CN-122022094-A - Self-organizing learning path optimization method and system based on learning behavior mining

CN122022094ACN 122022094 ACN122022094 ACN 122022094ACN-122022094-A

Abstract

The invention discloses a self-organizing learning path optimization method and a self-organizing learning path optimization system based on learning behavior mining, which relate to the field of path optimization, and comprise the following steps of S1, collecting data and building a library; the method comprises the steps of S2, presetting a judging standard, dividing and eliminating redundancy on data, integrating the data to form a data set, S3, preprocessing the data set, adopting an algorithm to refine core characteristics in a differentiation mode, building a double-dynamic model by combining a library, S4, setting a path standard, optimizing to obtain a multi-version pushing standard and path adaptation range, S5, calling a path candidate set to generate a personalized self-organizing learning path candidate scheme, S6, implementing the candidate scheme, integrating acquired data in real time to form a double report, and S7, presetting a threshold value to judge the double report. According to the invention, through full-period multidimensional learning data acquisition and hierarchical cognitive archive construction, the target refinement of learning adaptation rhythm and scene adaptation habit is realized.

Inventors

JIANG ZHICHAO
YE BO
GAO ZHIHONG

Assignees

武汉工学智联科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260121

Claims (10)

1. The self-organizing learning path optimization method based on learning behavior mining is characterized by comprising the following steps of: s1, acquiring learning behavior track data, knowledge point mastering data and learning scene adaptation data of a learner in a whole period, and constructing a hierarchical learner cognitive archive and a domain knowledge point feature library; S2, presetting a learner cognition level threshold, a learning path adaptation standard and a path conflict judgment standard, marking and classifying learning behavior track data, knowledge point mastering data and learning scene adaptation data, classifying scene adaptation, eliminating interference redundant data, and integrating to form a multidimensional multi-group learner knowledge point association data set; S3, preprocessing a multidimensional multi-group learner knowledge point association data set, adopting a behavior feature clustering algorithm and a time sequence rule mining algorithm according to data set differentiation, refining learning adaptation rhythm, core cognition short plates and scene adaptation habit, and constructing a plurality of groups of data weight dynamic allocation models and learner cognition knowledge point adaptation dynamic models by combining a domain knowledge point feature library; s4, setting a global path planning and local engagement criterion based on a plurality of groups of data weight dynamic allocation models, introducing a plurality of groups of data collaborative multi-objective optimization algorithm, defining a cognitive bearing threshold value and a progress interval, and then carrying out multi-round path node matching optimization to obtain a multi-version knowledge point pushing reference and a path dynamic adaptation range; s5, a path candidate configuration set is called from a hierarchical learner cognitive archive according to the multi-version knowledge point pushing standard and the path dynamic adaptation range, the path candidate configuration set is brought into a learner cognitive knowledge point adaptation dynamic model, and conflict verification and adaptation adjustment are carried out on the path candidate configuration set by combining a learning path adaptation standard and a path conflict judgment standard, so that a personalized self-organizing learning path candidate scheme is generated; S6, implementing a personalized self-organizing learning path candidate scheme, collecting path execution, cognitive adaptation and target achievement data in real time, synchronously recording cognitive fatigue, knowledge point solidification attenuation and adaptation degree drift data used for a long time by the path, and integrating to form a path success evaluation report and a long-period use influence report; S7, presetting a success threshold and a long-period use safety threshold, judging whether the path success evaluation report and the long-period use influence report reach standards, if the path success evaluation report and the long-period use influence report do not reach standards, extracting deviation factors and optimizing a plurality of groups of data weight dynamic allocation models, returning to the step S3 to execute the process of S6 again to generate paths, and if the paths reach standards, archiving the current paths and the long-period use parameters, and synchronously updating a hierarchical learner cognitive archive and a domain knowledge point feature library.
2. The self-organizing learning path optimization method based on learning behavior mining according to claim 1, wherein the preprocessing of the multidimensional multi-group learner knowledge point association data set adopts a behavior feature clustering algorithm and a time sequence rule mining algorithm according to data set differentiation, extracts learning adaptation rhythm, core cognition short board and scene adaptation habit, and builds a plurality of groups of data weight dynamic allocation models and learner cognition knowledge point adaptation dynamic models by combining a domain knowledge point feature library, and the method comprises the following steps: s31, preprocessing a multidimensional multi-group learner knowledge point associated data set, removing abnormal data and redundant data, respectively performing standardization and code conversion on different types of data, and splitting according to learner levels and learning stages to finish dimensional alignment to obtain a standardized data set; S32, respectively extracting learning adaptation rhythm, core cognitive shortboards and scene adaptation habits as three types of core characteristics by adopting a time sequence rule mining algorithm and a behavior characteristic clustering algorithm according to characteristic differences of a data set in a standardized data set; s33, combining the three types of core features with a domain knowledge point feature library, and constructing a plurality of groups of data weight dynamic allocation models by taking the data credibility, the cognitive target relativity and the scene priority as indexes; S34, combining three types of core features and knowledge point attributes in the domain knowledge point feature library, constructing an adaptive mapping relation between the learner features and the knowledge point attributes, and constructing a learner cognitive knowledge point adaptive dynamic model.
3. The self-organizing learning path optimization method based on learning behavior mining according to claim 1, wherein the steps of making global path planning and local engagement criteria based on a plurality of groups of data weight dynamic allocation models, introducing a plurality of groups of data cooperation multi-objective optimization algorithm, defining a cognitive bearing threshold and a progress interval, and performing multi-path node matching optimization to obtain a multi-version knowledge point pushing reference and a path dynamic adaptation range include the following steps: s41, based on a plurality of groups of data weight dynamic allocation models, determining the connection rule of knowledge point arrangement logic of the global path and the local nodes to form a global path planning and local connection rule; s42, introducing a plurality of groups of data to cooperate with a multi-objective optimizing algorithm, combining a learner cognition level threshold, defining a cognition bearing threshold and a learning progress interval, and determining optimizing constraint conditions; S43, carrying out multi-round path node matching optimization on knowledge points in a domain knowledge point feature library by taking global path planning and local engagement criteria as guidance and combining optimization constraint conditions; s44, generating a multi-version knowledge point pushing reference according to the optimizing result, defining a path dynamic adaptation range, and outputting the multi-version knowledge point pushing reference and the path dynamic adaptation range.
4. The self-organizing learning path optimization method based on learning behavior mining according to claim 1, wherein the step of retrieving a path candidate configuration set from a hierarchical learner cognitive archive according to a multi-version knowledge point pushing reference and a path dynamic adaptation range, bringing the path candidate configuration set into a learner cognitive knowledge point adaptation dynamic model, and performing conflict verification and adaptation adjustment on the path candidate configuration set in combination with a learning path adaptation standard and a path conflict judgment reference, and generating a personalized self-organizing learning path candidate scheme comprises the following steps: S51, according to the multi-version knowledge point pushing reference and the path dynamic adaptation range, the current cognitive level and learning stage information of the learner are called from the hierarchical learner cognitive archive, and the scene adaptation requirements are combined, so that a matched path candidate configuration set is screened; S52, introducing the path candidate configuration set into a learner cognitive knowledge point adaptation dynamic model, and calculating the adaptation degree score of each configuration set and the learner according to the learner characteristics and knowledge point attribute adaptation mapping relation to obtain a preliminary screening adaptation result; s53, according to the learned path adaptation standard and the path conflict judgment standard, carrying out conflict verification on the primary screening adaptation result, and carrying out adaptation adjustment by replacing knowledge points and adjusting arrangement sequence according to path candidate configuration with conflict items and adaptation degree scores which do not reach standards; S54, integrating the adjusted path candidate configuration to generate a personalized self-organizing learning path candidate scheme.
5. The method for optimizing a self-organizing learning path based on learning behavior mining according to claim 1, wherein the implementation of the personalized self-organizing learning path candidate scheme, the real-time acquisition of path execution, cognitive adaptation and target achievement data, and the synchronous recording of cognitive fatigue, knowledge point solidification attenuation and adaptation degree drift data used for a long time of the path, the integration of forming a path success evaluation report and a long-period use influence report comprises the following steps: S61, starting an implementation flow of a personalized self-organizing learning path candidate scheme, and acquiring path execution data such as learning node completion conditions, knowledge point interaction frequency and the like in a path execution process based on a real-time synchronization mechanism of a hierarchical learner cognitive archive; s62, acquiring cognitive adaptation data and target achievement data, presetting a time recording period, and recording cognitive fatigue degree, knowledge point solidification failure rate and adaptation degree drift values in the long-term use process according to the time recording period; S63, carrying out standardized integration on acquired path execution, cognitive adaptation and target achievement data to form a path achievement core data set, and carrying out trend analysis and quantization on cognitive fatigue degree, knowledge point solidification failure reduction rate and adaptation degree drift values to form a long-period influence data set; s64, generating a path success evaluation report based on the path success core data set, and generating a long-period use influence report by combining the long-period influence data set.
6. The self-organizing learning path optimization method based on learning behavior mining according to claim 2, wherein the characteristic difference of the data set in the standardized dataset is extracted by adopting a time sequence rule mining algorithm and a behavior feature clustering algorithm, and learning adaptation rhythm, core cognitive shortboards and scene adaptation habit are respectively extracted as three types of core features, and the method comprises the following steps: S321, dividing the standardized data set into a learning behavior track data set, a knowledge point mastering data set and a scene adaptation data set according to data types; s322, adopting a time sequence rule mining algorithm to the learning behavior track data set, extracting features such as learning duration, interval period, node completion sequence and the like, and extracting learning adaptation rhythm through feature correlation analysis; s323, clustering the knowledge point mastering data set by adopting a behavior feature clustering algorithm according to the knowledge point mastering degree scoring, wrong question type and other features to form different cognitive level clusters, and positioning the core cognitive short board; S324, a behavior feature cluster and time sequence rule mining algorithm is fused with the scene adaptation data set, learning behavior preference, adaptation time length and effect feedback under different scenes are analyzed, scene adaptation habit is refined, and learning adaptation rhythm, core cognitive shortboards and scene adaptation habit are integrated to obtain a core feature set.
7. The self-organizing learning path optimization method based on learning behavior mining according to claim 3, wherein the introducing of the multi-group data collaborative multi-objective optimization algorithm, the combination of the learner cognition level threshold, the definition of the cognition bearing threshold and the learning progress interval, the determination of the optimization constraint condition comprises the following steps: s421, introducing a plurality of groups of data collaborative multi-target optimizing algorithm, initializing algorithm core parameters, and definitely determining the input data dimension of the plurality of groups of data collaborative multi-target optimizing algorithm as the weight parameters output by the dynamic distribution model of the plurality of groups of data weights and the knowledge point attribute data in the domain knowledge point feature library; S422, based on refined learning adaptation rhythm, core cognitive shortboards and scene adaptation habit, combining a learner cognitive level threshold, quantitatively analyzing a cognitive load upper limit through an algorithm, and defining a learner cognitive bearing threshold; S423, combining the cognitive bearing threshold with the difficulty gradient and the association compactness of knowledge points in the learning adaptation rhythm and the domain knowledge point feature library, dividing target ranges of different learning stages, and defining a learning progress interval; s424, integrating the cognitive bearing threshold and the learning progress interval, and determining optimizing constraint conditions of a plurality of groups of data collaborative multi-objective optimizing algorithm.
8. The self-organizing learning path optimization method based on learning behavior mining according to claim 3, wherein the multi-round path node matching optimization for knowledge points in the domain knowledge point feature library by taking global path planning and local engagement criteria as guidance and combining optimization constraint conditions comprises the following steps: S431, loading global path planning, local engagement criteria and optimizing constraint conditions, combining weight parameters of a plurality of groups of data weight dynamic allocation models, and establishing a node matching scoring mechanism by taking knowledge point adaptation degree, relevance and difficulty gradient as core scoring dimensions; S432, first-pass screening is carried out on knowledge points in a domain knowledge point feature library based on a scoring mechanism, knowledge points which accord with a progress interval and a cognition bearing threshold value are selected as initial path nodes according to a global path arrangement logic, and relevance matching among the nodes is completed according to a local connection rule, so that a plurality of groups of initial path node combinations are formed; S433, carrying out multi-round iterative optimization on the initial path node combination, and obtaining a multi-round path node matching optimization result by adjusting the node sequence, replacing equivalent knowledge points and supplementing the transition node optimization adaptation degree.
9. The self-organizing learning path optimization method based on learning behavior mining according to claim 4, wherein the step of bringing the path candidate configuration set into a learner cognitive knowledge point adaptation dynamic model, calculating the adaptation degree score of each configuration set and the learner according to the learner characteristic and knowledge point attribute adaptation mapping relation, and obtaining a preliminary screening adaptation result comprises the following steps: S521, disassembling the path candidate configuration set into knowledge point attributes corresponding to each configuration, and taking the learning adaptation rhythm, the core cognitive short board and the scene adaptation habit as input parameters of a learner cognitive knowledge point adaptation dynamic model; S522, based on the adaptive mapping relation between the learner characteristics and the knowledge point attributes, carrying out adaptive degree quantization scoring on the knowledge point attributes of each configuration set and the core characteristics of the learner through a learner cognitive knowledge point adaptive dynamic model, and setting an adaptive degree threshold screening scoring standard reaching configuration set to obtain a preliminary screening adaptive result.
10. The self-organizing learning path optimization system based on learning behavior mining is used for realizing the self-organizing learning path optimization method based on learning behavior mining, and is characterized by comprising a data acquisition and database building module, a judging and integrating data module, a data model building module, a planning reference range module, a path candidate scheme module, an effect report generating module and a report judging and adjusting module; the data acquisition and library building module is used for acquiring learning behavior track data, knowledge point mastering data and learning scene adaptation data of a learner in the whole period, and building a hierarchical learner cognitive archive and a domain knowledge point feature library; the data integration judging module is used for presetting a learner cognition level threshold, a learning path adaptation standard and a path conflict judging standard, marking, classifying and scene adaptation dividing learning behavior track data, knowledge point mastering data and learning scene adaptation data, eliminating interference redundant data, and integrating to form a multidimensional multi-group learner knowledge point association data set; the data model construction module is used for preprocessing a multidimensional multi-group learner knowledge point association data set, adopting a behavior feature clustering algorithm and a time sequence rule mining algorithm according to data set differentiation, refining learning adaptation rhythm, a core cognition short board and scene adaptation habit, and constructing a plurality of groups of data weight dynamic allocation models and a learner cognition knowledge point adaptation dynamic model by combining a domain knowledge point feature library; The planning reference range module is used for making a global path planning and local engagement criterion based on a plurality of groups of data weight dynamic allocation models, introducing a plurality of groups of data cooperation multi-target optimizing algorithm, defining a cognitive bearing threshold value and a progress interval, and then carrying out multi-round path node matching optimizing to obtain a multi-version knowledge point pushing reference and path dynamic adaptation range; The path candidate scheme module is used for retrieving a path candidate configuration set from the hierarchical learner cognitive archive according to the multi-version knowledge point pushing standard and the path dynamic adaptation range, introducing the path candidate configuration set into the learner cognitive knowledge point adaptation dynamic model, and carrying out conflict verification and adaptation adjustment on the path candidate configuration set by combining the learning path adaptation standard and the path conflict judgment standard to generate a personalized self-organizing learning path candidate scheme; The effect report generation module is used for implementing a personalized self-organizing learning path candidate scheme, collecting path execution, cognitive adaptation and target achievement data in real time, synchronously recording cognitive fatigue, knowledge point solidification attenuation and adaptation degree drift data used for a long time of a path, and integrating a path success evaluation report and a long-period use influence report; The report judging and adjusting module is used for presetting a success qualification threshold and a long period use safety threshold, judging whether the path success evaluation report and the long period use influence report reach standards, extracting deviation factors and optimizing a plurality of groups of data weight dynamic allocation models if the path success evaluation report and the long period use influence report do not reach standards, returning to the step S3 to re-execute the step S6 to generate paths, archiving the current paths and the long period use parameters if the path success evaluation report and the long period use influence report reach standards, and synchronously updating a hierarchical learner cognitive archive and a domain knowledge point feature library.

Description

Self-organizing learning path optimization method and system based on learning behavior mining Technical Field The invention relates to the field of path optimization, in particular to a self-organizing learning path optimization method and system based on learning behavior mining. Background The individualized self-organizing learning path is used as a core supporting module of an intelligent education and online learning system, the adaptive accuracy and the dynamic adjustment capability of the individualized self-organizing learning path directly determine the learning efficiency, the cognitive load balance and the learning target achievement rate of a learner, the individualized self-organizing learning path becomes a key technical grip for promoting education digital transformation and realizing teaching according to the material, the dynamic coupling of multi-dimensional learning data and the cognitive state drift in the long-period learning process are used as core constraint factors for influencing the adaptive performance of the learning path, and the individualized self-organizing path is influenced by the dynamic superposition of multiple factors such as the difference of the learning level of the learner, the fluctuation of learning behavior time sequence, the dynamic change of learning scene, the association complexity of knowledge points and the like, so that the adaptive misalignment of the learning path and the learner demand and the unbalance of the pushing rhythm of the knowledge points are caused, the cognitive fatigue accumulation is caused, the core knowledge short board cannot accurately compensate, the learning enthusiasm is reduced, and the learning target deviation or the learning process is interrupted in serious cases. The self-organizing learning path optimization technology based on learning behavior mining not only solves the key means of insufficient path adaptation precision caused by multi-factor coupling, directly ensures the accurate matching of individual learning demands of learners and learning processes, reduces the cognitive load and improves the learning efficiency by dynamically adjusting path nodes and pushing rhythms, but also promotes the intelligent learning system to support the core of full-period adaptation and high-reliability upgrading, optimizes the path planning strategy by mining learning behavior rules, and simultaneously provides data basis for the cognitive modeling of learners and the construction of a knowledge point system of the next-generation intelligent education system, and avoids the low learning efficiency and the deteriorated learning experience caused by path adaptation misalignment. However, when the existing self-organizing learning path optimization method based on learning behavior mining is used, experience rules are relied on, time sequence rules of learning behavior tracks, coupling correlation values of learning scene adaptation requirements and long-period cognitive state data are ignored, action logics of complex learning processes such as dynamic cognitive change, behavior habit fluctuation and scene switching of learners cannot be covered, so that the path planning has one-sided performance of heavy knowledge pushing light cognitive adaptation, and the existing method mostly adopts a static model or a single algorithm to perform feature extraction and path matching, so that the problem of semantic alignment of multi-source heterogeneous data such as learning behaviors, cognitive states and scene information cannot be solved, and the problem of loose data feature association is easily influenced by interference data is caused. For the problems in the related art, no effective solution has been proposed at present. Disclosure of Invention Aiming at the problems in the related art, the invention provides a self-organizing learning path optimization method and a self-organizing learning path optimization system based on learning behavior mining, so as to overcome the technical problems in the prior art. In order to achieve the above purpose, the specific technical scheme adopted by the invention is as follows: According to one aspect of the invention, a self-organizing learning path optimization method based on learning behavior mining comprises the following steps: s1, acquiring learning behavior track data, knowledge point mastering data and learning scene adaptation data of a learner in a whole period, and constructing a hierarchical learner cognitive archive and a domain knowledge point feature library; S2, presetting a learner cognition level threshold, a learning path adaptation standard and a path conflict judgment standard, marking and classifying learning behavior track data, knowledge point mastering data and learning scene adaptation data, classifying scene adaptation, eliminating interference redundant data, and integrating to form a multidimensional multi-group learner knowledge point association data set; S3, preproce