning resources for the third bias course.

Images