Personalized learning path recommendation method based on online learner portrait clustering mining, storage medium and equipment

Abstract

The invention discloses a personalized learning path recommendation method based on online learner portrait clustering mining, a storage medium and equipment, and the method comprises the steps: collecting the basic information, dynamic learning behavior sequence data and static result data of online learners, constructing a personalized learning feature sequence of the online learner and a time sequence learning behavior sequence of the online learner to obtain an online learner portrait sequence; performing category division on all online learner portrait sequences according to course comprehensive scores, mining a high-contribution personalized learning feature sequence from the personalized learning feature sequences of the online learners through an information gain rate, and clustering the online learner portrait sequences; mining a maximum frequent learning activity sequence pattern of various online learner groups; and recommending the personalized learning characteristics of the online learner with the highest comprehensive course score in the online learner group to other online learners in the same maximum frequent learning activity sequence mode.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of personalized learning recommendation, in particular to a personalized learning path recommendation method based on online learner portrait clustering mining, a storage medium and equipment. BACKGROUND

[0002] Online learning meets the learning needs of learners "anytime, anywhere" with the advantage of transcending time and space limitations. However, the current mainstream online learning platform adopts a one-to-many uniform resource allocation mode, ignoring the differences of learners in learning style, learning background, learning ability and learning goals, etc., which easily leads to the phenomenon of "learning wandering" of learners, significantly reducing the effect of online learning. Therefore, how to build personalized learning paths to reduce the learning blindness of learners, improve the course learning experience and adapt to their differentiated characteristics has become an urgent and important research topic.

[0003] Learning path is a set of learning resource or learning activity sequence generated according to the prior knowledge and learning goal of learners. The research on learning path is typically represented by the personalized learning path construction method for learning resources and the personalized learning path construction method for learning activity sequence.

[0004] The learning path construction method for learning resources mainly uses the semantic information of knowledge elements contained in learning resources to construct ontology and knowledge graph to generate learning paths. Typical research includes: constructing ontology knowledge base according to the relationship between knowledge elements in learning resources, and generating learning paths through the relationship between ontologies; regarding learning path recommendation as a sorting constraint problem, recommending learning paths based on Bayesian network according to the concept relationship of domain ontology; recommending knowledge points to provide personalized learning path recommendation through the construction of adaptive learning concept map. This kind of method can realize the comprehensive and fine recommendation of learning resources, but whether based on ontology or knowledge graph construction method, the construction of ontology and knowledge graph needs the participation of domain experts, which not only has a huge workload, but also is difficult to efficiently complete the labeling and other links. This leads to the fact that this kind of method can only realize the recommendation in a small range of specified learning resources, and the application is limited. In addition, the association between knowledge elements in ontology and knowledge graph has diversity, and part of the association may not conform to the learning order in the cognitive level of learners, so the learning path constructed by this way is bound to deviate from the actual needs of the individualization of learners.

[0005] The learning path construction method for learning activity sequence mainly generates the optimal learning path by analyzing the learning behavior characteristics exhibited by the online learners in the platform learning process. Typical researches include: a personalized learning system based on decision tree constructs personalized learning path for online learners; a new learning path recommendation system is constructed by combining multi-dimensional attribute collaborative filtering and sequence pattern mining algorithm; a learning path recommendation system is designed based on the preferences, previous performance and course credit requirements of online learners; the learning data, individual characteristics and learning ability of learners are integrated to analyze and track online learners and push the optimal learning path; genetic algorithm is used to generate effective learning path meeting individual needs by combining the learning style and knowledge level of online learners. This kind of learning path generation method for learning activity sequence solves the learning path recommendation problem by taking the learner as the center, but the fusion of the wisdom of online learner group and individual rule is not realized in most researches, so the recommended learning path still has certain limitations, and deviates from the personalized actual needs of learners, which is difficult to realize the targeted guidance in online learning process. SUMMARY

[0006] In view of the problems in the prior art, the present application provides a personalized learning path recommendation method based on online learner portrait clustering mining, a storage medium and equipment, which recommends personalized learning paths through the bidirectional fusion mechanism of matching individual characteristics of online learners to group categories and adapting group rules to individual needs, realizes accurate guidance of online learners in the learning process, effective avoidance of learning wandering, and efficient recommendation of personalized learning paths.

[0007] To achieve the above technical purposes, the present application adopts the following technical solutions: A personalized learning path recommendation method based on online learner portrait clustering mining, comprising the following steps: Step S1: collecting online learning information of online learners, including: basic information of online learners, dynamic learning behavior sequence data and static result data; Step S2: constructing the personalized learning feature sequence of online learners according to the basic information and dynamic learning behavior sequence data of online learners; Step S3: constructing the time sequence learning behavior sequence of online learners according to the dynamic learning behavior sequence data of online learners, and constructing the online learner portrait sequence in combination with the personalized learning feature sequence of online learners; Step S4: classifying all online learner portrait sequences according to the course comprehensive score, and mining the high-contribution personalized learning feature sequence from the personalized learning feature sequence of online learners through information gain rate; Step S5: Cluster the online learner profile sequences based on the high-contribution personalized learning feature sequences to mine the most frequent learning activity sequence patterns of various online learner groups; Step S6: Recommend the personalized learning path to other online learners based on the personalized learning characteristics of the online learner with the highest overall course score in the online learner group under the same maximum frequent learning activity sequence pattern.

[0008] Furthermore, the personalized learning characteristic sequence of the online learner includes: obtaining the characteristics of learning objectives, learning style and learning sprint pace through the basic information of the online learner; and obtaining the characteristics of learning time, learning behavior, learning time patterns and learning pace through the dynamic learning behavior sequence data of the online learner.

[0009] Furthermore, step S3 includes the following sub-steps: Step S3.1: Construct a time-series learning behavior sequence based on the dynamic learning behavior sequence data of online learners. ,in, Indicates the first The time under the timestamp Indicates that online learners in the first Learning interaction sequence data under each timestamp Indicates that online learners in the first Sequence data of learning task results scores under each timestamp; Step S3.2: Combine the online learner's basic information A, personalized learning feature sequence F, static outcome data S, and time-series learning behavior sequence. Constructing a sequence of online learner profiles ,in, This represents a sequence of timestamp intervals.

[0010] Furthermore, in step S4, personalized learning features with an information gain ratio higher than 0.15 are considered as high-contribution personalized learning features.

[0011] Furthermore, the calculation process for the information gain ratio is as follows:

[0012] in, Represents the j-th personalized learning feature Information gain rate Represents the j-th personalized learning feature Information gain , This represents the information entropy of all online learners. , This represents the number of categories into which the profile sequence of all online learners is divided based on their overall course scores. express index, Indicates the first The percentage of online learners in each category Represents the j-th personalized learning feature of all online learners. conditional entropy, , Indicates the first Categories The j-th personalized learning feature Information entropy C represents the first Categories The j-th personalized learning feature Number of categories, Indicates the index of C. Indicates the first The first category Personalized learning features The percentage of online learners; Represents the j-th personalized learning feature The splitting value, , Represents the j-th personalized learning feature Number of categories, express index, Indicates the first Personalized learning features under each category The percentage of online learners.

[0013] Furthermore, step S5 includes the following sub-steps: Step S5.1: Perform one-hot encoding on the high-contribution personalized learning feature sequences of all online learners, cluster them using the DBSCAN algorithm, and assign the online learner profile sequences to the corresponding categories; Step S5.2: For each category of online learner profile sequence, construct the learning activity sequence of each category group using the learning interaction sequence data in the online learner profile sequence; Step S5.3: Perform a global scan of the learning activity sequence for each category group, count the frequency of each individual learning activity in the learning activity sequence, and select the individual learning activities whose frequency meets the frequency requirement as frequent itemsets; Step S5.4: For each individual activity in the frequent itemset, using the individual activity as a prefix, extract all sequence fragments containing the prefix from the learning activity sequence of the corresponding category, and remove the prefix itself and the behavioral elements before the prefix to form a projection database containing only the learning activity sequence following the prefix. Step S5.5: Repeat steps S5.3-S5.4 for the learning activity sequences in the projection database until the projection database is empty or there are no frequent itemsets; Step S5.6: Form a sequence of frequent learning activities from the unidirectional activities in all frequent itemsets of each category group in a recursive order, and find the pattern of the most frequent learning activity sequence.

[0014] Furthermore, the process of finding the maximum frequent learning activity sequence pattern is as follows: for each category group, the frequent learning activity sequences are sorted in descending order of the number of individual activities, and two frequent learning activity sequences are compared one by one in sequence. and ,like All individual activities exist in sequence. In the middle, and The length is less than ,delete ,reserve All retained frequent learning activity sequences are combined into a maximum frequent learning activity sequence pattern.

[0015] Furthermore, the present invention also provides a computer-readable storage medium storing a computer program that enables a computer to execute the personalized learning path recommendation method based on online learner profile clustering mining.

[0016] Furthermore, the present invention also provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, it implements the personalized learning path recommendation method based on online learner profile clustering mining.

[0017] Compared with the prior art, the present invention has the following beneficial effects: (1) The personalized learning path recommendation method based on online learner profile clustering mining of the present invention constructs a personalized learning feature sequence and a time-series learning behavior sequence of online learners, thereby constructing an online learner profile sequence, which can more comprehensively and meticulously describe the online learner profile and provide accurate suggestions for subsequent personalized learning path recommendation; (2) The personalized learning path recommendation method based on online learner portrait clustering mining of the present application eliminates the shortcomings of only focusing on individual learning behavior in the past, mines high-contribution personalized learning feature sequences from the personalized learning feature sequences of online learners through information gain rate, can preferentially select features with high discrimination and non-sparse values, avoids excessive bias towards multi-value noise features, and balances feature discrimination and redundancy control; at the same time, the adaptability to data distribution is good, without assuming that the data conforms to a specific distribution, and the diversity demand of the personalized scene is adapted; (3) The personalized learning path recommendation method based on online learner portrait clustering mining of the present application clusters the online learner portrait sequences through the high-contribution personalized learning feature sequences, mines the maximum frequent learning activity sequence patterns of various online learner groups, on the one hand, makes up for the defects of traditional global frequent sequence mining ignoring group heterogeneity, makes the mining results more in line with the behavior logic of different groups, provides "group-level feature basis" for personalized recommendation, and then generates learning path recommendations with group adaptability and individual pertinence in combination with the maximum frequent learning activity sequence patterns of the group, effectively balances the accuracy and generality of the recommendation. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 The flowchart of the personalized learning path recommendation method based on online learner portrait clustering mining of the present application; Figure 2 The online learning information acquisition schematic diagram of the online learner. DETAILED DESCRIPTION

[0019] The technical solutions of the present application will be further explained and described below in combination with the drawings.

[0020] As Figure 1 The flowchart of the personalized learning path recommendation method based on online learner portrait clustering mining of the present application, the personalized learning path recommendation method comprises the following steps: Step S1: As Figure 2, the online learning information of the online learners is collected from the teaching management system and the online learning platform through the crawler technology and data export, including: basic information of the online learners, dynamic learning behavior sequence data and static result data of the online learners, wherein the basic information of the online learners includes courses, student ID, name, affiliated college and original educational level, the dynamic learning behavior sequence data of the online learners includes student login, student logout, student comments, student completed homework, teacher comments on homework, teacher comment replies and learning task result scores, and the static result data of the online learners includes usual scores, final exam scores and comprehensive scores. The collected online learning information of the online learners is subjected to data cleaning processing, data integration and data conversion, the problems of merging of multi-source data, unifying data format and eliminating attribute redundancy are solved, and consistent data set is formed, so as to guarantee the accuracy, consistency and accuracy of the data.

[0021] Step S2: constructing the personalized learning feature sequence of the online learners according to the basic information and the dynamic learning behavior sequence data of the online learners, as the basis for constructing the learner portrait sequence, describing various features and needs of the learners. As shown in Table 1, the personalized learning feature sequence of the online learners includes: obtaining the features of learning goal, learning style and learning sprint rhythm through the basic information of the online learners, and obtaining the features of learning time, learning behavior, learning time regularity and learning rhythm through the dynamic learning behavior sequence data of the online learners.

[0022] Table 1 Personalized features

[0023] Step S3: constructing the time sequence learning behavior sequence of the online learners according to the dynamic learning behavior sequence data of the online learners, converting the scattered dynamic learning behavior data into continuous trajectory with semantic association and rich semantic information, combining the personalized learning feature sequence of the online learners to construct the online learner portrait sequence, which can more comprehensively and meticulously describe the online learner portrait, and provide accurate suggestions for subsequent personalized learning path recommendation; including the following sub-steps: Step S3.1: performing time sequence consistency verification on the collected dynamic learning behavior sequence, eliminating duplicate records and outliers, and ensuring that each learning behavior data has a unique and continuous time mark; constructing the time sequence learning behavior sequence according to the dynamic learning behavior sequence data of the online learners , wherein, represents the time at the th time stamp, represents the learning interaction sequence data of the online learners at the th time stamp, represents the learning task result score sequence data of the online learners at the th time stamp; Step S3.2: the basic information A of the online learner, the personalized learning feature sequence F, the static result data S and the time sequence learning behavior sequence Constructing the online learner portrait sequence Wherein, represents a time stamp interval sequence, reflecting the change of learning rhythm.

[0024] Step S4: classifying all online learner portrait sequences according to course comprehensive scores, mining high contribution personalized learning feature sequences from the personalized learning feature sequences of online learners through information gain rate, revealing the influence degree of learning behavior patterns on learning effectiveness, retaining effective influence personalized learning features and excluding irrelevant contribution features. Specifically, the personalized learning features with information gain rate higher than 0.15 are regarded as high contribution personalized learning features, which can preferentially select features with high discrimination and non-sparse values, avoid excessive bias towards multi-value noise features, and balance feature discrimination and redundancy control; at the same time, it has good adaptability to data distribution, does not need to assume that the data conforms to a specific distribution, and adapts to the diversity needs of personalized scenarios.

[0025] The information gain rate in the application can solve the preference problem of information gain for multi-value personalized learning features, and automatically screen high contribution personalized learning feature sequences with strong correlation with learning effectiveness. The calculation process of the information gain rate in the application is as follows:

[0026] Wherein, represents the information gain rate of the jth personalized learning feature , represents the information gain of the jth personalized learning feature , , represents the information entropy of all online learners, , represents the number of categories according to the classification of all online learner portrait sequences according to course comprehensive scores, represents the index of , represents the proportion of the number of online learners in the jth category, represents the conditional entropy of the jth personalized learning feature of all online learners, , represents the information entropy of the jth personalized learning feature in the jth category , , represents the information entropy of the jth personalized learning feature in the jth category , and C represents the jth category The j-th personalized learning feature Number of categories Indicates the index of C. Indicates the first The first category Personalized learning features The percentage of online learners; Represents the j-th personalized learning feature The splitting value, , Represents the j-th personalized learning feature Number of categories express index, Indicates the first Personalized learning features under each category The percentage of online learners.

[0027] Step S5: Cluster the online learner profile sequences based on high-contribution personalized learning feature sequences to mine the most frequent learning activity sequence patterns for various online learner groups. This compensates for the shortcomings of traditional global frequent sequence mining, which ignores group heterogeneity, making the mining results more consistent with the behavioral logic of different groups and providing "group-level feature basis" for personalized recommendations. Combined with the most frequent learning activity sequence patterns of this group, a learning path recommendation that combines group adaptability and individual targeting is generated, effectively balancing the accuracy and generalization of the recommendation. This includes the following sub-steps: Step S5.1: Perform one-hot encoding on the high-contribution personalized learning feature sequences of all online learners, cluster them using the DBSCAN algorithm, and assign the online learner profile sequences to the corresponding categories. The clustered online learner groups can be used to mine the learning activity sequence patterns of each group, providing group feature basis for personalized learning path recommendation. Step S5.2: For each category of online learner profile sequence, construct the learning activity sequence of each category group using the learning interaction sequence data in the online learner profile sequence; Step S5.3: Perform a global scan of the learning activity sequence for each category group, count the frequency of each individual learning activity in the learning activity sequence, and select the individual learning activities whose frequency meets the frequency requirement as frequent itemsets; Step S5.4: For each individual activity in the frequent itemset, take the individual activity as a prefix, extract all sequence fragments containing the prefix from the learning activity sequence of the corresponding category, and remove the prefix itself and the behavioral elements before the prefix to form a projection database that only contains the learning activity sequence following the prefix, thereby reducing computational complexity. Step S5.5: Repeat steps S5.3-S5.4 for the learning activity sequences in the projection database until the projection database is empty or there is no frequent item set; Step S5.6: Form the frequent learning activity sequences from the one-way activities in all the frequent item sets in each category group in a recursive order, and find the maximum frequent learning activity sequence pattern.

[0028] The finding process of the maximum frequent learning activity sequence pattern in the present application directly mines on the learning activity sequences without generating candidate sequences, can efficiently process large-scale data, and significantly improves the mining efficiency. Specifically: for the frequent learning activity sequences in each category group, arrange the single activities in descending order of the number of single activities, compare two frequent learning activity sequences in order one by one and , if all the single activities in completely exist in in order, and the length of is less than , delete , keep , and combine all the kept frequent learning activity sequences into the maximum frequent learning activity sequence pattern. The finding process of the maximum frequent learning activity sequence pattern preferentially processes long sequences, and this operation can ensure that the longest frequent sequence is kept first in the subsequent deletion of redundant sequences, and then the contained short sequences are removed in batches based on the long sequences, thereby improving the screening efficiency.

[0029] Step S6: The maximum frequent learning activity sequence pattern can reflect the core learning behavior timing rules of different learner groups in the most concise form, and avoid the pattern redundancy problem in subsequent personalized path recommendation. The individualized learning features of the online learner with the highest comprehensive course result in the online learner group under the same maximum frequent learning activity sequence pattern are recommended to other online learners as the individualized learning path.

[0030] In one technical solution of the present application, a computer readable storage medium is also provided, which stores a computer program. The computer program enables a computer to implement the personalized learning path recommendation method based on online learner portrait clustering mining of the present application.

[0031] In one technical solution of the present application, an electronic device is also provided, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the personalized learning path recommendation method based on online learner portrait clustering mining of the present application is implemented.

[0032] In the embodiments disclosed in the present application, the computer storage medium can be a tangible medium which can contain or store programs for use by or in connection with an instruction execution system, apparatus or device. The computer storage medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus or device, or any suitable combination of the above. More specific examples of computer storage medium can include one or more wires, portable computer disks, hard drives, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), fiber optics, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above.

[0033] Those skilled in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed in the present application can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software manner depends on the specific application and design constraints of the technical solutions. The skilled person can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0034] The above is only the preferred embodiment of the present application, and the protection scope of the present application is not limited to the above-mentioned embodiments. Any technical solution falling within the concept of the present application shall fall within the protection scope of the present application. It should be noted that, for ordinary skilled persons in the art, some improvements and refinements without departing from the principles of the present application shall be considered within the protection scope of the present application.

Claims

1. A personalized learning path recommendation method based on online learner profile clustering mining, characterized in that, Includes the following steps: Step S1: Collect online learning information from online learners, including: basic information of online learners, dynamic learning behavior sequence data, and static outcome data; Step S2: Construct a personalized learning characteristic sequence for online learners based on their basic information and dynamic learning behavior sequence data; Step S3: Construct a time-series learning behavior sequence of online learners based on the dynamic learning behavior sequence data of online learners, and construct an online learner profile sequence by combining the personalized learning feature sequence of online learners; Step S4: Classify all online learner profile sequences according to their overall course scores, and extract high-contribution personalized learning feature sequences from the personalized learning feature sequences of online learners using the information gain ratio; Step S5: Cluster the online learner profile sequences based on the high-contribution personalized learning feature sequences to mine the most frequent learning activity sequence patterns of various online learner groups; Step S6: Recommend the personalized learning path to other online learners based on the personalized learning characteristics of the online learner with the highest overall course score in the online learner group under the same maximum frequent learning activity sequence pattern.

2. The personalized learning path recommendation method based on online learner profile clustering mining according to claim 1, characterized in that, The personalized learning characteristic sequence of online learners includes: characteristics of learning goals, learning styles and learning sprint pace obtained from the basic information of online learners; and characteristics of learning time invested, learning behavior, learning time patterns and learning pace obtained from the dynamic learning behavior sequence data of online learners.

3. The personalized learning path recommendation method based on online learner profile clustering mining according to claim 1, characterized in that, Step S3 includes the following sub-steps: Step S3.1: Construct a time-series learning behavior sequence based on the dynamic learning behavior sequence data of online learners. ,in, Indicates the first The time under the timestamp Indicates that online learners in the first Learning interaction sequence data under each timestamp Indicates that online learners in the first Sequence data of learning task results scores under each timestamp; Step S3.2: Combine the online learner's basic information A, personalized learning feature sequence F, static outcome data S, and time-series learning behavior sequence. Constructing a sequence of online learner profiles ,in, This represents a sequence of timestamp intervals.

4. The personalized learning path recommendation method based on online learner profile clustering mining according to claim 1, characterized in that, In step S4, personalized learning features with an information gain ratio higher than 0.15 are identified as high-contribution personalized learning features.

5. A personalized learning path recommendation method based on online learner profile clustering mining as described in claim 1 or 4, characterized in that, The calculation process for the information gain ratio is as follows: in, Represents the j-th personalized learning feature Information gain rate Represents the j-th personalized learning feature Information gain , This represents the information entropy of all online learners. , This represents the number of categories into which the profile sequence of all online learners is divided based on their overall course scores. express index, Indicates the first The percentage of online learners in each category Represents the j-th personalized learning feature of all online learners. conditional entropy, , Indicates the first Categories The j-th personalized learning feature Information entropy C represents the first Categories The j-th personalized learning feature Number of categories Indicates the index of C. Indicates the first The first category Personalized learning features The percentage of online learners; Represents the j-th personalized learning feature The splitting value, , Represents the j-th personalized learning feature Number of categories express index, Indicates the first Personalized learning features under each category The percentage of online learners.

6. The personalized learning path recommendation method based on online learner profile clustering mining according to claim 1, characterized in that, Step S5 includes the following sub-steps: Step S5.1: Perform one-hot encoding on the high-contribution personalized learning feature sequences of all online learners, cluster them using the DBSCAN algorithm, and assign the online learner profile sequences to the corresponding categories; Step S5.2: For each category of online learner profile sequence, construct the learning activity sequence of each category group using the learning interaction sequence data in the online learner profile sequence; Step S5.3: Perform a global scan of the learning activity sequence for each category group, count the frequency of each individual learning activity in the learning activity sequence, and select the individual learning activities whose frequency meets the frequency requirement as frequent itemsets; Step S5.4: For each individual activity in the frequent itemset, using the individual activity as a prefix, extract all sequence fragments containing the prefix from the learning activity sequence of the corresponding category, and remove the prefix itself and the behavioral elements before the prefix to form a projection database containing only the learning activity sequence following the prefix. Step S5.5: Repeat steps S5.3-S5.4 for the learning activity sequences in the projection database until the projection database is empty or there are no frequent itemsets; Step S5.6: Form a sequence of frequent learning activities from the unidirectional activities in all frequent itemsets of each category group in a recursive order, and find the pattern of the most frequent learning activity sequence.

7. The personalized learning path recommendation method based on online learner profile clustering mining according to claim 6, characterized in that, The process of finding the maximum frequent learning activity sequence pattern is as follows: For each category group, the frequent learning activity sequences are sorted in descending order by the number of individual activities, and two frequent learning activity sequences are compared one by one in sequence. and ,like All individual activities exist in sequence. In the middle, and The length is less than ,delete ,reserve All retained frequent learning activity sequences are combined into a maximum frequent learning activity sequence pattern.

8. A computer-readable storage medium storing a computer program, characterized in that, The computer program causes the computer to execute the personalized learning path recommendation method based on online learner profile clustering mining as described in any one of claims 1-7.

9. An electronic device, characterized in that, include: The memory, the processor, and the computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, it implements the personalized learning path recommendation method based on online learner profile clustering mining as described in any one of claims 1-7.

Images