CN113868515B - Self-adaptive learning path recommendation method based on ant colony algorithm - Google Patents

Self-adaptive learning path recommendation method based on ant colony algorithm Download PDF

Info

Publication number
CN113868515B
CN113868515B CN202111043932.0A CN202111043932A CN113868515B CN 113868515 B CN113868515 B CN 113868515B CN 202111043932 A CN202111043932 A CN 202111043932A CN 113868515 B CN113868515 B CN 113868515B
Authority
CN
China
Prior art keywords
knowledge
learner
learning
point
target
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202111043932.0A
Other languages
Chinese (zh)
Other versions
CN113868515A (en
Inventor
王磊
张佳婷
江巧永
费蓉
马永娟
王彬
王焱龙
罗颖
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xian University of Technology
Original Assignee
Xian University of Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Xian University of Technology filed Critical Xian University of Technology
Priority to CN202111043932.0A priority Critical patent/CN113868515B/en
Publication of CN113868515A publication Critical patent/CN113868515A/en
Application granted granted Critical
Publication of CN113868515B publication Critical patent/CN113868515B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/95—Retrieval from the web
    • G06F16/953—Querying, e.g. by the use of web search engines
    • G06F16/9535—Search customisation based on user profiles and personalisation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/35—Clustering; Classification
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
    • G06F16/367—Ontology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/30—Semantic analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/10—Services
    • G06Q50/20—Education
    • G06Q50/205—Education administration or guidance
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Business, Economics & Management (AREA)
  • General Engineering & Computer Science (AREA)
  • Databases & Information Systems (AREA)
  • Data Mining & Analysis (AREA)
  • General Health & Medical Sciences (AREA)
  • Computational Linguistics (AREA)
  • Health & Medical Sciences (AREA)
  • Educational Technology (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Educational Administration (AREA)
  • Tourism & Hospitality (AREA)
  • Strategic Management (AREA)
  • Biomedical Technology (AREA)
  • Evolutionary Computation (AREA)
  • Economics (AREA)
  • Primary Health Care (AREA)
  • Human Resources & Organizations (AREA)
  • Biophysics (AREA)
  • Marketing (AREA)
  • General Business, Economics & Management (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Animal Behavior & Ethology (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Electrically Operated Instructional Devices (AREA)

Abstract

The invention discloses an ant colony algorithm-based personalized learning path recommendation method, which specifically comprises the following steps: step 1, establishing a similar learner model; step 2, determining a similar learner set according to the similar learner model, and identifying adjacent learners of the target learner; step 3, establishing a knowledge point feature model; step 4, constructing a knowledge graph of the corresponding field; step 5, finding a learning starting point for learning by a learner; step 6, generating a personalized knowledge subgraph which is unique to each learner according to the learning starting point and the learning target; and 7, recommending the optimal path selection of the adjacent learners to the target learners by combining the ant colony algorithm, and recommending an optimal learning path for the learners. According to the invention, the behavior data of the learner on the learning platform is utilized, and the learning path is recommended to the learner by combining the domain knowledge graph from the personalized features of the learner.

Description

Self-adaptive learning path recommendation method based on ant colony algorithm
Technical Field
The invention belongs to the technical field of intelligent education methods, and relates to an ant colony algorithm-based personalized learning path recommendation method.
Background
The adaptive learning is a leap of the development quality of remote education, and education modernization requires education to provide high-quality personalized learning for students, so that learners can learn actively, learn according to own needs and learn according to a mode suitable for themselves. Along with the development of information technology and the increase of the number of network resources, the personalized demands of learners are increasing, however, the information overload and network navigation in big data age become the obstruction of the personalized learning of the learners, and the personalized learning demands of the learners are difficult to be met by the traditional unified lessons and arrangement work of teachers. Therefore, a learning scheme suitable for a learner is planned for the learner, so that the learner can learn more specifically, and further the learning efficiency of the learner is improved, so that the learning scheme becomes an important study subject.
The recommendation system is an important means for information filtering and is a very potential method for solving the information overload problem. Adaptive learning path recommendation is considered as an effective means to solve the phenomenon of knowledge lost in online learning. In order to complete learning objectives, learners in the adaptive learning guidance system need to learn a series of learning objects, and for learners with different learning objectives and different cognitive abilities, learning orders and content organization manners of the learning objects are different. How to complete the learning task in the shortest time by combining the sequences of the learning objects becomes the main content of the learning path research.
The existing learning path recommending method mostly adopts a unified learning course planning scheme, adopts a unified teaching mode for the same course, and is difficult to provide the most suitable learning route for different learners by learning according to the sequence of textbooks, and the learning efficiency is low.
Disclosure of Invention
The invention aims to provide an ant colony algorithm-based personalized learning path recommendation method, which utilizes behavior data of a learner on a learning platform, starts from personalized features of the learner and combines a domain knowledge graph to recommend a learning path to the learner.
The technical scheme adopted by the invention is that the personalized learning path recommending method based on the ant colony algorithm is implemented according to the following method:
step 1, establishing a similar learner model, wherein the similar learner model comprises a learning target, a learning style, a knowledge level and a learning motivation of a learner;
step 2, determining a similar learner set according to the established similar learner model, and identifying adjacent learners of the target learner;
step 3, establishing a knowledge point feature model;
step 4, constructing a knowledge graph of the corresponding field;
step 5, searching and matching according to the learning target of the similar learner and the knowledge graph established in the step 4, and finding a learning starting point of the learner to learn;
step 6, generating a personalized knowledge subgraph belonging to each learner by combining the knowledge graph according to the learning starting point and the learning target;
and 7, recommending the optimal path selection of the adjacent learners to the target learners by combining the ant colony algorithm, and recommending an optimal learning path for the learners.
The present invention is also characterized in that,
the learning targets of the learners in the step 1 are clustered by taking the learners selecting the same learning target as similar learners through statistics of behavior data of the learners in a learning platform;
the learning styles are divided into an aggregation type, a divergent type, an isomorphism type and an adjustment type according to the Knob learning style dividing method, the same learner can show various learning type trends, and the degrees of trends of different learners on various learning styles are different, so the vector S= (S) 1 ,s 2 ,s 3 ,s 4 ) Describing the style trend of a learner, wherein s 1 ,s 2 ,s 3 ,s 4 Respectively represent the degree of the learner belonging to the polymerization type, the divergence type, the isotype and the regulation type, and s is more than or equal to 0 i ≤1，i＝1,2,3,4，s 1 +s 2 +s 3 +s 4 ＝1；
The knowledge level of the learner is obtained through the learning test and is classified into three grades, namely general, medium and excellent, specifically:
after entering the self-adaptive learning guide system, a learner firstly selects a knowledge point to be learned, enters a pre-learning test page of the corresponding knowledge point, and obtains the initial knowledge level of the learner, namely the mastery degree W of the knowledge point through testing; mapping the knowledge point mastery degree of the learner to a certain value in a [0,1] interval through the test question score, and prescribing: when the learning degree W of a learner on a knowledge point is more than or equal to 0 and less than or equal to 0.3, the knowledge level of the learner is general; when the knowledge point mastering degree of the learner is more than 0.3 and less than or equal to 0.6, the knowledge level of the learner is medium; when the knowledge point mastering degree of the learner is more than 0.6 and less than or equal to 1, the knowledge level of the learner is excellent;
the learning motivation of the learner is divided into three levels of learning, mastering and proficiency, and the learning motivation of the learner is selected by the learner according to the importance degree of the corresponding knowledge points.
In the step 2, determining the similar learner set according to the established similar learner model specifically comprises the following steps: firstly, a learner selects a learning object and a learning motivation after entering a learning platform, and a learner cluster selecting the same learning object and learning motivation forms a similar learner set.
In the step 2, the identifying of the adjacent learner is specifically:
determining adjacent learners in the determined similar learner set according to the knowledge level and the learning style, specifically:
step 2.1, determining a learner group with similar knowledge level in the similar learner groupLet l= { L 1 ,L 2 ,…,L m The set of all learners who complete the learning task is denoted by l j J=1, 2, …, m, representing the knowledge level of the jth learner in the set of similar learners, let L 0 Representing the target learner, l 0 Representing the knowledge level of the target learner, using r 1 Adjacent parameters representing knowledge level, 0.ltoreq.r 1 Learner group with similar knowledge level less than or equal to 1Body->Expressed as:
step 2.2, the learner population having similar knowledge level obtained in step 2.1Determining a neighbor learner group L of a target learner according to the learning style of the learner o,similar The method specifically comprises the following steps:
by usingA learning style vector representing the target learner, +.>Respectively represent the degree of the target learner belonging to the aggregation type, the divergent type, the isotype and the regulation type, and is +.> Representation set->The learning style vector of the learner in question,
let r 2 Representing the adjacent parameters between learning styles, 0.ltoreq.r 2 ≤1，|s o -s k I represents vector s o And s k The euclidean distance between them, then the target learner's neighbor learner population is determined as:
the knowledge point feature model in the step 3 comprises the degree of the knowledge point adopting various expression modes and the difficulty coefficient d of the knowledge point, wherein the degree of the knowledge point adopting various expression modes is represented by a vector C, and C= (C) 1 ,c 2 ,c 3 ,c 4 ) Wherein c 1 ,c 2 ,c 3 ,c 4 The specific gravity of the knowledge content expressed by courseware PPT, video, audio and document in the knowledge content of one knowledge point is respectively represented, and is more than or equal to 0 and less than or equal to c i ≤1(i＝1,2,3,4),c 1 +c 2 +c 3 +c 4 =1; d is more than or equal to 0 and less than or equal to 1, and the closer d is to 0, the smaller the difficulty of the knowledge point is, the closer d is to 1, and the greater the difficulty of the knowledge point is.
The step 4 is specifically as follows:
step 4.1, classifying the electronic textbooks in the corresponding fields according to a three-layer knowledge framework, wherein the first layer is a outline of knowledge in the corresponding fields, namely chapters, the second layer is each corresponding section below the chapters, and the meta-knowledge is the third layer, wherein fields contained in the chapters comprise chapter ids, octonames, descriptions of the chapters and ids of the sections contained below the chapters; the fields contained in the section include a section id, a section name, a description of the section and a chapter id to which the section belongs; the meta knowledge is not subdivided, and the included fields comprise meta knowledge id, meta knowledge name, precursor knowledge point id, subsequent knowledge point id, other relation knowledge point id and section id, so that the construction of the ontology structure of the corresponding knowledge field is completed;
step 4.2, manually extracting entities, relations and attributes in the body structure, wherein the entities are chapters, sections and meta knowledge, the relations are the characteristics of the entity objects by connecting the two entities through edges, and the attributes are the characteristics of the entity objects, so that the initial construction of a knowledge graph is realized;
step 4.3, crawling hundred-degree encyclopedia data by using an octopus data acquisition device: leading in the network address of the collected corresponding domain knowledge points, then customizing the crawling fields required by keywords, abstract, english names, brief introduction and catalogues, and processing the imported corresponding domain knowledge points in batches to obtain crawled data;
step 4.4, manually marking the data obtained in the step 4.3, namely part of corpus sentence data, extracting entities from the marked data, carrying out non-repeated pairwise team formation on a plurality of entities contained in one corpus, storing the extracted result into a Mysql database, and carrying out de-duplication on the data with identical two entities from the extracted result;
and 4.5, carrying out relation extraction, manually judging the relation among the entities of a part of the data obtained in the step 3.4, generating data in the format of < entity, relation and corpus sentence >, and fusing the extracted triples < entity, relation and corpus sentence > with a knowledge graph preliminarily established manually and manually to realize the construction of the domain knowledge graph.
The step 5 is specifically as follows:
step 5.1, carrying out relation extraction on the knowledge graph established in the step 4 to obtain semantic relations among knowledge points, wherein the semantic relations comprise seven relations of dependence, inclusion, belonging to, near-sense, antisense and parity;
step 5.2, positioning learning targets of similar learners into the constructed hierarchical knowledge graph, if the target knowledge points are first-layer knowledge points, inquiring all second-layer sub-knowledge points of the knowledge points through father-son relations, then respectively inquiring third-layer sub-knowledge points of the knowledge points through all second-layer knowledge points, and randomly distributing the third-layer knowledge points to each similar learner as learning starting points;
if the target knowledge point is the knowledge point of the second layer, inquiring all the child knowledge points of the third layer of the knowledge point through father-son relationship, and randomly distributing the third layer of knowledge point to each similar learner as a learning starting point; if the target knowledge points are the third-layer meta-knowledge, the third-layer knowledge points are randomly allocated to each similar learner as learning starting points.
The personalized knowledge subgraph in the step 6 is: and forming a personalized knowledge subgraph of the learner by connecting one or more learning paths between the learning starting point and the learning target through knowledge points.
The step 7 is specifically as follows:
step 7.1, initializing various parameters in an ant colony algorithm: setting an influencing parameter alpha 1 ，α 2 Beta and information volatility, initializing pheromone tau of each road section ij (t) initializing the target learner L o Knowledge level of 1 o And learning style S o ；
Step 7.2, determining the adjacent learner population L in step 2 o,similar Each learner in the (a) serves as an ant, and each adjacent learner is randomly allocated with a learning starting point o according to the learning target of the target learner and the learning starting point allocation method in the step 5 i Namely, each ant is assigned with a learning starting point;
step 7.3, if the ant has already learned the assigned learning starting point knowledge point o i Traversing all next-level knowledge points o corresponding to the learning starting point according to the arrangement sequence of knowledge points in the personalized knowledge subgraph j Calculating next knowledge point o of ant selection j The probability of learning as the next knowledge point is:
wherein,from knowledge point o for ants i To knowledge point o j Is τ ij (t) represents the path o at the time point t i→ o j The number of pheromones left up, +.>Representing the level of knowledge and the knowledge point o of a learner j Degree of matching between difficulties->Representing target learner L o Learning wind of (a)Lattice and knowledge point o j Degree of matching between expression patterns of (a), alpha 1 And alpha 2 Respectively indicate->And->Influence parameters on decisions, beta representing pheromone tau ij (t) an influencing parameter for the decision;
the calculation is carried out according to the following formula:
wherein l o Representing learner L o Knowledge level d of (2) j Representing knowledge points o j Difficulty coefficient of (2);
the calculation is carried out according to the following formula:
wherein use is made ofA learning style vector representing the target learner, +.>Respectively indicates the degree of the target learner belonging to the aggregation type, the divergent type, the isotype and the regulation type,/-degree>Representing knowledge points o j A feature vector of the knowledge representation, wherein>Respectively represent a knowledge point o j Adopting the proportion of knowledge content expressed by courseware PPT, video, audio and documents to all knowledge content of the knowledge point;
j (i) represents at the knowledge point o i A set of all alternative paths for the next step of ants;
step 7.4, selecting the knowledge point o with the highest probability j As the next knowledge point learning, the learner learns from the knowledge point o after learning i To knowledge point o j Updating the pheromone on the road section, wherein the updated pheromone is as follows:
wherein ρ is the information volatility volatilization factor, ρ is more than or equal to 0 and less than or equal to 1,for learner to knowledge point o i To knowledge point o j Scoring of road segments, i.e. learner at road segment o i→ o j A pheromone generated thereon;
step 7.5, then the knowledge point o j Repeatedly executing the steps 7.3-7.4 as a starting point to select the next knowledge point until reaching an end point, and generating a learning path J (i), namely, generating a learning path J (i) by each adjacent learner;
step 7.6, for all road segments o on the resulting learned path J (i) i→ o j Updating the intensity of the pheromone, wherein the formula is as follows:
step 7.7, calculating the overall probability of each neighbor learner ultimately selecting and generating a learning path J (i)
That is, the product of the probabilities of selecting each path segment on path J (i) will be learned;
step 7.8, ranking the overall probabilities of all the adjacent learners generating the learning path J (i), and ranking the probabilitiesThe largest learning path is recommended to the learner.
The beneficial effects of the invention are as follows:
(1) The method effectively utilizes various behavior data of the learner on the learning platform, the learner learns and takes an examination in the self-adaptive learning guiding platform and participates in a series of behaviors discussed in the forum, and based on the data, a learner model can be established, and important indexes required by path recommendation, namely the cognitive level of the learner, can be obtained.
(2) The invention considers the influence of the difference of the learner individuals on the learning style on the learning path selection, recommends the learning path for the learner, and can meet the unified learning style of similar learner groups, improve the accuracy of the learning path recommendation and recommend the learning path which is more suitable for the learner to the learner.
(3) The invention establishes the knowledge graph in the junior middle school mathematics field, provides a visual knowledge point relation graph for the learner, enables the learner to acquire the relation among the knowledge points more intuitively, generates a personalized knowledge subgraph for the learner by extracting the semantic relation among the knowledge points, and provides a learning path guiding scheme for the learner in a targeted manner.
Drawings
Fig. 1 is a flowchart of a personalized learning path recommendation method based on an ant colony algorithm;
fig. 2 is a logic diagram of a personalized learning path recommendation method based on an ant colony algorithm.
Detailed Description
The invention will be described in detail below with reference to the drawings and the detailed description.
The invention discloses an ant colony algorithm-based personalized learning path recommendation method, which is implemented as shown in the flow chart of figures 1-2, and specifically comprises the following steps:
step 1, establishing a similar learner model, wherein the similar learner model comprises a learning target, a learning style, a knowledge level and a learning motivation of a learner;
the learning targets of the learners are clustered by taking the learners selecting the same learning target as similar learners through statistics of behavior data of the learners in a learning platform;
the learning styles are divided into an aggregation type, a divergent type, an isomorphism type and an adjustment type according to the Knob learning style dividing method, the same learner can show various learning type trends, and the degrees of trends of different learners on various learning styles are different, so the vector S= (S) 1 ,s 2 ,s 3 ,s 4 ) Describing the style trend of a learner, wherein s 1 ,s 2 ,s 3 ,s 4 Respectively represent the degree of the learner belonging to the polymerization type, the divergence type, the isotype and the regulation type, and s is more than or equal to 0 i ≤1，i＝1,2,3,4，s 1 +s 2 +s 3 +s 4 ＝1；
The knowledge level of the learner is obtained through the learning test and is classified into three grades, namely general, medium and excellent, specifically:
after entering the self-adaptive learning guide system, a learner firstly selects a knowledge point to be learned, enters a pre-learning test page of the corresponding knowledge point, and obtains the initial knowledge level of the learner, namely the mastery degree W of the knowledge point through testing; mapping the knowledge point mastery degree of the learner to a certain value in a [0,1] interval through the test question score, and prescribing: when the learning degree W of a learner on a knowledge point is more than or equal to 0 and less than or equal to 0.3, the knowledge level of the learner is general; when the knowledge point mastering degree of the learner is more than 0.3 and less than or equal to 0.6, the knowledge level of the learner is medium; when the knowledge point mastering degree of the learner is more than 0.6 and less than or equal to 1, the knowledge level of the learner is excellent;
the learning motivation of the learner is divided into three levels of learning, mastering and proficiency, and the learning motivation of the learner is selected by the learner according to the importance degree of the corresponding knowledge points;
step 2, determining a similar learner set according to the established similar learner model, and identifying adjacent learners of the target learner;
wherein, determining the similar learner set according to the established similar learner model specifically comprises: firstly, a learner selects a learning target and a learning motivation after entering a learning platform, and a learner cluster selecting the same learning target and learning motivation forms a similar learner set;
wherein, the identifying of the adjacent learner is specifically:
determining adjacent learners in the determined similar learner set according to the knowledge level and the learning style, specifically:
step 2.1, determining a learner group with similar knowledge level in the similar learner groupLet l= { L 1 ,L 2 ,…,L m The set of all learners who complete the learning task is denoted by l j J=1, 2, …, m, representing the knowledge level of the jth learner in the set of similar learners, let L 0 Representing the target learner, l 0 Representing the knowledge level of the target learner, using r 1 Adjacent parameters representing knowledge level, 0.ltoreq.r 1 Learner population with similar knowledge level +.1 +.>Expressed as:
in the step 2.2 of the method,learner population with similar knowledge level obtained in step 2.1Determining a neighbor learner group L of a target learner according to the learning style of the learner o,similar The method specifically comprises the following steps:
by usingA learning style vector representing the target learner, +.>Respectively represent the degree of the target learner belonging to the aggregation type, the divergent type, the isotype and the regulation type, and is +.> Representation set->The learning style vector of the learner in question,
let r 2 Representing the adjacent parameters between learning styles, 0.ltoreq.r 2 ≤1，|s o -s k I represents vector s o And s k The euclidean distance between them, then the target learner's neighbor learner population is determined as:
step 3, establishing a knowledge point feature model;
the knowledge point characteristic model comprises the degree of various expression modes adopted by the knowledge point and the difficulty coefficient d of the knowledge point, wherein the knowledge point adopts various modesThe degree of expression is represented by vector C, c= (C) 1 ,c 2 ,c 3 ,c 4 ) Wherein c 1 ,c 2 ,c 3 ,c 4 The specific gravity of the knowledge content expressed by courseware PPT, video, audio and document in the knowledge content of one knowledge point is respectively represented, and is more than or equal to 0 and less than or equal to c i ≤1(i＝1,2,3,4),c 1 +c 2 +c 3 +c 4 =1; d is more than or equal to 0 and less than or equal to 1, and the closer d is to 0, the smaller the difficulty of a knowledge point is, the closer d is to 1, and the greater the difficulty of the knowledge point is;
step 4, constructing a knowledge graph of the corresponding field; the method comprises the following steps:
step 4.1, classifying the electronic textbooks in the corresponding fields according to a three-layer knowledge framework, wherein the first layer is a outline of knowledge in the corresponding fields, namely chapters, the second layer is each corresponding section below the chapters, and the meta-knowledge is the third layer, wherein fields contained in the chapters comprise chapter ids, octonames, descriptions of the chapters and ids of the sections contained below the chapters; the fields contained in the section include a section id, a section name, a description of the section and a chapter id to which the section belongs; the meta knowledge is not subdivided, and the included fields comprise meta knowledge id, meta knowledge name, precursor knowledge point id, subsequent knowledge point id, other relation knowledge point id and section id, so that the construction of the ontology structure of the corresponding knowledge field is completed;
step 4.2, manually extracting entities, relations and attributes in the body structure, wherein the entities are chapters, sections and meta knowledge, the relations refer to that two entities are connected through edges, the attributes refer to characteristics of entity objects, the relations are connection relations between the two entities, for example, the entities are rational numbers and positive numbers, the rational numbers comprise positive numbers, and the relations between the rational numbers and the positive numbers comprise relations; the attribute is the feature of the knowledge point of rational number, and the feature of the knowledge point of positive number. And (3) injection: in general, only entities and relations are used for constructing the map, and the attributes are things contained in the entities and are not reflected on the map. For example, a person is an entity, and the attribute of the person is something that is inherent in the person, for example: < positive number, synonymous, negative number > triplet indicates that the relationship of positive and negative numbers is a synonymous relationship, and the attribute contains 17 kinds: definition, addition, subtraction, multiplication, division, exchange, combination, allocation, meaning, property, ground truth, algorithm, property theorem, decision theorem, inference, ground property, decision method, formula, algorithm, for example: the attribute with the rational number is proposed time and the reason number is proposed in 0580 years before the male element, and the attribute value is presented in 0580 years before the male element; thus constructing a three-level netlike junior middle school mathematics knowledge graph of chapter-section-element knowledge manually from top to bottom; realizing the preliminary construction of the knowledge graph;
step 4.3, crawling hundred-degree encyclopedia data by using an octopus data acquisition device: leading in the network address of the collected corresponding domain knowledge points, then customizing the crawling fields required by keywords, abstract, english names, brief introduction and catalogues, and processing the imported corresponding domain knowledge points in batches to obtain crawled data;
step 4.4, manually marking the data obtained in the step 4.3, namely part of corpus sentence data, extracting entities from the marked data, carrying out non-repeated pairwise team formation on a plurality of entities contained in one corpus, storing the extracted result into a Mysql database, and carrying out de-duplication on the data with identical two entities from the extracted result;
step 4.5, carrying out relation extraction, manually judging the relation among the entities of a part of the data obtained in the step 3.4, generating data in the format of < entity, relation and corpus sentence >, and fusing the extracted triples < entity, relation and corpus sentence > with a knowledge graph preliminarily established manually and manually to realize the construction of the domain knowledge graph;
step 5, searching and matching according to the learning target of the similar learner and the knowledge graph established in the step 4, and finding a learning starting point of the learner to learn; the method comprises the following steps:
step 5.1, carrying out relation extraction on the knowledge graph established in the step 4 to obtain semantic relations among knowledge points, wherein the semantic relations comprise seven relations of dependence, inclusion, belonging to, near-sense, antisense and parity;
step 5.2, positioning learning targets of similar learners into the constructed hierarchical knowledge graph, if the target knowledge points are first-layer knowledge points, inquiring all second-layer sub-knowledge points of the knowledge points through father-son relations, then respectively inquiring third-layer sub-knowledge points of the knowledge points through all second-layer knowledge points, and randomly distributing the third-layer knowledge points to each similar learner as learning starting points;
if the target knowledge point is the knowledge point of the second layer, inquiring all the child knowledge points of the third layer of the knowledge point through father-son relationship, and randomly distributing the third layer of knowledge point to each similar learner as a learning starting point; if the target knowledge points are the third-layer element knowledge, randomly distributing the third-layer knowledge points as learning starting points to each similar learner;
step 6, generating a personalized knowledge subgraph belonging to each learner by combining the knowledge graph according to the learning starting point and the learning target; wherein, the personalized knowledge subgraph is: one or more learning paths connected by knowledge points between a learning starting point and a learning target form a personalized knowledge subgraph of a learner;
the learning starting point and the learning target are unique for each learner, but the learning starting point obtained by traversing the target in the knowledge graph is not unique, and the path from the learning starting point to the learning end point is not unique, because one knowledge point has at least one relational node related to the learning starting point, for example, the mixed operation of the rational numbers comprises four sub-nodes of addition of the rational numbers, subtraction of the rational numbers, multiplication of the rational numbers and division of the rational numbers, four branches are separated, and at least four paths exist as a result. Thus, a learning objective may correspond to one or more learning origins. One or more learning paths connected by knowledge points between the learning starting point and the learning target form a personalized knowledge subgraph of the learner;
step 7, recommending optimal path selection of adjacent learners to target learners by combining ant colony algorithm, and recommending an optimal learning path for the learners; the method comprises the following steps:
step 7.1, initializing various parameters in an ant colony algorithm: setting influencing parametersα 1 ，α 2 Beta and information volatility, initializing pheromone tau of each road section ij (t) initializing the target learner L o Knowledge level of 1 o And learning style S o ；
Step 7.2, determining the adjacent learner population L in step 2 o,similar Each learner in the (a) serves as an ant, and each adjacent learner is randomly allocated with a learning starting point o according to the learning target of the target learner and the learning starting point allocation method in the step 5 i Namely, each ant is assigned with a learning starting point;
step 7.3, if the ant has already learned the assigned learning starting point knowledge point o i Traversing all next-level knowledge points o corresponding to the learning starting point according to the arrangement sequence of knowledge points in the personalized knowledge subgraph j Calculating next knowledge point o of ant selection j The probability of learning as the next knowledge point is:
wherein,from knowledge point o for ants i To knowledge point o j Is τ ij (t) represents the path o at the time point t i→ o j The number of pheromones left up, +.>Representing the level of knowledge and the knowledge point o of a learner j Degree of matching between difficulties->Representing target learner L o Learning style and knowledge point o j Degree of matching between expression patterns of (a), alpha 1 And alpha 2 Respectively indicate->And->Influence parameters on decisions, beta representing pheromone tau ij (t) an influencing parameter for the decision;
the calculation is carried out according to the following formula:
wherein l o Representing learner L o Knowledge level d of (2) j Representing knowledge points o j Difficulty coefficient of (2);
the calculation is carried out according to the following formula:
wherein use is made ofA learning style vector representing the target learner, +.>Respectively indicates the degree of the target learner belonging to the aggregation type, the divergent type, the isotype and the regulation type,/-degree>Representing knowledge points o j A feature vector of the knowledge representation, wherein>Respectively represent oneKnowledge point o j Adopting the proportion of knowledge content expressed by courseware PPT, video, audio and documents to all knowledge content of the knowledge point;
j (i) represents at the knowledge point o i A set of all alternative paths for the next step of ants;
step 7.4, selecting the knowledge point o with the highest probability j As the next knowledge point learning, the learner learns from the knowledge point o after learning i To knowledge point o j Updating the pheromone on the road section, wherein the updated pheromone is as follows:
wherein ρ is the information volatility volatilization factor, ρ is more than or equal to 0 and less than or equal to 1,for learner to knowledge point o i To knowledge point o j Scoring of road segments, i.e. learner at road segment o i→ o j A pheromone generated thereon;
step 7.5, then the knowledge point o j Repeatedly executing the steps 7.3-7.4 as a starting point to select the next knowledge point until reaching an end point, and generating a learning path J (i), namely, generating a learning path J (i) by each adjacent learner;
step 7.6, for all road segments o on the resulting learned path J (i) i →o j Updating the intensity of the pheromone, wherein the formula is as follows:
step 7.7, calculating the overall probability of each neighbor learner ultimately selecting and generating a learning path J (i)
That is, the product of the probabilities of selecting each path segment on path J (i) will be learned;
step 7.8, ranking the overall probabilities of all the adjacent learners generating the learning path J (i), and ranking the probabilitiesThe largest learning path is recommended to the learner.

Claims (6)

1. The personalized learning path recommending method based on the ant colony algorithm is characterized by comprising the following steps of:
step 1, establishing a similar learner model, wherein the similar learner model comprises a learning target, a learning style, a knowledge level and a learning motivation of a learner; the learning targets of the learners are clustered by taking the learners selecting the same learning target as a class of similar learners through statistics of behavior data of the learners in a learning platform; the learning styles are divided into an aggregation type, a divergent type, an isomorphism type and an adjustment type according to the Knob learning style dividing method, the same learner can show various learning type trends, and the degrees of trends of different learners on various learning styles are different, so the vector S= (S) 1 ,s 2 ,s 3 ,s 4 ) Describing the style trend of a learner, wherein s 1 ,s 2 ,s 3 ,s 4 Respectively represent the degree of the learner belonging to the polymerization type, the divergence type, the isotype and the regulation type, and s is more than or equal to 0 i ≤1，i＝1,2,3,4，s 1 +s 2 +s 3 +s 4 =1; the knowledge level of the learner is obtained through the learning test and is classified into three grades, namely general, medium and excellent, specifically: after entering the self-adaptive learning guide system, a learner firstly selects a knowledge point to be learned, enters a pre-learning test page of the corresponding knowledge point, and obtains the initial knowledge level of the learner, namely the mastery degree W of the knowledge point through testing; the knowledge points of learners are masteredThe test question score maps to [0,1]]A certain value in the interval, and specifies: when the learning degree W of a learner on a knowledge point is more than or equal to 0 and less than or equal to 0.3, the knowledge level of the learner is general; when the knowledge point mastering degree of the learner is more than 0.3 and less than or equal to 0.6, the knowledge level of the learner is medium; when the knowledge point mastering degree of the learner is more than 0.6 and less than or equal to 1, the knowledge level of the learner is excellent; the learning motivation of the learner is divided into three levels of learning, mastering and proficiency, and the learning motivation of the learner is selected by the learner according to the importance degree of the corresponding knowledge points;
step 2, determining a similar learner set according to the established similar learner model, and identifying adjacent learners of the target learner; the recognition of the adjacent learner is specifically:
determining adjacent learners in the determined similar learner set according to the knowledge level and the learning style, specifically:
step 2.1, determining a learner group with similar knowledge level in the similar learner groupLet l= { L 1 ,L 2 ,…,L m The set of all learners who complete the learning task is denoted by l j J=1, 2, …, m, representing the knowledge level of the jth learner in the set of similar learners, let L 0 Representing the target learner, l 0 Representing the knowledge level of the target learner, using r 1 Adjacent parameters representing knowledge level, 0.ltoreq.r 1 Learner population with similar knowledge level +.1 +.>Expressed as:
step 2.2, the learner population having similar knowledge level obtained in step 2.1Determining a neighbor learner group L of a target learner according to the learning style of the learner o,similar The method specifically comprises the following steps:
by usingA learning style vector representing the target learner, +.>Respectively represent the degree of the target learner belonging to the aggregation type, the divergent type, the isotype and the regulation type, and is +.> Representation set->The learning style vector of the learner in question, let r 2 Representing the adjacent parameters between learning styles, 0.ltoreq.r 2 ≤1，|s o -s k I represents vector s o And s k The euclidean distance between them, then the target learner's neighbor learner population is determined as:
step 3, establishing a knowledge point feature model;
step 4, constructing a knowledge graph of the corresponding field; the method comprises the following steps:
step 4.1, classifying the electronic textbooks in the corresponding fields according to a three-layer knowledge framework, wherein the first layer is a outline of knowledge in the corresponding fields, namely chapters, the second layer is each corresponding section below the chapters, and the meta-knowledge is the third layer, wherein fields contained in the chapters comprise chapter ids, octonames, descriptions of the chapters and ids of the sections contained below the chapters; the fields contained in the section include a section id, a section name, a description of the section and a chapter id to which the section belongs; the meta knowledge is not subdivided, and the included fields comprise meta knowledge id, meta knowledge name, precursor knowledge point id, subsequent knowledge point id, other relation knowledge point id and section id, so that the construction of the ontology structure of the corresponding knowledge field is completed;
step 4.2, manually extracting entities, relations and attributes in the body structure, wherein the entities are chapters, sections and meta knowledge, the relations are the characteristics of the entity objects by connecting the two entities through edges, and the attributes are the characteristics of the entity objects, so that the initial construction of a knowledge graph is realized;
step 4.3, crawling hundred-degree encyclopedia data by using an octopus data acquisition device: leading in the network address of the collected corresponding domain knowledge points, then customizing the crawling fields required by keywords, abstract, english names, brief introduction and catalogues, and processing the imported corresponding domain knowledge points in batches to obtain crawled data;
step 4.4, manually marking the data obtained in the step 4.3, namely part of corpus sentence data, extracting entities from the marked data, carrying out non-repeated pairwise team formation on a plurality of entities contained in one corpus, storing the extracted result into a Mysql database, and carrying out de-duplication on the data with identical two entities from the extracted result;
step 4.5, carrying out relation extraction, manually judging the relation among the entities of a part of the data obtained in the step 3.4, generating data in the format of < entity, relation and corpus sentence >, and fusing the extracted triples < entity, relation and corpus sentence > with a knowledge graph preliminarily established manually and manually to realize the construction of the domain knowledge graph;
step 5, searching and matching according to the learning target of the similar learner and the knowledge graph established in the step 4, and finding a learning starting point of the learner to learn;
step 6, generating a personalized knowledge subgraph belonging to each learner by combining the knowledge graph according to the learning starting point and the learning target;
and 7, recommending the optimal path selection of the adjacent learners to the target learners by combining the ant colony algorithm, and recommending an optimal learning path for the learners.
2. The personalized learning path recommendation method based on ant colony algorithm according to claim 1, wherein the determining the similar learner set according to the established similar learner model in the step 2 is specifically: firstly, a learner selects a learning object and a learning motivation after entering a learning platform, and a learner cluster selecting the same learning object and learning motivation forms a similar learner set.
3. The personalized learning path recommendation method based on ant colony algorithm according to claim 2, wherein the knowledge point feature model in step 3 includes the degree of knowledge point adopting various expression modes and the difficulty coefficient d of knowledge point, wherein the degree of knowledge point adopting various expression modes is represented by vector C, c= (C) 1 ,c 2 ,c 3 ,c 4 ) Wherein c 1 ,c 2 ,c 3 ,c 4 The specific gravity of the knowledge content expressed by courseware PPT, video, audio and document in the knowledge content of one knowledge point is respectively represented, and is more than or equal to 0 and less than or equal to c i ≤1(i＝1,2,3,4),c 1 +c 2 +c 3 +c 4 =1; d is more than or equal to 0 and less than or equal to 1, and the closer d is to 0, the smaller the difficulty of the knowledge point is, the closer d is to 1, and the greater the difficulty of the knowledge point is.
4. The personalized learning path recommendation method based on the ant colony algorithm according to claim 3, wherein the step 5 is specifically:
step 5.1, carrying out relation extraction on the knowledge graph established in the step 4 to obtain semantic relations among knowledge points, wherein the semantic relations comprise seven relations of dependence, inclusion, belonging to, near-sense, antisense and parity;
step 5.2, positioning learning targets of similar learners into the constructed hierarchical knowledge graph, if the target knowledge points are first-layer knowledge points, inquiring all second-layer sub-knowledge points of the knowledge points through father-son relations, then respectively inquiring third-layer sub-knowledge points of the knowledge points through all second-layer knowledge points, and randomly distributing the third-layer knowledge points to each similar learner as learning starting points;
if the target knowledge point is the knowledge point of the second layer, inquiring all the child knowledge points of the third layer of the knowledge point through father-son relationship, and randomly distributing the third layer of knowledge point to each similar learner as a learning starting point; if the target knowledge points are the third-layer meta-knowledge, the third-layer knowledge points are randomly allocated to each similar learner as learning starting points.
5. The personalized learning path recommendation method based on the ant colony algorithm according to claim 4, wherein the personalized knowledge subgraph in the step 6 is: and forming a personalized knowledge subgraph of the learner by connecting one or more learning paths between the learning starting point and the learning target through knowledge points.
6. The personalized learning path recommendation method based on the ant colony algorithm according to claim 5, wherein the step 7 is specifically:
step 7.1, initializing various parameters in an ant colony algorithm: setting an influencing parameter alpha 1 ，α 2 Beta and information volatility, initializing pheromone tau of each road section ij (t) initializing the target learner L o Knowledge level of 1 o And learning style S o ；
Step 7.2, determining the proximity in step 2Learner group L o,similar Each learner in the (a) serves as an ant, and each adjacent learner is randomly allocated with a learning starting point o according to the learning target of the target learner and the learning starting point allocation method in the step 5 i Namely, each ant is assigned with a learning starting point;
step 7.3, if the ant has already learned the assigned learning starting point knowledge point o i Traversing all next-level knowledge points o corresponding to the learning starting point according to the arrangement sequence of knowledge points in the personalized knowledge subgraph j Calculating next knowledge point o of ant selection j The probability of learning as the next knowledge point is:
wherein,from knowledge point o for ants i To knowledge point o j Is τ ij (t) represents the path o at the time point t i→ o j The number of pheromones left up, +.>Representing the level of knowledge and the knowledge point o of a learner j Degree of matching between difficulties->Representing target learner L o Learning style and knowledge point o j Degree of matching between expression patterns of (a), alpha 1 And alpha 2 Respectively indicate->And->For decision makingIs represented by beta, beta represents pheromone tau ij (t) an influencing parameter for the decision;
the saidThe calculation is carried out according to the following formula:
wherein l o Representing learner L o Knowledge level d of (2) j Representing knowledge points o j Difficulty coefficient of (2);
the saidThe calculation is carried out according to the following formula:
wherein use is made ofA learning style vector representing the target learner, +.>Respectively indicates the degree of the target learner belonging to the aggregation type, the divergent type, the isotype and the regulation type,/-degree>Representing knowledge points o j A feature vector of the knowledge representation, wherein>Respectively represent a knowledge point o j Knowledge content expressed by courseware PPT, video, audio and document occupies all knowledge pointsSpecific gravity of the knowledge content;
j (i) represents at the knowledge point o i A set of all alternative paths for the next step of ants;
step 7.4, selecting the knowledge point o with the highest probability j As the next knowledge point learning, the learner learns from the knowledge point o after learning i To knowledge point o j Updating the pheromone on the road section, wherein the updated pheromone is as follows:
wherein ρ is the information volatility volatilization factor, ρ is more than or equal to 0 and less than or equal to 1,for learner to knowledge point o i To knowledge point o j Scoring of road segments, i.e. learner at road segment o i→ o j A pheromone generated thereon;
step 7.5, then the knowledge point o j Repeatedly executing the steps 7.3-7.4 as a starting point to select the next knowledge point until reaching an end point, and generating a learning path J (i), namely, generating a learning path J (i) by each adjacent learner;
step 7.6, for all road segments o on the resulting learned path J (i) i→ o j Updating the intensity of the pheromone, wherein the formula is as follows:
step 7.7, calculating the overall probability of each neighbor learner ultimately selecting and generating a learning path J (i)
That is, the product of the probabilities of selecting each path segment on path J (i) will be learned;
step 7.8, ranking the overall probabilities of all the adjacent learners generating the learning path J (i), and ranking the probabilitiesThe largest learning path is recommended to the learner.
CN202111043932.0A 2021-09-07 2021-09-07 Self-adaptive learning path recommendation method based on ant colony algorithm Active CN113868515B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111043932.0A CN113868515B (en) 2021-09-07 2021-09-07 Self-adaptive learning path recommendation method based on ant colony algorithm

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111043932.0A CN113868515B (en) 2021-09-07 2021-09-07 Self-adaptive learning path recommendation method based on ant colony algorithm

Publications (2)

Publication Number Publication Date
CN113868515A CN113868515A (en) 2021-12-31
CN113868515B true CN113868515B (en) 2024-02-09

Family

ID=78994637

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111043932.0A Active CN113868515B (en) 2021-09-07 2021-09-07 Self-adaptive learning path recommendation method based on ant colony algorithm

Country Status (1)

Country Link
CN (1) CN113868515B (en)

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113127731A (en) * 2021-03-16 2021-07-16 西安理工大学 Knowledge graph-based personalized test question recommendation method

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20160005323A1 (en) * 2014-07-03 2016-01-07 Mentorum Solutions Inc. Adaptive e-learning system and method

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113127731A (en) * 2021-03-16 2021-07-16 西安理工大学 Knowledge graph-based personalized test question recommendation method

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
程岩 ; .在线学习中基于群体智能的学习路径推荐方法.系统管理学报.2011,(02),全文. *

Also Published As

Publication number Publication date
CN113868515A (en) 2021-12-31

Similar Documents

Publication Publication Date Title
CN111460249B (en) Personalized learning resource recommendation method based on learner preference modeling
Pardos et al. Connectionist recommendation in the wild: on the utility and scrutability of neural networks for personalized course guidance
CN111753098A (en) Teaching method and system based on cross-media dynamic knowledge graph
CN104318340B (en) Information visualization methods and intelligent visible analysis system based on text resume information
Christodoulopoulos et al. A group formation tool in an e-learning context
CN112784608B (en) Test question recommending method and device, electronic equipment and storage medium
CN111428051B (en) Multi-intelligence fused self-adaptive learning knowledge graph construction method and system
CN113344053B (en) Knowledge tracking method based on examination question different composition representation and learner embedding
CN113127731A (en) Knowledge graph-based personalized test question recommendation method
CN114065040A (en) Individual learning path and learning resource recommendation method based on discipline knowledge graph
Tam et al. Toward a complete e-learning system framework for semantic analysis, concept clustering and learning path optimization
CN113239209A (en) Knowledge graph personalized learning path recommendation method based on RankNet-transformer
CN114201684A (en) Knowledge graph-based adaptive learning resource recommendation method and system
CN113626499A (en) Student portrait mining implementation method based on big data warehouse technology
Huo et al. Towards personalized learning through class contextual factors-based exercise recommendation
CN112966092A (en) Knowledge graph personalized semantic recommendation method based on basic education
CN112380429A (en) Exercise recommendation method and device
CN113868515B (en) Self-adaptive learning path recommendation method based on ant colony algorithm
Troussas et al. Comparative analysis of algorithms for student characteristics classification using a methodological framework
CN115438152B (en) Simple answer scoring method and system based on multi-neural network and knowledge graph
Cam et al. Discovery of Course Success Using Unsupervised Machine Learning Algorithms.
Hamim et al. Student profile modeling: an overview model
Kallel et al. Building Collaborative e-learning teams in a smart education environment
Yan et al. Personalization exercise recommendation framework based on knowledge concept graph
CN114547481A (en) MOOC course hybrid recommendation method based on fuzzy logic and teaching plan

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant