An online behavior-based learning style inclination adaptability diagnosis system and method

Abstract

The present application relates to a kind of learning style tendency adaptive diagnostic system and method based on online behavior, including data storage module, data acquisition module, learning style tendency diagnostic module, adaptive service module and user interface module;Data storage module supports the storage of learning material, interactive activity, process behavior, learning style index, user information and related additional information;Data acquisition module carries out acquisition and record to the behavior generated by learner on learning content object in learning process;Learning style tendency diagnostic module is based on the classification index of learning style and the behavior of learner accessing different types of learning content object, and the calculation of learning style is carried out;Adaptive service module provides learning material, learning activity and learning mode suitable for the style of learner based on diagnosed learning style;User interface layer realizes the support of adaptive learning based on multi-terminal to user.The present application promotes the promotion of online learning adaptability and learning effectiveness.

Description

Technical Field

[0001] This invention relates to an adaptive diagnostic system and method for learning style preferences based on online behavior. Specifically, it involves a technical approach that predicts and diagnoses learners' learning style preferences based on their learning behaviors in different scenarios, combined with learning style theory and its reflection in learning content, and provides adaptive services for them. This belongs to the field of adaptive diagnostic technology for learning styles in online learning. Background Technology

[0002] In the online learning environment, learners' individual characteristics and learning needs are becoming increasingly diverse, leading to vastly different demands for learning resources. Learning style (Zhao Hong et al., 2015), as an important manifestation of learners' preferences, reflects their tendency to select and use resources and their learning strategies during online learning. Accurately identifying learners' preferences and providing them with learning resources suitable for their style has a significant impact on ensuring the effectiveness of online learning. Therefore, learning style has become an important factor in online learning design. Some researchers have begun to integrate learning style into the design of online learning platform functions and provide learners with personalized content. However, current research on the diagnosis of learning style is mostly based on questionnaires and scales. Although learning style is a relatively stable trait, learners' learning style tendencies differ at specific stages. As learning progresses and learning scenarios change, learners' learning styles exhibit a certain degree of dynamic change (Mario, S., Thomas, M. & Thomas, H (2015). Learning style analysis in adaptive GBL application toteach SQL[J]. Computers & Education, 86: 105-119.). This also leads to diagnostic biases caused by lagging and non-real-time updating methods such as scales and questionnaires, making it difficult for online learning systems to provide learners with the most adaptive learning materials and activities.

[0003] In the process of learning style diagnosis methods and system design, researchers currently focus mainly on scale-based diagnosis. For example, some researchers have designed adaptive educational game structure models for learners of different styles based on the cognitive differences between sequential and comprehensive learners, using pre-tests to support primary school students' learning of four arithmetic operations (Li Xin (2015). Design Model and Experimental Study of Educational Game Structure Incorporating "Adaptability" [J]. Journal of Distance Education, 33(02):97-103.); other researchers have provided a more adaptive presentation for learners of different learning styles through questionnaires (Yang, T., Hwang, G. & Yang, S. (2013). Development of an Adaptive Learning System with Multiple Perspectives based on Students' Learning Styles and Cognitive Styles [J]. Educational Technology & Society, 16(4):185-200.). However, overall, the above methods lack dynamic perception of learner style changes and are difficult to meet learners' needs for adaptive learning. Therefore, some researchers have begun to focus on dynamically calculating learning styles through learners' behavior. For example, Jiang Qiang et al. analyzed the characteristics of different learning styles and constructed a method based on Bayesian networks to mine learners' behavioral patterns and infer learning styles (Jiang Qiang, Zhao Wei, Wang Pengjiao (2012). Research on the construction of user learning style model based on network learning behavior pattern mining [J]. E-Education Research, 33(11):55-61.). In addition, Yin Chuantao et al. constructed an adaptive recommendation method and system based on learning styles, which predicts learners' learning styles through learners' learning behavior on websites. On the one hand, learning resources are labeled with learning style vectors, and on the other hand, learners are defined with 8-dimensional vectors of learning styles. Then, the learners' learning styles are calculated through the behaviors generated by the learners, which are used to match specific learning resources (Yin Chuantao, Zhang Xiaoyan, Sun Honglu, Qiao Lei, Guan Minghui. An adaptive recommendation method and system based on learning styles [P]. Beijing: CN109213863B, 2022-05-24.); Lin Yi et al. constructed an adaptive virtual reality teaching method and system based on a learning style model. In this method, the style discrimination based on behavior, the teaching optimization based on learning style, and the feedback based on optimization are completed in three steps (Lin Yi, Wang Shunbo, Lan Yangfan, Wu Bingkun. An adaptive virtual reality teaching method and system based on a learning style model [P]. Fujian Province: CN111179135A, 2020-05-19.).

[0004] In summary, current research and practice on learning style assessment and diagnosis are gradually shifting from questionnaire-based assessment to data-driven assessment. Some researchers have begun to explore using learners' online learning behaviors to assess their learning styles, using single-dimensional behaviors such as the frequency of "browsing" and "staying" on different types of documents during online learning to support the correspondence between behavior and learning style. The problems are as follows: (1) Learners' learning styles are reflected in their interaction with diverse learning materials and activities throughout the entire online learning process. Different materials and activities play different roles in the diagnosis of different dimensions of learning style. Current research and patents have not yet addressed the relationship between this content object and learning style. (1) The design and consideration of the associations in different dimensions are required, so it is necessary to construct the constraint relationship between the learning content object and the learning style; (2) The behaviors of learners in different content materials and activities are complex and diverse, and different behaviors have different strengths and weaknesses in diagnosing learning styles. Current patents and research only consider the frequency of behaviors and fail to consider the weight of different behaviors in evaluating learning styles. Therefore, it is necessary to construct a multi-level interactive behavior model for different content objects in online learning; (3) The learning behaviors used to diagnose learning styles have a weakening effect over time, and current research and patents have not paid enough attention to this point. Therefore, it is necessary to construct a learning style tendency diagnosis model based on time evolution. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to overcome the shortcomings of the prior art and provide a learning style tendency adaptive diagnosis system and method based on online behavior. By diagnosing the learner's learning style, the system determines the learner's preference for learning materials and provides adaptive services, thereby promoting the improvement of online learning adaptability and learning effectiveness.

[0006] Technical solution of the present invention:

[0007] In a first aspect, the present invention provides a learning style tendency adaptive diagnostic system based on online behavior, comprising: a data storage module, a data acquisition module, a learning style tendency diagnostic module, an adaptive service module, and a user interface module;

[0008] The data storage module establishes six sub-databases: a learning material database, an interactive activity database, a process behavior database, a learning style model database, an individual information database, and a supplementary information database. The learning material database stores different types of learning materials and their sets during the learner's online learning process. The interactive activity database stores different types of interactive activities and their sets during the learner's online learning process. The process behavior database stores all learning behaviors and their frequencies generated by the learner through interaction with learning materials and activities during online learning. The learning style model database stores well-known learning style models in the field of education and their different dimensions. The individual information database stores the learner's individual information, including basic information, device information, and preference information. The supplementary information database stores relevant supplementary information, including well-known typical learning scenarios and learning modes in the current online learning field. The learning mode refers to the organization of different types of learning materials during the learning process. Furthermore, this supplementary information database provides an extensible interface for supplementing relevant supplementary information.

[0009] Data acquisition module: Used to support the standardized acquisition of learning behaviors generated by learners in well-known typical learning scenarios in the current online learning field. During the acquisition process, the acquired data needs to be preprocessed and formatted to facilitate data classification and storage; the well-known typical learning scenarios include MOOC course learning, online self-study, online collaborative learning and blended learning;

[0010] Learning style tendency diagnosis module: Construct a learning style tendency diagnosis algorithm that integrates three-dimensional features of "learning style model", "learning content object", and "learning behavior". This algorithm links the "learning style model", "learning content object", and "learning behavior" to form a learning style tendency of learners using a series of learning behaviors generated on the learning content object as an intermediary. The learning content object includes learning materials and interactive activities, and the learning behavior refers to a series of behaviors generated on the learning content object.

[0011] To support the dynamic evolution of learning style preference diagnosis, the learning style preference diagnosis module also incorporates the learning behavior occurrence time parameter as a dynamic learning style preference diagnosis method. The dynamic learning style preference diagnosis method constructs a time-based learning behavior weighting function, which calculates the weighted learning behavior at different time stages when collecting learning behavior, thereby strengthening the most recently occurring learning behavior and weakening the learning behavior of a longer period of time, and dynamically fine-tuning the learning style preference diagnosis over time, ultimately obtaining the learning style preference.

[0012] Adaptive Service Module: Based on learning style preferences, it overlays individual information and supplementary information stored in the individual information database and supplementary information database to form a learner feature vector. The feature vector determines the type of learning materials and learning mode that the learner is suited for. Based on the vector similarity calculation method, it selects the learning materials and interactive activities with the highest matching degree with the learner from the learning material database and interactive activity database of the data storage module for recommendation and output.

[0013] User Interface Module: This module encapsulates the functions of the learning style diagnosis module and the adaptive service module, providing learners with an online learning interface through smart terminals. In supporting learners' learning process, the user interface module provides learners with adaptive learning materials and interactive activities, and also ensures data collection for the data acquisition module based on the sensing and recording devices equipped on the terminal.

[0014] Furthermore, the learning style preference diagnosis algorithm in the learning style preference diagnosis module is implemented as follows:

[0015] (1) Merge the set of learning materials in the learning material library and the set of interactive activities in the interactive activity library to form a related set of the two, and determine the learning content object;

[0016] (2) Extract the set of all learning behaviors and operation frequencies based on the learning content object to realize the binary association between "learning content object" and "learning behavior" and obtain the association information;

[0017] (3) Establish the association between different dimensions in the learning style model and the learning content object, that is, associate different learning content objects with different dimensions in the learning style model to obtain the association between "learning style model" and "learning content object", superimpose the association information in step (2), and obtain the set of all learning behaviors and operation frequencies of different dimensions of each learning style model in turn; then set the calculation weight of different dimensions of each learning style model according to the learner's individual information, additional information and the types of all learning behaviors on the learning content object;

[0018] (4) Under natural learning conditions, calculate the learning style tendency scores of all learners in different dimensions of the learning style model, and sort the learning style tendency scores of each dimension to obtain the learning style tendency of the learners.

[0019] Furthermore, the learning style preference diagnosis algorithm is specifically implemented as follows:

[0020] (1) Merge the learning material set MaterialClassification and the interactive activity set ActivityClassification to form the associated set ContentObjectTypeSet of all learning content objects, which is then determined as the learning content object;

[0021] The learning material set MaterialClassification = {learning material category 1, learning material category 2, learning material category 3, learning material category 4, ..., learning material category n};

[0022] The set of interactive activities is ActivityClassification = {Activity Class 1, Activity Class 2, Activity Class 3, ..., Activity Class m};

[0023] Related collection ContentObjectTypeSet=

[0024] {ContentObject1,ContentObject2,……,ContentObject i}

[0025] ContentObject i ContentObject represents a specific category of learning materials or interactive activities within the ActivityClassification or MaterialClassification. i ∈{learning material category 1, learning material category 2, ..., learning material category n, interactive activity category 1, interactive activity category 2, ..., interactive activity category m}, representing different types of learning materials;

[0026] For each type of ContentObject in different learning content object categories i It includes the Material set of all learning materials and interactive activities generated by learners. i,j ,Right now:

[0027] ContentObject i ={Material i,1 Material i,2 Material i,j};

[0028] Among them, Material i,j This indicates that the learning material category is ContentObject. iThe j-th specific learning material or interactive activity (such as ContentObject) in the ... i If it represents video material, then Material i,j This represents the j-th video in the learning materials of the video category.

[0029] (2) Based on all learning materials and interactive activities in the associated set ContentObjectTypeSet, retrieve the learner's Material for each learning material or interactive activity from the process behavior library. i,j The set of learning behaviors and their operation frequencies generated on different learning content objects constitutes CB, which is the set of all learning behaviors and operation frequencies of learners on different learning content objects, thereby associating learning content objects with learning behaviors.

[0030] CB = B⊙M={

[0031] B1=COFrequency{R 11 ,R 12 ,…R 1t},

[0032] B2=COFrequency{R 21 ,R 22 ,…R 2t},

[0033] …

[0034] B i =COFrequency{R i1 ,R i2 B ij , …, R it},

[0035] …

[0036] B n =COFrequency{R n1 ,R n2 ,…R nt}

[0037] }

[0038] Where B is the complete set of all learning behaviors of the learner, and ⊙M represents the learner's learning behavior in M, i.e., Material. i,j Operational frequency of all learning behaviors; B n R represents the set of learning content objects and operation frequencies in all nth type of learning behavior during the learner's learning process, and is a subset of the universal set of learning behaviors B for a specific learning behavior type; here, R is used. ij This represents the behavior corresponding to any learning content object in CB (such as R).11 ,R 12 …), which means the set of learning content objects and operation frequencies of the j-th type in the i-th type of learning behavior, where i∈{1,2,……,n}, j∈{1,2,……,t};

[0039] (3) Based on the CB obtained in step (2), establish the association between different dimensions in the learning style model and the learning content object, that is, map different types of content in the learning content object to different dimensions in the learning style model. Using the different dimensions of each learning style model as a distinction, obtain the learning content object and operation frequency that can reflect the different dimensions in each learning style model, and form a subset CB of the different dimensions StyleK of the learning style model. styleK :

[0040] CB styleK = {

[0041] BS1 = COFrequency styleK {RSK 11 RSK 12 ,…RSK 1t},

[0042] BS2 = COFrequency styleK {RSK 21 RSK 22 ,…RSK 2t},

[0043] …

[0044] BS i = COFrequency styleK {RSK i1 RSK i2 ,…RSK ij ,…RSK it},

[0045] …

[0046] BS n = COFrequency styleK {RSK n1 RSK n2 ,…RSK nt},

[0047] };

[0048] Among them: RSK i1 ~ RSK itThe operation frequency of the learning content object in the i-th type of learning behavior can reflect the different dimensions of StyleK of the current learning style model, where RSK is not a dimension of the current learning style model. ij Assign 0 directly, where i∈{1,2,……,n}, j∈{1,2,……,t}, RSK ij For R ij A subset of behavioral frequencies along the style dimension of the styleK learning process;

[0049] (4) Based on the above CB, and combining individual information and additional information settings (the two are combined into PersonalInfo), calculate the weights wi for different types of learning behaviors. Based on wi, obtain the learner's final learning style tendency coefficient StyleScore; where the maximum value of each dimension in the learning style model is the learner's learning style tendency in the current dimension:

[0050] StyleScore =Max(W CB)

[0051] in:

[0052] CB = B ⊙ M

[0053] Where: B =

[0054] M =

[0055] W = [w1, w2,…,wi,…,wn]

[0056] Therefore, StyleScore =

[0057] Let wi be the i-th element in W, wi = {Behavior, BehaviorType, Scale, Note}, where i ranges from 1 to n, and Behavior represents the name of a specific learning behavior. Then, its corresponding Bij (i, j do not represent any meaning, but are only used for reference and counting, and are the values ​​in the i-th row and j-th column of the vector space) represents a specific learning behavior that occurred. M The BehaviorType represents the object type corresponding to the behavior that occurred, the Scale represents the weight of different types of learning behavior, and the Note represents the learning style reflected by the learning behavior, i.e., the different dimensions of the learning style model.

[0058] The scales corresponding to the different types of learning behaviors constitute a set of weights called ScaleSet:

[0059] ScaleSet = {

[0060] PersonalInfo{Scale 11 ,Scale 12 ,……,Scale 1m},

[0061] CreateBehavior{Scale 21 ,Scale 22 ,……,Scale 2n},

[0062] InteractiveBehavior{Scale 31 ,Scale 32 ,……,Scale 3s},

[0063] ReceiveBehavior{Scale 31 ,Scale 32 ,……,Scale 3t}

[0064] }

[0065] Where: PersonalInfo{ Scale 11 ,Scale 12 ,……,Scale 1m} represents the weights used in the calculation process of superimposing different types of individual information and additional information. CreateBehavior{ Scale 21 ,Scale 22 ,……,Scale 2n} represents the weights corresponding to different creative behaviors during the calculation process, InteractiveBehavior{ Scale 31 ,Scale 32 ,……,Scale 3s} represents the weights corresponding to different interactive behaviors during the calculation process, ReceiveBehavior{ Scale 31 ,Scale 32 ,……,Scale 3t} represents the weights corresponding to different receptive behaviors in the calculation process; the creative behavior refers to the behavior type at the level of creation in the learning behavior (such as creating resources), the interactive behavior refers to the behavior type at the level of interaction in the learning behavior (such as commenting), and the receptive behavior refers to the behavior type at the level of reception in the learning behavior (such as browsing).

[0066] Finally, the values ​​of different dimensions in the StyleScore are sorted in order of magnitude, and the largest value is the learner's learning style tendency (Tend).

[0067] Furthermore, based on classifications in the field of learning science, a deep learning behavior model (DSM) is constructed, encompassing three levels of interactive behavior and reflecting different depths of learner learning: the Double Helix Deep Learning Behavior Model.

[0068] DSM = {

[0069] Receptive behaviors: {browsing, searching}

[0070] Interactive behaviors: {Request collaboration, join learning, edit, comment, annotate, download, favorite, follow, share and recommend, watch video, drag progress, skip current content}

[0071] Creative behaviors: {Creating knowledge points, creating resources, creating activities, creating evaluation schemes, posting, and approving}

[0072] }

[0073] To support learning behavior statistics, BehaviorClassification is constructed to integrate all types of learning behaviors corresponding to the above learning behavior models, supporting the acquisition of all learning behaviors B of the learner, i.e., the complete set of learning behaviors:

[0074] BehaviorClassification={Browse, Search, Request Collaboration, Join Learning, Edit, Comment, Annotate, Download, Favorite, Follow, Share & Recommend, Watch Video, Drag Progress, Skip Current Content, Create Knowledge Point, Create Resource, Create Activity, Create Evaluation Scheme, Publish Post, Review, {Extended Behaviors}}.

[0075] Furthermore, the well-known learning style models in the field of education include: the Felder-Silverman model, the Kolb model, and the VARK model;

[0076] The Felder-Silverman learning style model is represented by different dimensions as follows: FS{Visual, Verbal, Intuitive, Perceptive, Active, Reflective, Global, Sequential};

[0077] The Kolb learning style models of different dimensions are represented as follows: Kolb {assimilation, accommodation, divergence, centralization};

[0078] The different dimensions of the VARK learning style model are: VARK {visual, auditory, reading and writing, experiential}.

[0079] Furthermore, the data acquisition module collects data based on the built-in sensors of the learning device and the log points of the learning platform. During the acquisition process, the collected data is preprocessed and formatted according to the data from different sources. The standardized acquisition of the data is based on the built-in sensors of mobile phones, tablets and PCs, the log points of the online learning platform, and the learner's interactive activities.

[0080] Furthermore, to consider the role of behaviors occurring at different times in influencing the diagnosis of learning style tendencies, this invention establishes a time-based behavior weighting method. The time-based behavior weighting function method is as follows: First, the set of learning behaviors is divided into k segments according to their chronological order, and these k segments are sequentially denoted as p1, p2, ..., p... k The corresponding time nodes ts for the k segments are 1 / k, 2 / k, ..., k-1 / k, 1. Then, the logistic function is used to map the weights of the learning behavior.

[0081] W(ts) =

[0082] W(ts) represents the weight of the learning behavior within the time period corresponding to the ts-th node, where the logistic function is the weighted parameter that adjusts the behavior at the current time node by using time ts as a variable.

[0083] Ultimately, this will enable the calculation of learners' learning style preferences that evolve dynamically over time.

[0084] Furthermore, the recommendation process is based on vector similarity, that is, constructing representation vectors MR of learning materials and interactive activities according to the learning style tags suitable for learning materials and activities, obtaining the similarity between the representation vector MR and Tend, and the learning materials and interactive activities with the highest similarity are the ones to be recommended.

[0085] Secondly, the present invention provides a method for implementing a learning style tendency adaptive diagnostic system based on online behavior, which is implemented as follows:

[0086] (1) Construct a learning style tendency diagnosis algorithm that integrates three-dimensional features of "learning style model", "learning content object" and "learning behavior". The algorithm associates the "learning style model", "learning content object" and "learning behavior" to form a learning style tendency of learners using a series of learning behaviors generated on the learning content object as the medium. The learning style model is stored in the learning style model library of the data storage module. The learning content object is stored in the learning material library and interactive activity library of the data storage module. The learning behavior is stored in the process behavior library of the data storage module and is collected through the data acquisition module.

[0087] (2) Real-time diagnosis of learners' learning style preferences when learners engage in online learning:

[0088] (21) First, extract all the learning materials and interactive activities that learners interact with during the online learning process, and merge them to obtain the associated set of all learning content objects, ContentObjectTypeSet, which is the learning content object;

[0089] The learning material set MaterialClassification = {learning material category 1, learning material category 2, learning material category 3, learning material category 4, ..., learning material category n}

[0090] The set of interactive activities is ActivityClassification = {Activity Class 1, Activity Class 2, Activity Class 3, ..., Activity Class m};

[0091] Related collection ContentObjectTypeSet=

[0092] {ContentObject1,ContentObject2,……,ContentObject i}

[0093] ContentObject i ContentObject represents a specific category of learning materials or interactive activities within the ActivityClassification or MaterialClassification. i ∈{learning material category 1, learning material category 2, ..., learning material category n, interactive activity category 1, interactive activity category 2, ..., interactive activity category m}, representing different types of learning materials;

[0094] For each type of ContentObject i It includes the Material set of all learning materials and interactive activities generated by learners. i,j ,Right now:

[0095] ContentObject i ={Material i,1 Material i,2 Material i,j};

[0096] Among them, Material i,jThis indicates that the content object category is ContentObject. i The j-th specific learning material or interactive activity (such as ContentObject) in the ... i If the content refers to video-based learning materials, then Material i,j This indicates the j-th video in the video category.

[0097] (22) Based on all the learning materials and interactive activities in ContentObjectTypeSet, retrieve the different Material types for each learning material or interactive activity from the process behavior library. i,j The set of learning behaviors and operation frequencies generated above constitutes CB, which associates the learning content object with the learning behavior.

[0098] CB = B⊙M={

[0099] B1=COFrequency{R 11 ,R 12 ,…R 1t},

[0100] B2=COFrequency{R 21 ,R 22 ,…R 2t},

[0101] …

[0102] B i =COFrequency{R i1 ,R i2 B ij , …, R it},

[0103] …

[0104] B n =COFrequency{R n1 ,R n2 ,…R nt}

[0105] }

[0106] Where B is the complete set of all learning behaviors of the learner, and ⊙M represents the learner's learning behavior in M, i.e., Material. i,j The frequency of all learning behaviors; R ij Let B represent the set of learning content objects and operation frequencies of the j-th type in the i-th type of learning behavior, where i∈{1,2,……,n}, j∈{1,2,……,t}; nThis represents the set of learning content objects and operation frequencies in all nth type of learning behavior during the learner's learning process;

[0107] (23) Based on CB, establish the association between different dimensions in the learning style model and the learning content object, that is, map different types of content in the learning content object to different dimensions in the learning style model. Using the different dimensions of each learning style model as the distinction, obtain the learning content object and operation frequency that can reflect each different dimension in each learning style model, forming a subset of CB in each different dimension of the learning style model. For styleK learning style, the following set is obtained:

[0108] CB styleK = {

[0109] BS1 = COFrequency styleK {RSK 11 RSK 12 ,…RSK 1t},

[0110] BS2 = COFrequency styleK {RSK 21 RSK 22 ,…RSK 2t},

[0111] …

[0112] BS i = COFrequency styleK {RSK i1 RSK i2 ,…RSK ij ,…RSK it},

[0113] …

[0114] BS n = COFrequency styleK {RSK n1 RSK n2 ,…RSK nt},

[0115] };

[0116] Among them: RSK i1 ~ RSK it The operation frequency of the learning content object in the i-th type of learning behavior can reflect the different dimensions of StyleK of the current learning style model, where RSK is not a dimension of the current learning style model. ijAssign 0 directly, where i∈{1,2,……,n}, j∈{1,2,……,t}, RSK ij For R ij A subset of behavioral frequencies along the style dimension of the styleK learning process;

[0117] (24) Establish the association between different dimensions in the learning style model and the learning content object, that is, associate different learning content objects with different dimensions in the learning style model to obtain the association between "learning style model" and "learning content object", superimpose the binary association between "learning content object" and "learning behavior", and obtain all behaviors and their frequency sets of different dimensions of each learning style model in turn. Then, combine individual information and additional information to set the calculation weights wi of different types of learning behaviors, wi={Behavior, BehaviorType, Scale, Note}.Scale, where Behavior represents the specific learning behavior name, BehaviorType represents different types of learning behavior, Scale represents the weight of different types of learning behavior, and Note represents the learning style index reflected by the learning behavior; the set of the above calculation weights ScaleSet is as follows:

[0118] ScaleSet = {

[0119] PersonalInfo{Scale 11 ,Scale 12 ,……,Scale 1m},

[0120] CreateBehavior{Scale 21 ,Scale 22 ,……,Scale 2n},

[0121] InteractiveBehavior{Scale 31 ,Scale 32 ,……,Scale 3s},

[0122] ReceiveBehavior{Scale 31 ,Scale 32 ,……,Scale 3t}

[0123] }

[0124] Where: PersonalInfo{ Scale 11 ,Scale 12 ,……,Scale 1m} represents the weights used in the calculation process of superimposing different types of individual information and additional information. CreateBehavior{ Scale 21 ,Scale 22 ,……,Scale 2n} represents the weights corresponding to different creative behaviors during the calculation process, InteractiveBehavior{ Scale 31 ,Scale 32 ,……,Scale 3s} represents the weights corresponding to different interactive behaviors during the calculation process, ReceiveBehavior{ Scale 31 ,Scale 32 ,……,Scale 3t} represents the weights corresponding to different receptive behaviors in the calculation process; the creative behavior refers to the behavior type at the level of creation in the learning behavior (such as creating resources), the interactive behavior refers to the behavior type at the level of interaction in the learning behavior (such as commenting), and the receptive behavior refers to the behavior type at the level of reception in the learning behavior (such as browsing).

[0125] (25) Finally, the learner’s final learning style preference coefficient StyleScore is obtained by associating the above weights, learner behavior and learning content objects;

[0126] StyleScore =Max(W CB);

[0127] CB = B⊙M ;

[0128] B =

[0129] M =

[0130] W = [w1, w2, ..., wi, ..., wn];

[0131]

[0132] Let wi be the i-th element in W, and wi = {Behavior, BehaviorType, Scale, Note}, where i ranges from 1 to n;

[0133] (26) The maximum value of each dimension in the learning style model is the learner's learning style tendency in the current dimension. The values ​​of different dimensions in StyleScore are sorted in order of size, and the largest value is the learner's learning style tendency Tend.

[0134] Thirdly, the present invention provides an electronic device, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus;

[0135] Memory, used to store computer programs;

[0136] A processor is used to execute computer programs stored in memory, which, when executed, implement the system or the method described herein.

[0137] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the system or the method described herein.

[0138] The advantages of this invention compared to existing technologies are as follows: Providing learners with learning materials and services is crucial for learning effectiveness, but current adaptive learning systems lack dynamic, systematic, and adaptive diagnostic methods for learning styles, leading to problems such as insufficient matching and sustainability of adaptive services. This invention constructs a three-dimensional adaptive diagnostic method for learning style tendencies, based on a "learning style model - learning content object - learning behavior model." Using learning content objects as an intermediary, it diagnoses learners' learning styles through learning behaviors. On one hand, it collects learners' behaviors and interactions on diverse content objects during online learning; on the other hand, it constructs the constraint relationships between learning styles and different content objects, as well as the weight sets of different learning behaviors in reflecting learning styles. This supports real-time, dynamic diagnosis of learning style tendencies, allowing learners to assess the dynamic changes in their learning style tendencies based on their continuously generated behaviors during online learning. This provides conditions for improving adaptive learning service capabilities and supports the improvement of learners' online learning adaptability and learning effectiveness, promoting increased learning adaptability and efficiency. Attached Figure Description

[0139] Figure 1 This is a schematic diagram of the learning style adaptation diagnostic system according to an embodiment of the present invention;

[0140] Figure 2 This is a schematic diagram illustrating the operation of a data acquisition and storage module according to an embodiment of the present invention;

[0141] Figure 3 This is a schematic diagram illustrating the learning style annotation for learning materials or activities according to an embodiment of the present invention;

[0142] Figure 4 This is a flowchart illustrating the learning style diagnosis module according to an embodiment of the present invention;

[0143] Figure 5This is the result of a learner's learning style diagnosis according to one embodiment of the present invention;

[0144] Figure 6 This is a logical diagram illustrating the operation of an adaptive service module according to an embodiment of the present invention. Detailed Implementation

[0145] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0146] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments.

[0147] like Figure 1 As shown, this invention discloses an adaptive diagnostic system for learning style preferences based on online behavior, comprising: a data storage module, a data acquisition module, a learning style preference diagnostic module, an adaptive service module, and a user interface module. The data storage module stores learning materials, interactive activities, procedural behaviors, learning style indicators, user information, and related supplementary information supporting learning style diagnosis. The data acquisition module collects and records learners' behaviors on specific learning content objects (learning materials + interactive activities) during the learning process. The learning style diagnostic module models and calculates learning styles based on learning style classification indicators and learners' behaviors when accessing different types of learning content objects. The adaptive service module provides learners with learning materials, learning activities, and learning modes adapted to their diagnosed learning styles. Finally, the user interface layer supports user adaptive learning across multiple terminals. This invention is applicable to online learning processes, determining learners' preferences for learning materials and learning styles through learning style diagnosis, providing adaptive services, and thereby promoting the improvement of online learning adaptability and learning effectiveness.

[0148] like Figure 1 As shown, the data storage module includes six sub-databases: learning material database, interactive activity database, process behavior database, learning style model database, individual information database, and supplementary information database.

[0149] The learning resource library stores different types of learning materials and collections used by learners during their online learning process;

[0150] Interactive activity library, storing different types of interactive activities and sets of them during learners' online learning process;

[0151] The process behavior library stores the set of all learning behaviors and operation frequencies generated by learners during online learning through interaction with learning materials and activities.

[0152] The learning style model library stores well-known learning style models in the field of education, along with sets of different dimensions. Well-known learning style models in the field of education include: the Felder-Silverman model, the Kolb model, and the VARK model.

[0153] The Felder-Silverman model is represented in different dimensions as follows: FS{Visual, Verbal, Intuitive, Perceptive, Active, Reflective, Global, Sequential};

[0154] The Kolb learning style models of different dimensions are represented as follows: Kolb {assimilation, accommodation, divergence, centralization};

[0155] The different dimensions of the VARK learning style model are: VARK {visual, auditory, reading and writing, experiential}.

[0156] An individual information database stores learners' individual information, including basic information, device information, and preference information;

[0157] The supplementary information database stores relevant supplementary information, including well-known typical learning scenarios and learning modes in the current online learning field. The learning mode refers to the organization of different types of learning materials during the learning process. In addition, the supplementary information database also provides an extensible interface for supplementing relevant supplementary information.

[0158] like Figure 1 As shown, the data acquisition module is used to support the standardized acquisition of learner behaviors in well-known typical learning scenarios in the current online learning field. During the acquisition process, the acquired data needs to be preprocessed and formatted to facilitate data classification and storage. Well-known typical learning scenarios include MOOC course learning, online self-study, online collaborative learning, and blended learning. The data acquisition module collects data based on the built-in sensors of learning devices and the log points of the learning platform. During the acquisition process, the collected data is preprocessed to format data from different sources. The standardized acquisition of the data is based on the built-in sensors of mobile phones, tablets, and PCs, the log points of online learning platforms, and learner interaction activities.

[0159] like Figure 2 As shown, the specific implementation of the data acquisition module is as follows:

[0160] (1) For the data to be entered into the database, it is classified according to its data characteristics during the collection process: learning materials, interactive activities, process behavior, learning style information, individual information and additional information;

[0161] (2) Learning style information, individual information and additional information are mainly stored directly by predefinition and import or user input;

[0162] (3) For the learning materials and interactive activities, which correspond to the learning content objects in the learning style diagnosis module, it is necessary to annotate the uploaded learning materials and interactive activities in a semi-automatic (manual + machine annotation) manner. The annotation information is mainly based on the learning style classification, individual user preferences, and scenario information. The semi-automatic annotation is as follows: Figure 3 As shown in the figure, labels are set for different learning content objects to indicate learning styles. The labeling is automatically done according to the constraint relationship between learning content objects and learning styles in Table 1. Inaccurate labels can be manually revised. For example, if a learning content object is a concept map, then according to Table 1, it should be visual at the perception level, insightful and contemplative at the information processing level, and global at the content reasoning level. Similarly, for other types of learning content objects, the learning style can be associated with different learning content objects according to the correspondence in Table 1.

[0163] Table 1. Constraints on the Relationship between Learning Content Target and Learning Style

[0164]

[0165] (4) For process behavior data, which corresponds to the learning behavior model of the learning style diagnosis module, it is necessary to clarify the learning content object associated with it. At the same time, since the process behavior mainly depends on user generation during the online learning process, the method of real-time log of user online operation object is mainly adopted in the process of obtaining learning content object, and its format is standardized in the xAPI method to form behavior = {learning behavior ID, learner ID (learning behavior subject), learning behavior type, learning content object ID (learning behavior object), learning behavior time, learning behavior duration, learning behavior scenario, learning behavior result};

[0166] (5) Finally, the six types of data are stored in the database according to their attributes, providing conditions for subsequent learning style tendency diagnosis and adaptive services.

[0167] like Figure 4As shown, the learning style tendency diagnosis module constructs a learning style tendency diagnosis algorithm that integrates three-dimensional features: "learning style model," "learning content object," and "learning behavior." This algorithm links the "learning style model," "learning content object," and "learning behavior" to form a system that uses the learning content object as an intermediary and calculates the learner's learning style tendency using a series of learning behaviors generated on the learning content object. The learning content object includes learning materials and interactive activities, and the learning behavior refers to a series of learning behaviors generated on the learning content object.

[0168] The learning style preference diagnosis algorithm in the learning style preference diagnosis module is implemented as follows:

[0169] (1) Merge the set of learning materials in the learning material library and the set of interactive activities in the interactive activity library to form a related set of the two, and determine the learning content object;

[0170] (2) Extract the set of all learning behaviors and operation frequencies based on the learning content object to realize the binary association between "learning content object" and "learning behavior" and obtain the association information;

[0171] (3) Establish the association between different dimensions in the learning style model and the learning content object, that is, associate different learning content objects with different dimensions in the learning style model to obtain the association between "learning style model" and "learning content object", superimpose the association information in step (2), and obtain the set of all learning behaviors and operation frequencies of different dimensions of each learning style model in turn; then set the calculation weight of different dimensions of each learning style model according to the learner's individual information, additional information and the types of all learning behaviors on the learning content object;

[0172] (4) Under natural learning conditions, calculate the learning style tendency scores of all learners in different dimensions of the learning style model, and sort the learning style tendency scores of each dimension to obtain the learning style tendency of the learners.

[0173] The specific implementation of the above learning style preference diagnosis algorithm is as follows:

[0174] (1) Merge the learning material set MaterialClassification and the interactive activity set ActivityClassification to form the associated set ContentObjectTypeSet of all learning content objects, which is then determined as the learning content object;

[0175] The learning material set MaterialClassification = {learning material category 1, learning material category 2, learning material category 3, learning material category 4, ..., learning material category n};

[0176] For example, MaterialClassification = {video, audio, animation, image, text, chart, course link}

[0177] Videos, audio, animations, images, text, charts, course links, etc. are all categories of learning materials.

[0178] The set of interactive activities is ActivityClassification = {Activity Class 1, Activity Class 2, Activity Class 3, ..., Activity Class m};

[0179] For example: ActivityClassification={Discussion and Exchange, Debate, Practice and Test, Peer Review, Content Curation, Assignment Submission, SWOT Analysis, Learning Reflection}, where Discussion and Exchange, Debate, Practice and Test, Peer Review, Content Curation, Assignment Submission, SWOT Analysis, and Learning Reflection are the interactive activity categories.

[0180] Related collection ContentObjectTypeSet=

[0181] {ContentObject1,ContentObject2,……,ContentObject i}

[0182] ContentObject i ContentObject represents a specific category of learning materials or interactive activities within the ActivityClassification or MaterialClassification. i ∈{learning material category 1, learning material category 2, ..., learning material category n, interactive activity category 1, interactive activity category 2, ..., interactive activity category m}, representing different types of learning materials;

[0183] For each type of ContentObject in different learning content object categories i It includes the Material set of all learning materials and interactive activities generated by learners. i,j ,Right now:

[0184] ContentObject i ={Materiali,1 Material i,2 Material i,j};

[0185] Among them, Material i,j This indicates that the content object category is ContentObject. i The j-th specific learning material (e.g., ContentObject) in the dataset i If it represents video material, then Material i,j This represents the j-th video in the learning materials of the video category.

[0186] For example: ContentObjectTypeSet = {video, audio, animation, image, text, chart, course link, learning activity {discussion and exchange, debate, practice test, peer review, content curation, assignment submission, SWOT analysis, learning reflection}};

[0187] (2) Based on all learning materials and interactive activities in the associated set ContentObjectTypeSet, retrieve the learner's Material for each learning material or interactive activity from the process behavior library. i,j The set of learning behaviors and their operation frequencies generated on different learning content objects constitutes CB, which is the set of all learning behaviors and operation frequencies of learners on different learning content objects, thereby associating learning content objects with learning behaviors.

[0188] CB = B⊙M={

[0189] B1=COFrequency{R 11 ,R 12 ,…R 1t},

[0190] B2=COFrequency{R 21 ,R 22 ,…R 2t},

[0191] …

[0192] B i =COFrequency{R i1 ,R i2 B ij , …, R it},

[0193] …

[0194] B n =COFrequency{R n1,R n2 ,…R nt}

[0195] }

[0196] Where B is the complete set of all learning behaviors of the learner, and ⊙M represents the learner's learning behavior in M, i.e., Material. i,j Operational frequency of all learning behaviors; B n R represents the set of learning content objects and operation frequencies in all nth type of learning behavior during the learner's learning process, and is a subset of the universal set of learning behaviors B for a specific learning behavior type; here, R is used. ij This represents the behavior corresponding to any learning content object in CB (such as R). 11 ,R 12 …), which means the set of learning content objects and operation frequencies of the j-th type in the i-th type of learning behavior, where i∈{1,2,……,n}, j∈{1,2,……,t};

[0197] For example, B i Represents browsing behavior, R ij R represents the frequency with which the learner views the j-th learning content object (such as a photosynthesis video). For example, if the learner views the video 10 times, then R... ij It is 10.

[0198] (3) Based on the CB obtained in step (2), establish the association between different dimensions in the learning style model and the learning content object, that is, map different types of content in the learning content object to different dimensions in the learning style model. Using the different dimensions of each learning style model as a distinction, obtain the learning content object and operation frequency that can reflect the different dimensions in each learning style model, and form a subset CB of the different dimensions StyleK of the learning style model. styleK :

[0199] CB styleK = {

[0200] BS1 = COFrequency styleK {RSK 11 RSK 12 ,…RSK 1t},

[0201] BS2 = COFrequency styleK {RSK 21 RSK 22 ,…RSK 2t},

[0202] …

[0203] BSi = COFrequency styleK {RSK i1 RSK i2 ,…RSK ij ,…RSK it},

[0204] …

[0205] BS n = COFrequency styleK {RSK n1 RSK n2 ,…RSK nt},

[0206] };

[0207] Among them: RSK i1 ~ RSK it The operation frequency of the learning content object in the i-th type of learning behavior can reflect the different dimensions of StyleK of the current learning style model, where RSK is not a dimension of the current learning style model. ij Assign 0 directly, where i∈{1,2,……,n}, j∈{1,2,……,t}, RSK ij For R ij A subset of behavioral frequencies along the style dimension of the styleK learning process;

[0208] If the current learning style is calculated as visual, and the resource the learner browses is audio, since this browsing behavior cannot be added to the calculation of the visual style score, then the corresponding RSK for that audio will be... ij It can be directly assigned a value of 0, but when calculating based on verbal style, this value is the frequency value read from the database.

[0209] (4) Based on the above CB, and combining individual information and additional information (the two are combined into PersonalInfo), calculate the weights W for different types of learning behaviors. Based on W, obtain the learner's final learning style preference coefficient StyleScore; where the maximum value of each dimension in the learning style model is the learner's learning style preference coefficient in the current dimension:

[0210] StyleScore =Max(W CB)

[0211] in:

[0212] CB = B⊙M

[0213] Where: B =

[0214] M =

[0215] W = [w1, w2, ..., wi, ..., wn]

[0216] Therefore, StyleScore =

[0217] Let wi be the i-th element in W, wi = {Behavior, BehaviorType, Scale, Note}, where i ranges from 1 to n, Behavior represents the name of a specific learning behavior, then its corresponding Bij (i, j are used as references and are the values ​​in the i-th row and j-th column of the vector space) represents a specific learning behavior, Mi represents the object type corresponding to the behavior, BehaviorType represents different types of learning behaviors, Scale represents the weight of different types of learning behaviors, and Note represents the learning style reflected by the learning behavior, i.e., the indicators of different dimensions of the learning style model;

[0218] The scales corresponding to the different types of learning behaviors constitute a set of weights called ScaleSet:

[0219] ScaleSet = {

[0220] PersonalInfo{Scale 11 ,Scale 12 ,……,Scale 1m},

[0221] CreateBehavior{Scale 21 ,Scale 22 ,……,Scale 2n},

[0222] InteractiveBehavior{Scale 31 ,Scale 32 ,……,Scale 3s},

[0223] ReceiveBehavior{Scale 31 ,Scale 32 ,……,Scale 3t}

[0224] }

[0225] Where: PersonalInfo{ Scale 11 ,Scale12 ,……,Scale 1m} represents the weights used in the calculation process of superimposing different types of individual information and additional information. CreateBehavior{ Scale 21 ,Scale 22 ,……,Scale 2n} represents the weights corresponding to different creative learning behaviors during the calculation process, InteractiveBehavior{ Scale 31 ,Scale 32 ,……,Scale 3s} represents the weights corresponding to different interactive learning behaviors during the calculation process, ReceiveBehavior{ Scale 31 ,Scale 32 ,……,Scale 3t} represents the weights corresponding to different receptive learning behaviors in the calculation process; the creative behavior refers to the learning behavior type at the level of creation (such as creating resources), the interactive behavior refers to the learning behavior type at the level of interaction (such as commenting), and the receptive behavior refers to the learning behavior type at the level of reception (such as browsing).

[0226] Table 2 describes the weights corresponding to the above behaviors and how they reflect learning styles. For example, the behavior of "creating resources" is a creative behavior, and its weight is Scale. 22 If the learner creates a video, it reflects a visual learning style.

[0227] Table 2 Learning Behavior Empowerment Constraints Table

[0228]

[0229] Based on classifications within the field of learning science, a deep learning behavior model (DSM) is constructed, encompassing three levels of interactive learning behavior and reflecting different depths of learner learning. This model reflects the learner's varying levels of learning depth.

[0230] DSM = {

[0231] Receptive behaviors: {browsing, searching}

[0232] Interactive behaviors: {Request collaboration, join learning, edit, comment, annotate, download, favorite, follow, share and recommend, watch video, drag progress, skip current content}

[0233] Creative behaviors: {Creating knowledge points, creating resources, creating activities, creating evaluation schemes, posting, and approving}

[0234] }

[0235] To support learning behavior statistics, BehaviorClassification is constructed to integrate all types of learning behaviors corresponding to the above learning behavior models, supporting the acquisition of all learning behaviors B of the learner, i.e., the complete set of learning behaviors:

[0236] BehaviorClassification={Browse, Search, Request Collaboration, Join Learning, Edit, Comment, Annotate, Download, Favorite, Follow, Share & Recommend, Watch Video, Drag Progress, Skip Current Content, Create Knowledge Point, Create Resource, Create Activity, Create Evaluation Scheme, Publish Post, Review, {Extended Behaviors}}.

[0237] Finally, the values ​​of different dimensions in the StyleScore are sorted in ascending order, and the highest value represents the learner's learning style tendency (Tend). For example, if the calculated values ​​of the eight sub-dimensions of the learner's learning process are StyleScore = {Visual: 0.85, Verbal: 0.2, Intuitive: 0.22, Perceiving: 0.72, Active: 0.66, Reflective: 0.33, Global: 0.82, Sequential: 0.1}, then the learner's learning style is Visual, Global, Perceiving, and Active, with Visual being the highest tendency. Figure 5 shows the learning style types corresponding to different learners based on the above calculations.

[0238] To support the dynamic evolution of learning style preference diagnosis, the learning style preference diagnosis module also incorporates the learning behavior occurrence time parameter as a dynamic learning style preference diagnosis method. The dynamic learning style preference diagnosis method constructs a time-based learning behavior weighting function, which calculates the weighted learning behavior at different time stages when collecting learning behavior, thereby strengthening the most recently occurring learning behavior and weakening the learning behavior of a longer period of time, and dynamically fine-tuning the learning style preference diagnosis over time, ultimately obtaining the learning style preference.

[0239] The implementation method of the time-based learning behavior weighting function is as follows: First, the set of learning behaviors is divided into k segments according to the time sequence, and the k segments are denoted as p1, p2, ..., p3 according to the time sequence. k The corresponding time nodes ts for the k segments are 1 / k, 2 / k, ..., k-1 / k, 1. Then, the logistic function is used to map the weights of the learning behavior.

[0240] W(ts) =

[0241] W(ts) represents the weight of the learning behavior within the time period corresponding to the ts-th node, where the logistic function is the weighted parameter that adjusts the behavior at the current time node by using time ts as a variable.

[0242] Ultimately, this will enable the calculation of learners' learning style preferences that evolve dynamically over time.

[0243] Adaptive Service Module: Based on learning style preferences, it overlays individual information and supplementary information stored in the individual information database and supplementary information database to form a learner feature vector. The feature vector determines the type of learning materials and learning mode that the learner is suited for. Based on the vector similarity calculation method, it selects the learning materials and interactive activities with the highest matching degree with the learner from the learning material database and interactive activity database of the data storage module for recommendation and output.

[0244] The recommendation process is based on vector similarity, that is, constructing representation vectors MR of learning materials and interactive activities according to the learning style tags suitable for learning materials and activities, obtaining the similarity between the representation vector MR and Tend, and the learning materials and interactive activities with the highest similarity are the ones to be recommended.

[0245] like Figure 6 As shown, the specific implementation of the adaptive service module of this invention is as follows:

[0246] (1) Input the learner's calculated learning style tendency Tend;

[0247] (2) Extract information such as individual preferences and scenarios from the learner individual information database and the supplementary information database, and use it to select the most suitable learning mode in combination with the learner's learning style tendency (the learning mode here is the way the learner carries out learning, which is a service after the learning style is evaluated).

[0248] (3) The learning modes described in step (2) include inquiry-based learning, self-directed learning, and collaborative learning. Their organizational forms are divided into global type (presenting the overall logic first, and then presenting the details of the content) and sequential type (presenting the details of each part of the content first, and then presenting the overall logic). The learning mode suitable for the learner is determined according to the learner's learning style, scenario and individual preferences.

[0249] (4) Extract learning materials and interactive activities suitable for learners based on their learning styles and preferences;

[0250] (5) Associate the learning materials and interactive activities with the learning mode steps (the learning mode steps are predefined by experts, and the association method is not within the scope of this invention) to obtain adaptive learning services and output them to learners.

[0251] User Interface Module: This module encapsulates the functions of the learning style diagnosis module and the adaptive service module, providing learners with an online learning interface through smart terminals. In supporting learners' learning process, the user interface module provides learners with adaptive learning materials and interactive activities, and also ensures data collection for the data acquisition module based on the sensing and recording devices equipped on the terminal.

[0252] In summary, this invention is applicable to the online learning process by constructing a three-dimensional feature-supported learning style tendency diagnostic model of "learning style model-content object-learning behavior model" and embedding it into the online learning system. It collects learners' interactive behaviors in real time, determines their preferences for learning materials and content through the diagnosis of learners' learning styles, realizes real-time diagnosis of learners' learning style tendencies, and then provides them with learning materials and interactive activities that match their style, thereby promoting the improvement of learning adaptability and efficiency.

[0253] The parts of this invention not described in detail are well known in the field.

[0254] Through the above description of the embodiments, those skilled in the art can clearly understand that the above embodiments can be implemented by software, or by using software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solutions of the above embodiments can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, mobile hard drive, etc.), including several instructions to cause a computer device (such as a personal computer, server, or network device, etc.) to execute the system or method described in the various embodiments of the present invention.

[0255] Although the illustrative specific embodiments of the present invention have been described above to enable those skilled in the art to understand the invention, it should be understood that the invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes will be obvious as long as they are within the spirit and scope of the invention as defined and determined by the appended claims, and all inventions utilizing the concept of the present invention are protected.

Claims

1. A learning style preference adaptive diagnostic system based on online behavior, characterized in that, include: The module includes a data storage module, a data acquisition module, a learning style preference diagnosis module, an adaptive service module, and a user interface module. Data storage module: Establish six data sub-databases, namely, learning material database, interactive activity database, process behavior database, learning style model database, individual information database, and supplementary information database; The learning resource library stores different types of learning materials and collections used by learners during their online learning process; Interactive activity library, storing different types of interactive activities and sets of them during learners' online learning process; The process behavior library stores the set of all learning behaviors and operation frequencies generated by learners during online learning through interaction with learning materials and activities. A learning style model library that stores well-known learning style models in the field of education and their different dimensions; An individual information database stores learners' individual information, including basic information, device information, and preference information; The supplementary information database stores relevant supplementary information, including well-known typical learning scenarios and learning modes in the current online learning field. The learning mode refers to the organization of different types of learning materials in the learning process. In addition, the supplementary information database also provides an extensible interface for supplementing relevant supplementary information. Data acquisition module: Used to support the standardized acquisition of learner behaviors in well-known typical learning scenarios in the current online learning field. During the acquisition process, the acquired data needs to be preprocessed and formatted to facilitate data classification and storage; the well-known typical learning scenarios include course learning, online self-study, online collaborative learning and blended learning; Learning Style Tendency Diagnosis Module: This module constructs a learning style tendency diagnosis algorithm that integrates three-dimensional features: "learning style model," "learning content object," and "learning behavior." The algorithm links these three elements, using the learning content object as an intermediary, and calculates the learner's learning style tendency based on a series of learning behaviors generated on that learning content object. The learning content object includes learning materials and interactive activities, and the learning behavior refers to a series of learning actions generated on the learning content object. Adaptive Service Module: Based on learning style preferences, it overlays individual information and supplementary information stored in the individual information database and supplementary information database to form a learner feature vector. The feature vector determines the type of learning materials and learning mode that the learner is suited for. Based on the vector similarity calculation method, it selects the learning materials and interactive activities with the highest matching degree with the learner from the learning material database and interactive activity database of the data storage module for recommendation and output. User Interface Module: This module encapsulates the functions of the learning style diagnosis module and the adaptive service module, providing learners with an online learning interface through smart terminals. In supporting learners' learning process, the user interface module provides learners with adaptive learning materials and interactive activities on the one hand, and ensures data collection by the data acquisition module based on the sensing and recording devices equipped on the terminal on the other hand.

2. The learning style tendency adaptive diagnostic system based on online behavior according to claim 1, characterized in that: To support the dynamic evolution of learning style preference diagnosis, the learning style preference diagnosis module also incorporates the learning behavior occurrence time parameter as a dynamic learning style preference diagnosis method. The dynamic learning style preference diagnosis method constructs a time-based learning behavior weighting function, which calculates the weight of learning behaviors at different time stages when collecting learning behaviors, strengthens the most recently occurring learning behaviors, weakens the learning behaviors that occurred a long time ago, and dynamically fine-tunes the learning style preference diagnosis over time, ultimately obtaining the learning style preference.

3. The learning style tendency adaptive diagnostic system based on online behavior according to claim 1 or 2, characterized in that: The learning style preference diagnosis process in the learning style preference diagnosis module is implemented as follows: (1) Merge the set of learning materials in the learning material library and the set of interactive activities in the interactive activity library to form a related set of the two, and determine the learning content object; (2) Extract the set of all learning behaviors and operation frequencies based on the learning content object to realize the binary association between "learning content object" and "learning behavior" and obtain the association information; (3) Establish the association between different dimensions in the learning style model and the learning content object, that is, associate different learning content objects with different dimensions in the learning style model to obtain the association between "learning style model" and "learning content object", and superimpose the association information in step (2) to obtain the set of all learning behaviors and operation frequencies of different dimensions of each learning style model in turn; then set the calculation weight of different dimensions of each learning style model according to the learner's individual information, additional information and the types of all learning behaviors on the learning content object. (4) Calculate the learning style tendency scores of all learners in different dimensions of the learning style model, and sort the learning style tendency scores of each dimension to obtain the learning style tendency of the learners.

4. The learning style tendency adaptive diagnostic system based on online behavior according to claim 1 or 2, characterized in that: The learning style preference diagnosis algorithm is specifically implemented as follows: (1) Merge the learning material set MaterialClassification and the interactive activity set ActivityClassification to form the associated set ContentObjectTypeSet of all learning content objects, which is then determined as the learning content object; The learning material set MaterialClassification = {learning material category 1, learning material category 2, learning material category 3, learning material category 4, ..., learning material category n}; The set of interactive activities is ActivityClassification = {Activity Class 1, Activity Class 2, Activity Class 3, ..., Activity Class m}; Related collection ContentObjectTypeSet= {ContentObject1,ContentObject2,……,ContentObject i } ContentObject i ContentObject represents a specific category of learning materials or interactive activities within the ActivityClassification or MaterialClassification. i ∈{learning material category 1, learning material category 2, ..., learning material category n, interactive activity category 1, interactive activity category 2, ..., interactive activity category m}, representing different learning content object types; For each type of ContentObject in different learning content object categories i It includes the Material set of all learning materials and interactive activities generated by learners. i,j ,Right now: ContentObject i ={Material i,1 , Material i,2 ,……, Material i,j }; Material i,j This indicates that the content object category is ContentObject. i The j-th specific learning material or activity in the process; (2) Based on all learning materials and interactive activities in the ContentObjectTypeSet associated set of learning content objects, retrieve the Material of each learning material or interactive activity from the process behavior library. i,j The set of learning behaviors and their operation frequencies generated on different learning content objects constitutes CB, which is the set of all learning behaviors and operation frequencies of learners on different learning content objects, thereby associating learning content objects with learning behaviors. CB = B⊙M={ B1=COFrequency{R 11 ,R 12 ,…R 1t }, B2=COFrequency{R 21 ,R 22 ,…R 2t }, … B i =COFrequency{R i1 ,R i2 ,…, B ij , …, R it }, … B n =COFrequency{R n1 ,R n2 ,…R nt } } Where B is the complete set of all learning behaviors generated by the learner, and ⊙M represents the learner's behavior in M, i.e., Material. i,j Operational frequency of all learning behaviors; B n R represents the set of learning content objects and operation frequencies in all nth type of learning behavior during the learner's learning process, and is a subset of the universal set of learning behaviors B for a specific learning behavior type; here, R is used. ij Let represent the learning behavior corresponding to any learning content object in CB. Its meaning is the set of learning content objects of the j-th type and operation frequencies in the i-th type of learning behavior, where i∈{1,2,……,n} and j∈{1,2,……,t}; (3) Based on the CB obtained in step (2), establish the association between different dimensions in the learning style model and the learning content object, that is, map different types of content in the learning content object to different dimensions in the learning style model. Using the different dimensions of each learning style model as a distinction, obtain the learning content object and operation frequency that can reflect the different dimensions in each learning style model, and form a subset CB of the different dimensions StyleK of the learning style model. styleK : CB styleK = { BS1 = COFrequency styleK {RSK 11 , RSK 12 ,…RSK 1t }, BS2 = COFrequency styleK {RSK 21 , RSK 22 ,…RSK 2t }, … BS i = COFrequency styleK {RSK i1 , RSK i2 ,…RSK ij ,…RSK it }, … BS n = COFrequency styleK {RSK n1 , RSK n2 ,…RSK nt }, }； Among them: RSK i1 ~ RSK it The operation frequency of the learning content object in the i-th type of learning behavior can reflect the different dimensions of StyleK of the current learning style model, where RSK is not a dimension of the current learning style model. ij Assign 0 directly, where i∈{1,2,……,n}, j∈{1,2,……,t}, RSK ij For R ij In StyleK, a subset of learning behavior frequencies along the style dimension; (4) Based on CB, and combining individual information and additional information, set the calculation weights wi for different types of learning behaviors. Based on wi, obtain the learner's final learning style tendency coefficient StyleScore; where the maximum value of each dimension in the learning style model is the learner's learning style tendency in the current dimension: StyleScore =Max(W* CB styleK ) in: CB = B ⊙ M Where: B = M represents the type of material that reflects StyleK, M = W = [w1, w2, ..., wi, ..., wn] Therefore, StyleScore = Let wi be the i-th element in W, wi = {Behavior, BehaviorType, Scale, Note}.Scale, where i ranges from 1 to n, Behavior represents the name of a specific learning behavior, and its corresponding Bij represents a specific learning behavior that occurred. i and j are used as references and are the values ​​in the i-th row and j-th column of the vector space. The learning object type corresponding to the learning behavior is represented by BehaviorType, different types of learning behavior are represented by Scale, different weights of different types of learning behavior are represented by Note, and different dimensions of the learning style reflected by the learning behavior are represented by the indicators of the learning style model. The scales corresponding to the different types of learning behaviors constitute the complete set of weights, ScaleSet (wi is the set of weights corresponding to the behaviors generated during the calculation process): ScaleSet = { PersonalInfo{Scale 11 ,Scale 12 ,……,Scale 1m }, CreateBehavior{Scale 21 ,Scale 22 ,……,Scale 2n }, InteractiveBehavior{Scale 31 ,Scale 32 ,……,Scale 3s }, ReceiveBehavior{Scale 31 ,Scale 32 ,……,Scale 3t } } Where: PersonalInfo{ Scale 11 ,Scale 12 ,……,Scale 1m } represents the weights used in the calculation process of superimposing different types of individual information and additional information. CreateBehavior{ Scale 21 ,Scale 22 ,……,Scale 2n } represents the weights corresponding to different creative learning behaviors during the calculation process, InteractiveBehavior{ Scale 31 ,Scale 32 ,……,Scale 3s } represents the weights corresponding to different interactive learning behaviors during the calculation process, ReceiveBehavior{ Scale 31 ,Scale 32 ,……,Scale 3t } represents the weights corresponding to different receptive learning behaviors in the calculation process; the creative learning behaviors refer to the learning behaviors at the creative level, the interactive learning behaviors refer to the learning behaviors at the interactive level, and the receptive learning behaviors refer to the learning behaviors at the receiving level; the behaviors corresponding to the above weights include BehaviorClassification={Browse, Search, Apply for Collaboration, Join Learning, Edit, Comment, Annotate, Download, Favorite, Follow, Share Recommendation, Watch Video, Drag Progress, Skip Current Content, Create Knowledge Point, Create Resource, Create Activity, Create Evaluation Scheme, Publish Post, Review, {Extended Behavior}}; Finally, the values ​​of different dimensions in the StyleScore are sorted in order of magnitude, and the largest value is the learner's learning style tendency (Tend).

5. The learning style tendency adaptive diagnostic system based on online behavior according to claim 1 or 2, characterized in that: The learning behavior classification is based on a classification system derived from the field of learning science. It constructs a deep learning behavior model (DSM) encompassing three levels of learning behavior and reflecting different depths of learner learning. This model is known as the Double Helix Deep Learning Behavior Model. The learning behavior classifications in the DSM correspond to the three behavior weights (CreateBehavior, InteractiveBehavior, ReceiveBehavior) in the ScaleSet. DSM = { Receptive learning behaviors: {browsing, searching} Interactive learning behaviors include: {Requesting collaboration, joining a learning session, editing, commenting, annotating, downloading, saving, following, sharing and recommending, watching videos, dragging the progress bar, and skipping current content}. Creative learning behaviors: {Creating knowledge points, creating resources, creating activities, creating assessment schemes, posting, and approving} }。 6. The learning style tendency adaptive diagnostic system based on online behavior according to claim 1 or 2, characterized in that: The well-known learning style models in the field of education include: the Felder-Silverman model, the Kolb model, and the VARK model; The Felder-Silverman learning style model is represented by different dimensions as follows: FS{Visual, Verbal, Intuitive, Perceptive, Active, Reflective, Global, Sequential}; The Kolb learning style models of different dimensions are represented as follows: Kolb {assimilation, accommodation, divergence, centralization}; The different dimensions of the VARK learning style model are: VARK {visual, auditory, reading and writing, experiential}.

7. The learning style tendency adaptive diagnostic system based on online behavior according to claim 1 or 2, characterized in that: The data acquisition module collects data based on the built-in sensors of the learning device and the log data of the learning platform. During the acquisition process, the collected data is preprocessed and formatted according to the data from different sources. The standardized acquisition of the data is based on the built-in sensors of mobile phones, tablets and PCs, the log data of the online learning platform, and the learner's interactive activities.

8. The learning style tendency adaptive diagnostic system based on online behavior according to claim 2, characterized in that: The implementation method of the time-based learning behavior weighting function is as follows: First, the set of learning behaviors is divided into k segments according to the time sequence, and these segments are denoted as p1, p2, ..., p1 according to the order in which the events occur. k The time nodes ts corresponding to the k segments are 1 / k, 2 / k, ..., k-1 / k, 1. Next, the logistic function is used to map the weights of the learning behavior. By adjusting the weighting parameters of the behavior at the current time node using time ts as a variable, the weights of the learning behavior within the time segment corresponding to the ts-th node are obtained. W(ts)= W(ts) represents the calculated weight of the learning behavior within the time period corresponding to the ts-th node; Ultimately, this will enable the calculation of learners' learning style preferences that evolve dynamically over time.

9. The learning style tendency adaptive diagnostic system based on online behavior according to claim 1, characterized in that: The recommendation process in the adaptive service module is based on vector similarity. That is, a representation vector MR is constructed for the learning materials and interactive activities according to the learning style tags suitable for the learning materials and interactive activities. The similarity between the representation vector MR and Tend is obtained. The learning materials and interactive activities with the highest similarity are the ones to be recommended.

10. A method for implementing an online behavior-based adaptive diagnostic system for learning style preferences, as described in any one of claims 1-9, characterized in that: (1) Construct a learning style tendency diagnosis algorithm that integrates three-dimensional features of "learning style model", "learning content object" and "learning behavior". The algorithm associates the "learning style model", "learning content object" and "learning behavior" to form a learning style tendency of learners using a series of learning behaviors generated on the learning content object as the medium. The learning style model is stored in the learning style model library of the data storage module; The learning content objects are stored in the learning material library and interactive activity library of the data storage module; the learning behaviors are stored in the process behavior library of the data storage module and are collected through the data acquisition module. (2) Real-time diagnosis of learners' learning style preferences when learners engage in online learning: (21) First, extract all the learning materials and interactive activities that learners interact with during the online learning process, and merge them to obtain the associated set of all learning content objects, ContentObjectTypeSet, which is the learning content object; The learning material set MaterialClassification = {learning material category 1, learning material category 2, learning material category 3, learning material category 4, ..., learning material category n} The set of interactive activities is ActivityClassification = {Activity Class 1, Activity Class 2, Activity Class 3, ..., Activity Class m}; Related collection ContentObjectTypeSet= {ContentObject1,ContentObject2,……,ContentObject i } ContentObject i ContentObject represents a specific category of learning materials or interactive activities within the ActivityClassification or MaterialClassification. i ∈{learning material category 1, learning material category 2, ..., learning material category n, interactive activity category 1, interactive activity category 2, ..., interactive activity category m}, representing different learning content object types; For each type of ContentObject i It includes the Material set of all learning materials and interactive activities generated by learners. i,j ,Right now: ContentObject i ={Material i,1 , Material i,2 ,……, Material i,j }; Among them, Material i,j This indicates that the content object type is ContentObject. i The j-th specific learning material in the series; (22) Based on all learning materials and interactive activities in the associated set ContentObjectTypeSet, retrieve the different Material types for each learning material or interactive activity from the process behavior library. i,j The set of learning behaviors and operation frequencies generated above constitutes CB, which associates the learning content object with the learning behavior. CB = B⊙M={ B1=COFrequency{R 11 ,R 12 ,…R 1t }, B2=COFrequency{R 21 ,R 22 ,…R 2t }, … B i =COFrequency{R i1 ,R i2 ,…, B ij , …, R it }, … B n =COFrequency{R n1 ,R n2 ,…R nt } } Where B is the complete set of all learning behaviors of the learner, and ⊙M represents the learner's learning behavior in M, i.e., Material. i,j The frequency of all learning behaviors; R ij Let B represent the set of learning content objects and operation frequencies of the j-th type in the i-th type of learning behavior, where i∈{1,2,……,n}, j∈{1,2,……,t}; n This represents the set of learning content objects and operation frequencies in all nth type of learning behavior during the learner's learning process; (23) Based on CB, establish the association between different dimensions in the learning style model and the learning content objects, that is, map different types of content in the learning content objects to different dimensions in the learning style model. Using the different dimensions of each learning style model as the distinction, obtain the learning content objects and operation frequency that can reflect each different dimension in each learning style model, forming a subset of CB in the different dimensions of StyleK of the learning style model. styleK : CB styleK = { BS1 = COFrequency styleK {RSK 11 , RSK 12 ,…RSK 1t }, BS2 = COFrequency styleK {RSK 21 , RSK 22 ,…RSK 2t }, … BS i = COFrequency styleK {RSK i1 , RSK i2 ,…RSK ij ,…RSK it }, … BS n = COFrequency styleK {RSK n1 , RSK n2 ,…RSK nt }, }； Among them: RSK i1 ~ RSK it The operation frequency of the learning content object in the i-th type of learning behavior can reflect the different dimensions of StyleK of the current learning style model, where RSK is not a dimension of the current learning style model. ij Assign 0 directly, where i∈{1,2,……,n}, j∈{1,2,……,t}, RSK ij For R ij A subset of learning behavior frequencies along the styleK learning style dimension; (24) Establish the association between different dimensions of the learning style model and the learning content object, that is, associate different learning content objects with different dimensions of the learning style model to obtain the association between "learning style model" and "learning content object", and superimpose the binary association between "learning content object" and "learning behavior" to obtain all learning behaviors and their frequency sets for different dimensions of each learning style model. Then, combine individual information and additional information to set the calculation weight W for different types of learning behaviors, W={Behavior, BehaviorType, Scale, Note}.Scale, where Behavior represents the specific learning behavior name, BehaviorType represents different types of learning behavior, Scale represents the weight of different types of learning behavior, and Note represents the learning style index reflected by the learning behavior; the calculation weight wi is taken from the complete set ScaleSet of all behavior weights defined by the system, as follows: ScaleSet = { PersonalInfo{Scale 11 ,Scale 12 ,……,Scale 1m }, CreateBehavior{Scale 21 ,Scale 22 ,……,Scale 2n }, InteractiveBehavior{Scale 31 ,Scale 32 ,……,Scale 3s }, ReceiveBehavior{Scale 31 ,Scale 32 ,……,Scale 3t } } Where: PersonalInfo{ Scale 11 ,Scale 12 ,……,Scale 1m } represents the weights used in the calculation process of superimposing different types of individual information and additional information. CreateBehavior{ Scale 21 ,Scale 22 ,……,Scale 2n } represents the weights corresponding to different creative learning behaviors during the calculation process, InteractiveBehavior{ Scale 31 ,Scale 32 ,……,Scale 3s } represents the weights corresponding to different interactive learning behaviors during the calculation process, ReceiveBehavior{ Scale 31 ,Scale 32 ,……,Scale 3t } represents the weights corresponding to different receptive behaviors in the calculation process; the creative learning behavior refers to the learning behavior type at the creative level in the learning behavior, the interactive learning behavior refers to the learning behavior type at the interactive level in the learning behavior, and the receptive learning behavior refers to the learning behavior type at the receiving level in the learning behavior. (25) Finally, the weights W of all learner behaviors, learner behaviors, and learning content objects are correlated and calculated to obtain the learner's final learning style preference coefficient StyleScore; StyleScore =Max(W CB) CB = B ⊙ M B = M = W = [w1, w2, ..., wi, ..., wn] ： Let wi be the i-th element in W, and wi = {Behavior, BehaviorType, Scale, Note}, where i ranges from 1 to n; (26) The maximum value of each dimension in the learning style model represents the learner's learning style preference in that dimension. In StyleScore, the values ​​of different dimensions are sorted in order of magnitude, and the largest value represents the learner's learning style preference (Tend).

Images