CN103455576A - Thinking-map-based e-learning resource recommendation method - Google Patents

Abstract

The invention discloses a thinking-map-based e-learning resource recommendation method, which comprises the following steps of analyzing a learning behavior log of a learner accessing a knowledge-map-based e-learning system, calculating a learning time length threshold value, and preprocessing the learning behavior log; constructing a learning path network and a learning transaction graph set; mining a thinking map (a bubble map or a double bubble map) for a specified knowledge element set on the basis of the learning path network; feeding back a mining result to the learner to realize thinking-map-based e-learning knowledge recommendation. According to the method, cognitive-strategy-based map type knowledge recommendation service can be provided for an e-learner, and the learning efficiency of the e-learner is further improved.

Description

A kind of Network Learning Resource recommend method based on the thinking map

Technical field

The invention belongs to Network Learning Resource recommended technology field, relate to a kind of Network Learning Resource recommend method based on the thinking map.

Background technology

Along with the development of infotech, huge variation has also occurred in the form that the mankind seek knowledge.E-Learning can break through the space-time restriction, be convenient to the important means that the characteristics such as resource sharing and support autonomous learning have become people's continuing education and obtained knowledge because of it.Existing e-Learning mode of learning Main Basis behaviorism theory, this theory thinks that study is external knowledge simple " resettlement " process to learner's memory.And cognitive theory is thought, study is that cognitive subject passes through assimilation and the continuous process that forms new cognitive structure of adaptive mechanism, and the internal association of knowledge has vital role in this process.Existing mode of learning lacks the ability of expressing this internal association, and the person that is difficult to effectively not help learning completes Knowledge Construction, easily causes cognitive overload.Therefore, the achievement in research of scholar according to cognitive science arranged, proposed the concept of Knowledge Map.Knowledge Map has been described blocks of knowledge that certain field comprises and the internal association between blocks of knowledge, but does not point out how to learn these blocks of knowledge, learns the mode of thinking of these blocks of knowledge.

The thinking map, as a kind of visual language instrument of supporting mode of thinking training, can promote the concrete study of e-Learning learner to dissimilar knowledge.In the middle of current educational practice, the thinking map is multiplex in the learner is carried out to skill training, even for the learner carries out the knowledge recommendation service based on the thinking map, the thinking map is also mostly is to be drawn by counselor or experienced learner, need larger artificial mark workload, be difficult to promote in intelligent e-Learning system.

The Chinese invention patent that application number is 201210178807, disclose a kind of personalized network education resource recommend method.This patent discloses a kind of personalized network education resource recommend method based on the user interest degree, the person accesses the behavioral data of the learning System of extension-based thematic map to comprise the analytic learning, the concept that acquisition learner and group thereof are relevant to learning content and the learning interest path change pattern of Knowledge Element, then according to the relations such as front and back order between the learning object of the learning interest path change pattern of learner's individuality and place group thereof and expansion thematic map, realize initiatively recommending the personalized recommendation of suitable education resource to the learner.It has following features:

1, provided e-Learning learner's learning interest recognition methods, for the identification learning person sometime the section in learning interest changing model and corresponding interest-degree;

2, extract interest from the content of learner's study, for extracting the interested knowledge point of user.

3, provided the method that e-Learning learner's interest is deleted and merged, for adjusting user's learning interest changing model;

4, provide the method for e-Learning learner's interest prediction, for predicting learner's point of interest in the future, thereby carried out the education resource recommendation according to learner's interest.

Look into new discovery according to above-mentioned, there are following three aspect problems in existing Network Learning Resource recommend method: 1, lack the e-learning recommend method based on cognitive strategy.2, for types such as tactic knowledge, procedural knowledge and declarative knowledges, lack the concrete study mode of thinking.3, the thinking map is as a kind of learning tool of cognitive strategy, and supports mode of thinking training, but lacks a kind of method automatically generated by machine.

Summary of the invention

The object of the present invention is to provide a kind of Network Learning Resource recommend method based on the thinking map, in the middle of the Knowledge Map system, according to learner's request, automatically generate relative illustrated education resource.

To achieve these goals, the present invention adopts following technical scheme:

A kind of Network Learning Resource recommend method based on the thinking map is characterized in that: comprise following operation:

1) gather the learning behavior daily record produced in learner and the learning System reciprocal process based on Knowledge Map, study duration in the analytic learning user behaviors log, obtain study duration threshold value, and with this threshold value to learning behavior daily record filtered, obtain effective daily record;

2) build learning path network and the set of study transaction graph according to the daily record data after filtering;

3) excavate according to extraneous services request and learning path network, the set of study transaction graph bubble figure or two foaming figure; Excavate foaming figure when request is a Knowledge Element, excavates two foaming figure when request is two Knowledge Elements; First initialization candidate subgraph set when excavating, then in the middle of the learning path networking, finding has limit to be connected with candidate's subgraph and this limit has the node of maximum weights, this node is added to this candidate's subgraph in limit therewith, forms new candidate's subgraph set; Repeat this process, until the node number of candidate's subgraph reaches given threshold value, finally choose foaming figure or the two foaming figure of candidate's subgraph of relative support maximum as Result;

4) Result is recommended to the learner, and carry out gathering learning log in mutual process learner and learning system.

The described step of obtaining effective daily record is as follows:

The learning behavior log sheet produced in learner and the learning System reciprocal process based on Knowledge Map is shown { clickInfor _i; ClickInfor=(userId wherein, objectId, timeStamp, clickLength), userId means that the learner identifies, objectId means the corresponding learning object sign of this daily record, and timeStamp means the start time of the corresponding learning process of this daily record, and clickLength means the study duration of the corresponding learning process of this daily record;

At first all click behaviors in the learning behavior daily record are analyzed, are counted the corresponding relation between study duration clickLength and corresponding number of clicks clickCount, the click behavior in the learning behavior daily record is regarded as to the set of two types of clicks:

A kind of user of being and system are carried out the mutual generation of generality, and its number of clicks f (t) obeys Weibull distribution with the rule that stops duration t variation

t>=0;

Another kind is the daily record that learner's true learning process produces, and its number of clicks g (t) is with the rule Gaussian distributed that stops duration t variation $g\left(t\right)=C*e^{(-{((t-a)/b)}^c)};$

Using clickLength as independent variable t, clickCount regards the value sum of f (t) and g (t) as, and f (t) and g (t) are carried out to matching:

$f\left(t\right)=\left\{\begin{array}{cc}499*t^{0.09}*e^{(-0.12*t^{1.09})}& t>0\\ 0& t\≤0\end{array}\right.;$

$g\left(t\right)=94.14*e^{(-{((t-38.98)/21.32)}^2);}$

The fiducial interval that is 90% according to g (t) degree of confidence, the click stop duration threshold value of choosing learner's learning behavior is clickLengthThreshold;

Then filter out learning behavior daily record { clickInfor _iin all click duration clickLength _ithe clickInfor of<clickLengthThreshold _i, obtain effective daily record collection { clickInf _i.Described according to effect daily record collection { clickInf _ibuild learning path network LM step as follows:

The learning path net list is shown LM (LU, LE, lb (LU), lb (LE)), and wherein, LU means the Knowledge Element set that the learner learnt;

LE means the set of each nonoriented edge on learning path that the learner forms in learning process;

Lb (LU) means all node v in LU _ithe set that forms of mark, to arbitrary node v _i∈ LU, lb (v _i)=(v _i, ccv (v _i), ltv (v _i), lqv (v _i)); Ccv (v _i) mean node v _ithe accumulative total number of clicks, ltv (v _i) mean node v _iaccumulative total stop duration; Lqv (v _i) mean to learn node v _idifferent learners' quantity;

Lb (LE) means all e in LE _ithe set that forms of mark, to any limit e _i∈ LE, lb (e _i)=(e _i, jce (e _i), lte (e _i), lqe (e _i)); Jce (e _i) mean along e _ithe accumulative total number of hops, lte (e _i) mean along e _iaccumulative total study duration, lqe (e _i) mean along e _ithe different learners' of redirect accumulated quantity; (v _i, v _j) and (v _j, v _i) all mean the same limit;

At first add up effective daily record collection { clickInf _iin all learning objects sign objectId _i, form set LU, according to effective daily record collection { clickInf _ithe statistics LU in each node v _icorresponding accumulative total number of clicks ccv (v _i), accumulative total stops duration ltv (v _i), different learners' quantity lqv (v _i), form set lb (LU);

Next adds up effective daily record collection { clickInf _iin each learner userId _kcorresponding daily record collection { clickInf _k1, clickInf _k2..., clickInf _kn, and by limit (objectId _ki, objectId _{k (i+1)}) form in set LE; According to effective daily record collection { clickInf _ithe statistics LE in each limit e _iaccumulative total number of hops jce (e _i), accumulative total is learnt duration lte (e _i), different learners' accumulated quantity lqe (e _i), form set lb (LE).

The effective daily record collection of described basis { clickInf _ibuild study transaction graph set TG step as follows:

Study transaction graph set expression is TG (tg ₁, tg ₂..., tg _m), wherein, tg _ithe non-directed graph that means the learning path formation that the learner once logins, tg _i=(TV _i, TE _i); TV _imean study transaction graph tg _inode set,

lU means the Knowledge Element set that the learner learnt, TE _imean study transaction graph tg _ilimit set,

(v _i, v _j) and (v _j, v _i) mean identical limit; M means that the learner logins the number of times of study;

At first add up effective daily record collection { clickInf _iin each learner userId _kthe corresponding daily record collection of learning process { clickInf after each login _k1, clickInf _k2..., clickInf _kn, the corresponding study transaction graph of this process tg _i=(TV _i, TE _i) be:

TV _i={ objectId _k1, objectId _k2..., objectId _kn, objectId _knmean n the learning object that k learner learns;

Two learning objects form a limit,

TE _i={(objectId _k1,objectId _k2),(objectId _k2,objectId _k3),...,(objectId _k(n-1),objectId _kn)}；

Study transaction graph set TG is according to effective daily record collection { clickInf _ianalyze all study transaction graph tg draw _ithe set formed.

Describedly excavate with study transaction graph set TG the foaming figure or the two foaming figure that obtain according to learning path network LM and be expressed as cm (DV, CV, CE), wherein, the set that the learning objective Knowledge Element that DV is input forms, excavate foaming figure while in DV, a Knowledge Element only being arranged, if when two Knowledge Elements are arranged in DV excavate two foaming figure; CV is the set of Knowledge Element node in bubble figure or two foaming figure; CE is the set on limit between node in bubble figure or two foaming figure,

the concrete steps of excavating the foaming figure of take centered by learning objective Knowledge Element set DV or two foaming figure according to learning path network LM and study transaction graph set TG as:

1) if element number in DV | DV|=1, initialization 1-CM={ (DV, CV, CE) }, CV=DV,

if | DV|=2, initialization 2-CM={ (DV, CV, CE) }, CV=DV,

wherein, k-CM is the set with candidate's subgraph composition of k node;

2) initialization both candidate nodes set VS={v _i| dis (DV, v _i, LM)≤r}-DV, initialization candidate limit set ES={ (v _i, v _j) | v _i, v _j∈ VS, (v _i, v _j) ∈ LE},

Wherein, dis (DV, v _i, LM)=min{dis (v _i, v _j, LM) }, v _j∈ DV; Dis (v _i, v _j, LM) mean v _iwith v _jdistance in LM;

3) each node in VS is added respectively in CV, limit corresponding in ES is added in CE, form new k+1 node candidate sub collective drawing (k+1)-CM, the method that adds ingress and limit is: at first by node v _jadd in CV, then computing node v _jwith each node v in DV _ibetween weight the weight ((v on limit _i, v _j));

Weight ((v _i, v _j))=jce ((v _i, v _j))/[dis (DV, v _i, LM)+dis (DV, v _j, LM)], v _i∈ DV, by weight ((v _i, v _j)) maximum limit adds in CE and form (k+1)-cm=(DV, CV ∪ { v _j, CE ∪ (v _i, v _j)), and k+1 node candidate subgraph (k+1)-cm is added in k+1 node candidate sub collective drawing (k+1)-CM; If to certain node v in VS _j, all weight ((v _i, v _j)) all equal 0, do not add this node, k-cm to mean that certain has candidate's subgraph of k node;

4) to any two the candidate's subgraph k-cm in k-CM _i(DV, k-CV _i, k-CE _i) and k-cm _j(DV, k-CV _j, k-CE _j): if | k-CV _i∩ k-CV _j|=k-1 merges k-cm _iwith k-cm _jfor new (k+1)-cm _h(DV, k-CV _i∪ k-CV _j, (k+1)-CE _h), wherein, (k+1)-CE _hfor k-CV _i∪ k-CV _jthe set that limit in the maximum spanning tree of middle node forms; K-CV _i∪ k-CV _jin any two node v _iwith v _jbetween weight weight ((v _i, v _j))=jce ((v _i, v _j))/[dis (DV, v _i, LM)+dis (DV, v _j, LM)].

5) judgement (k+1)-cm _hwhether meet constraint condition, if meet, by (k+1)-cm _hadd in (k+1)-CM; Otherwise, by (k+1)-cm _hdelete; Being expressed as of constraint condition:

Wherein, | CV| is the number of nodes in bubble figure or two foaming figure cm;

$(\mathrm{ccv}\left(D\right)-\mathrm{ccv}\left(v_i\right))\_r=\frac{\left[\underset{v_k\&Element;\mathrm{DV}}{\mathrm{\&Sigma;}}(\mathrm{ccv}\left(v_k\right)-\mathrm{ccv}\left(v_i\right))\right]}{\left|\mathrm{DV}\right|}*\frac{\left|\mathrm{LU}\right|}{\underset{v_h\&Element;\mathrm{LU}}{\mathrm{\&Sigma;}}\mathrm{ccv}\left(v_h\right)};$

$(\mathrm{deg}\left(D\right)-\mathrm{deg}\left(v_i\right))\_r=\frac{\left[\underset{v_k\&Element;\mathrm{DV}}{\mathrm{\&Sigma;}}(\mathrm{deg}\left(v_k\right)-\mathrm{deg}\left(v_i\right))\right]}{\left|\mathrm{DV}\right|}*\frac{\left|\mathrm{LU}\right|}{\underset{v_h\&Element;\mathrm{LU}}{\mathrm{\&Sigma;}}\mathrm{deg}\left(v_h\right)};$

In formula: deg (v _i) expression node v _ithe degree value; Sup (cm, TG) _ r means the relative support of TG to cm;

Sup (cm, TG) means the support of TG to cm, | and { tg _j| tg _j∈ TG& LearnEqual (cm, tg _j)=1}|;

Study transaction graph set TG={tg ₁, tg ₂..., tg _m;

Tg _i=(TV _i, TE _i) mean the non-directed graph that learning path that the learner once logins forms;

TV _imean study transaction graph tg _inode set,

TE _imean study transaction graph tg _ilimit set,

(v _i, v _j) and (v _j, v _i) mean identical limit;

learnEqual(cm _i,tg _j)=1

s.t.

$\mathrm{for}\&ForAll;v_k\&Element;{\mathrm{CV}}_i,\&Exists;v_h\&Element;{\mathrm{CV}}_i\mathrm{makesdis}(v_k,v_h,{\mathrm{tg}}_j)\&le;\mathrm{interestDis}$

Wherein, cm _i=(DV _i, CV _i, CE _i) be i candidate's subgraph, DV _ithe target study node set that means candidate's subgraph, CV _ifor the node set of candidate's subgraph, CE _ifor the limit set of candidate's subgraph, set interestDis=5;

6) judge in newly-generated k-CM whether have two candidate's subgraph k-cm _iwith k-cm _jmeet | k-CV _i∩ k-CV _j|=k-1; If exist, return to 4) continue; Otherwise, if k-CM is empty, return to candidate's subgraph of relative support sup (cm, TG) _ r maximum in (k-1)-CM; If k-CM is not empty, return to candidate's subgraph of relative support maximum in k-CM.

Described for the figure that bubbles, in constraint condition, each threshold value is:

$\left\{\begin{array}{c}\mathrm{nodesThreshold}=7\\ \sup \_r\_\mathrm{Threshold}=0.05\\ \mathrm{ccv}\_r\_\mathrm{Threshold}=3\\ \mathrm{deg}\_r\_\mathrm{Threshold}=1.5\end{array}\right.$

For two foaming figure, in constraint condition, each threshold value is:

$\left\{\begin{array}{c}\mathrm{nodesThreshold}=10\\ \sup \_r\_\mathrm{Threshold}=0.075\\ \mathrm{ccv}\_r\_\mathrm{Threshold}=0\\ \mathrm{deg}\_r\_\mathrm{Threshold}=0\end{array}\right..$

The explicit queries that described request is the learner, the click of learner to certain Knowledge Element in Knowledge Map, or be the mark of field master of instruction for the teaching exercise.

Network Learning Resource recommend method based on the thinking map be input as learning objective Knowledge Element set DV, be output as foaming figure or two foaming figure cm, the concrete steps that education resource is recommended are:

1) daily record is carried out to pre-service, analytic learning duration threshold value, and learning log is filtered and obtained effective daily record collection { clickInf _i;

2) according to effective daily record collection { clickInf _ibuild learning path network LM and learn transaction graph set TG;

3) excavate foaming figure or the two foaming figure cm centered by DV;

4) Result recommended to the learner and gathered learning log.

Compared with prior art, the present invention has following useful technique effect:

Network Learning Resource recommend method based on the thinking map provided by the invention, based on two kinds of visualization tools in the thinking map: bubble figure and two figure of foaming, utilize learner's learning behavior daily record, Network Learning Resource is recommended.It has the characteristics of following four aspects:

The resource of 1, recommending is based on cognitive strategy.

2, recommend foaming figure can help the learner to learn better and understand description/with reference to the knowledge of type.

3, recommend two foaming figure can help the learner to learn better and understand the knowledge of reference/description type type.

4, mining algorithm can generate foaming figure and two foaming figure fast; According to learner's request, automatically generate relative illustrated education resource in the middle of the Knowledge Map system.

5, the present invention can provide the graphic knowledge recommendation service based on cognitive strategy for Web-based Learners, and then improves its learning efficiency.

The accompanying drawing explanation

Fig. 1 is that the inventive method relates to the mechanism figure that the Network Learning Resource based on the thinking map is recommended.

The topological structure exemplary plot that Fig. 2 is Knowledge Map.

Fig. 3 is the course learning prototype system structural drawing based on Knowledge Map.

Fig. 4 is the topological structure exemplary plot of bubble figure and two foaming figure.

Embodiment

For a more clear understanding of the present invention, below in conjunction with accompanying drawing, the present invention is described in further detail.The explanation of the invention is not limited.

Referring to Fig. 1, a kind of Network Learning Resource recommend method based on the thinking map comprises following operation:

1) gather the learning behavior daily record produced in learner and the learning System reciprocal process based on Knowledge Map, study duration in the analytic learning user behaviors log, obtain study duration threshold value, and with this threshold value to learning behavior daily record filtered, obtain effective daily record;

2) build learning path network and the set of study transaction graph according to the daily record data after filtering;

3) excavate according to extraneous services request and learning path network, the set of study transaction graph bubble figure or two foaming figure; Excavate foaming figure when request is a Knowledge Element, excavates two foaming figure when request is two Knowledge Elements; First initialization candidate subgraph set when excavating, then in the middle of the learning path networking, finding has limit to be connected with candidate's subgraph and this limit has the node of maximum weights, this node is added to this candidate's subgraph in limit therewith, forms new candidate's subgraph set; Repeat this process, until the node number of candidate's subgraph reaches given threshold value, finally choose foaming figure or the two foaming figure of candidate's subgraph of relative support maximum as Result;

Result is recommended to the learner, and carry out gathering learning log in mutual process learner and learning system.

Below be specifically described.The explanation of some concepts of using in given first the present invention:

1) log collection and pre-service

About Knowledge Element

It is Knowledge Element that weighing-appliance has the knowledge unit of complete knowledge representation.

For example the definition of the service in computer network is exactly a Knowledge Element.

About Knowledge Map

Daily record of the present invention source mainly is based on the Knowledge Map learning System.Knowledge Map is described the entity of a certain domain knowledge.Mainly by two parts, formed: the one, Knowledge Element; The 2nd, study dependence, i.e. incidence relation between Knowledge Element.Knowledge Map can be expressed as KM (KU, KE), and wherein KU means the Knowledge Element set, and KE means to learn the dependence set.

Fig. 2 is the instance graph of course Knowledge Map.

1.1 the collection of daily record

Required Data Source of the present invention is the click user behaviors log that the learner produces at the learning System learning based on Knowledge Map, and the prototype structure figure of learning system as shown in Figure 3.The log information that this method gathers mainly comprises: learner's sign, learning object (Knowledge Element) sign, the start time of learning process, the study duration of learning process.

1.2 daily record preprocess method

Because gathered raw data contains a large amount of invalid datas, learning records as too short as the study duration etc., thus need to carry out laying a solid foundation of pre-service follow-up work to original log.

The learning behavior log sheet produced in learner and the learning System reciprocal process based on Knowledge Map is shown { clickInfor _i; ClickInfor=(userId wherein, objectId, timeStamp, clickLength), userId means that the learner identifies, objectId means the corresponding learning object sign of this daily record, and timeStamp means the start time of the corresponding learning process of this daily record, and clickLength means the study duration of the corresponding learning process of this daily record; The concrete steps of daily record preprocess method are as follows:

Step 1: all click user behaviors logs are analyzed, counted the corresponding relation stopped between duration clickLength and corresponding number of clicks clickCount;

Step 2: user's click behavior is regarded as to the set of two types of clicks: a kind of user of being and learning system are carried out the mutual generation of generality, and its number of clicks f (t) obeys Weibull distribution with the rule that stops duration t variation $f\left(t\right)=C*t^{(b-1)}*e^{(-a\&CenterDot;t^b)}$ t≥0；

Another kind is the daily record that learner's true learning process produces, and its number of clicks g (t) is with the rule Gaussian distributed that stops duration t variation

Using clickLength as independent variable t, clickCount regards the value sum of f (t) and g (t) as, and f (t) and g (t) are carried out to matching:

$f\left(t\right)=\left\{\begin{array}{cc}499*t^{0.09}*e^{(-0.12*t^{1.09})}& t>0\\ 0& t\&le;0\end{array}\right.;$

$g\left(t\right)=94.14*e^{(-{((t-38.98)/21.32)}^2);}$

The click stop duration threshold value that the fiducial interval that is 90% according to g (t) degree of confidence is chosen learner's learning behavior is clickLengthThreshold.

Step 3: filter out original learning log collection { clickInfor _iin all click duration clickLength _ithe clickInfor of<clickLengthThreshold _i, obtain new effective daily record collection { clickInf _i.

2) learning path network struction

About the learning path network

The non-directed graph that the learning path produced while being learnt in the learning System based on Knowledge Map by the learner forms is learning path network LM.It is defined as follows:

LM=(LU,LE,lb(LU),lb(LE))

Wherein LU is the Knowledge Element set that the learner learnt, the set of each nonoriented edge on the path that LE forms in learning process for the learner,

lb (LU) is all v in LU _ithe set that forms of mark, to arbitrary node v _i∈ LU, lb (v _i)=(v _i, ccv (v _i), ltv (v _i), lqv (v _i)); Ccv (v _i) mean v _iaccumulative total number of clicks (click count of vertice, hereinafter to be referred as ccv); Ltv (v _i) mean v _iaccumulative total stop duration (learn time of vertice, hereinafter to be referred as ltv); Lqv (v _i) mean to learn v _idifferent learners' quantity (learner quantity of vertice, hereinafter to be referred as lqv).Lb (LE) is all e in LE _ithe set that forms of mark, to any limit e _i∈ LE, lb (e _i)=(e _i, jce (e _i), lte (e _i), lqe (e _i)); Jce (e _i) mean along e _iaccumulative total number of hops (jump count of edge, hereinafter to be referred as jce); Lte (e _i) mean along e _iaccumulative total study duration (learn time of edge, hereinafter to be referred as lte); Lqe (e _i) mean along e _ithe different learners' of redirect accumulated quantity (learner quantity of edge, hereinafter to be referred as lqe).Because LM is non-directed graph, therefore (v _i, v _j) and (v _j, v _i) all mean the same limit.

2.1 learning path network establishing method

The concrete construction method of learning path network is as follows:

Step 1: according to the daily record preprocess method add up effective log information clickInfor={clickInf _i.

Step 2: according to all clickInf _iin all learning object objectId _iform set LU, each node v in statistics LU _icorresponding accumulative total number of clicks ccv (v _i), accumulative total stops duration ltv (v _i), different learners' quantity lqv (v _ithereby) generation lb (LU);

Step 3: add up effective daily record collection { clickInf _iin each learner userId _kcorresponding daily record collection clickInf _k={ clickInf _k1, clickInf _k2..., clickInf _knand by the limit (objectId of learning object formation _ki, objectId _{k (i+1)}) be incorporated in LE each limit e in statistics LE _iaccumulative total number of hops jce (e _i), accumulative total is learnt duration lte (e _i), different learners' accumulated quantity lqe (e _i).

The structure of 3 study transaction graph set

About the study transaction graph

Study transaction graph set expression is TG (tg ₁, tg ₂..., tg _m), wherein, tg _ithe non-directed graph that means the learning path formation that the learner once logins, tg _i=(TV _i, TE _i); TV _imean study transaction graph tg _inode set,

lU means the Knowledge Element set that the learner learnt, TE _imean study transaction graph tg _ilimit set,

(v _i, v _j) and (v _j, v _i) mean identical limit; M means that the learner logins the number of times of study;

3.1 the building method of study transaction graph set

The set of study transaction graph is the union of learning each time affairs, and its concrete constitution step is as follows:

Step 1: by the daily record preprocess method, add up effective log information clickInfor={clickInf _i.

Step 2: add up effective daily record collection { clickInf _iin each learner userId _kcorresponding daily record collection clickInf _k={ clickInf _k1, clickInf _k2..., clickInf _kn}

Step 3: according to effective daily record collection { clickInf of each learner of gained in step 2 _i, find out corresponding learning object { objectId _kjthereby generation TV _i={ objectId _k1, objectId _k2..., objectId _kn;

Step 4:

Generate TE _i={ (objectId _k1, objectId _k2), (objectId _k2, objectId _k3) ..., (objectId _{k (n-1)}, objectId _kn);

Step 5: the TV generated by step 3 _itE with step 4 generation _ilearning of structure affairs tg _i=(TV _i, TE _i), thereby learning of structure affairs set TG=is (tg ₁, tg ₂..., tg _m); Study transaction graph set TG is according to effective daily record collection { clickInf _ianalyze all study transaction graph tg draw _ithe set formed.

The excavation of 4 foaming figure and two foaming figure

About the figure that bubbles

Non-directed graph by a learning objective Knowledge Element and a plurality of description/form with reference to Knowledge Element is foaming figure.

Hyerle has proposed for representing one group of visualized graphs instrument of 8 kinds of basic mode of thinking, i.e. thinking map according to cognitive psychology, Semantics Theory.Foaming figure belongs to a kind of instrument of the figure for the reference mode of thinking in the thinking map.The left figure of Fig. 4 is the figure that bubbles, and the node that wherein is labeled as " D " is the learning objective Knowledge Element, is labeled as the Knowledge Element that the node of " b " is the description/reference to the learning objective Knowledge Element.

About two foaming figure

Title is two foaming figure by two learning objective Knowledge Elements and non-directed graph that a plurality of contrast/the comparison Knowledge Element forms.It is a kind of for contrasting the figure instrument of the mode of thinking in the thinking map that two foaming figure belong to.The right figure of Fig. 4 is the figure that bubbles, and the node that wherein is labeled as " D1 " and " D2 " is two learning objective Knowledge Elements, is labeled as the Knowledge Element that the node of " b " is the contrast to the learning objective Knowledge Element/comparison.

About the learning state equivalence

learnEqual(cm _i,tg _j)=1

s.t.

$\mathrm{for}\&ForAll;v_k\&Element;{\mathrm{CV}}_i,\&Exists;v_h\&Element;{\mathrm{CV}}_i\mathrm{makesdis}(v_k,v_h,{\mathrm{tg}}_j)\&le;\mathrm{interestDis}$

Wherein, cm _i=(DV _i, CV _i, CE _i) be i candidate's subgraph, DV _ithe target study node set that means candidate's subgraph, CV _ifor the node set of candidate's subgraph, CE _ifor the limit set of candidate's subgraph, set interestDis=5.

4.1 bubble, figure describes with the mining algorithm of two foaming figure

The present invention is intended to wish according to the learner object knowledge unit of study, excavates other Knowledge Elements that description/reference relation or contrast/comparison are arranged with object knowledge unit, and presents to the learner with the form of foaming figure or two foamings.Utilizing learner's daily record excavation with v ₀in process for the foaming figure of learning objective node, relate generally to learning path network LM and generate, study affairs atlas TG generates, the generation of candidate's sub collective drawing CM, constraint condition f (cm) chooses, the judgement of study equivalent state, the problems such as the relative support calculating of candidate's subgraph.

Concrete, excavate with study transaction graph set TG the foaming figure or the two foaming figure that obtain according to learning path network LM and be expressed as cm (DV, CV, CE), wherein, the set that the learning objective Knowledge Element that DV is input forms, excavate foaming figure while in DV, a Knowledge Element only being arranged, if when two Knowledge Elements are arranged in DV excavate two foaming figure; CV is the set of Knowledge Element node in bubble figure or two foaming figure; CE is the set on limit between node in bubble figure or two foaming figure, the concrete steps of excavating the foaming figure of take centered by learning objective Knowledge Element set DV or two foaming figure according to learning path network LM and study transaction graph set TG as:

1) if element number in DV | DV|=1, initialization 1-CM={ (DV, CV, CE) }, CV=DV,

if | DV|=2, initialization 2-CM={ (DV, CV, CE) }, CV=DV,

wherein, k-CM is the set with candidate's subgraph composition of k node;

2) initialization both candidate nodes set VS={v _i| dis (DV, v _i, LM)≤r}-DV, initialization candidate limit set ES={ (v _i, v _j) | v _i, v _j∈ VS, (v _i, v _j) ∈ LE},

Wherein, dis (DV, v _i, LM)=min{dis (v _i, v _j, LM) }, v _j∈ DV; Dis (v _i, v _j, LM) mean v _iwith v _jdistance in LM;

3) each node in VS is added respectively in CV, limit corresponding in ES is added in CE, form new k+1 node candidate sub collective drawing (k+1)-CM, the method that adds ingress and limit is: at first by node v _jadd in CV, then computing node v _jwith each node v in DV _ibetween weight the weight ((v on limit _i, v _j));

Weight ((v _i, v _j))=jce ((v _i, v _j))/[dis (DV, v _i, LM)+dis (DV, v _j, LM)], v _i∈ DV, by weight ((v _i, v _j)) maximum limit adds in CE and form (k+1)-cm=(DV, CV ∪ { v _j, CE ∪ (v _i, v _j)), and k+1 node candidate subgraph (k+1)-cm is added in k+1 node candidate sub collective drawing (k+1)-CM; If to certain node v in VS _j, all weight ((v _i, v _j)) all equal 0, do not add this node, k-cm to mean that certain has candidate's subgraph of k node;

4) to any two the candidate's subgraph k-cm in k-CM _i(DV, k-CV _i, k-CE _i) and k-cm _j(DV, k-CV _j, k-CE _j): if | k-CV _i∩ k-CV _j|=k-1 merges k-cm _iwith k-cm _jfor new (k+1)-cm _h(DV, k-CV _i∪ k-CV _j, (k+1)-CE _h), wherein, (k+1)-CE _hfor k-CV _i∪ k-CV _jthe set that limit in the maximum spanning tree of middle node forms; K-CV _i∪ k-CV _jin any two node v _iwith v _jbetween weight weight ((v _i, v _j))=jce ((v _i, v _j))/[dis (DV, v _i, LM)+dis (DV, v _j, LM)].

5) judgement (k+1)-cm _hwhether meet constraint condition, if meet, by (k+1)-cm _hadd in (k+1)-CM; Otherwise, by (k+1)-cm _hdelete; Being expressed as of constraint condition:

Wherein, | CV| is the number of nodes in bubble figure or two foaming figure cm;

$(\mathrm{ccv}\left(D\right)-\mathrm{ccv}\left(v_i\right))\_r=\frac{\left[\underset{v_k\&Element;\mathrm{DV}}{\mathrm{\&Sigma;}}(\mathrm{ccv}\left(v_k\right)-\mathrm{ccv}\left(v_i\right))\right]}{\left|\mathrm{DV}\right|}*\frac{\left|\mathrm{LU}\right|}{\underset{v_h\&Element;\mathrm{LU}}{\mathrm{\&Sigma;}}\mathrm{ccv}\left(v_h\right)};$

$(\mathrm{deg}\left(D\right)-\mathrm{deg}\left(v_i\right))\_r=\frac{\left[\underset{v_k\&Element;\mathrm{DV}}{\mathrm{\&Sigma;}}(\mathrm{deg}\left(v_k\right)-\mathrm{deg}\left(v_i\right))\right]}{\left|\mathrm{DV}\right|}*\frac{\left|\mathrm{LU}\right|}{\underset{v_h\&Element;\mathrm{LU}}{\mathrm{\&Sigma;}}\mathrm{deg}\left(v_h\right)};$

In formula: deg (v _i) expression node v _ithe degree value; Sup (cm, TG) _ r means the relative support of TG to cm;

Sup (cm, TG) means the support of TG to cm, | and { tg _j| tg _j∈ TG& LearnEqual (cm, tg _j)=1}|;

Study transaction graph set TG={tg ₁, tg ₂..., tg _m;

Tg _i=(TV _i, TE _i) mean the non-directed graph that learning path that the learner once logins forms;

TV _imean study transaction graph tg _inode set,

TE _imean study transaction graph tg _ilimit set,

(v _i, v _j) and (v _j, v _i) mean identical limit;

learnEqual(cm _i,tg _j)=1

s.t.

$\mathrm{for}\&ForAll;v_k\&Element;{\mathrm{CV}}_i,\&Exists;v_h\&Element;{\mathrm{CV}}_i\mathrm{makesdis}(v_k,v_h,{\mathrm{tg}}_j)\&le;\mathrm{interestDis}$

Wherein, cm _i=(DV _i, CV _i, CE _i) be i candidate's subgraph, DV _ithe target study node set that means candidate's subgraph, CV _ifor the node set of candidate's subgraph, CE _ifor the limit set of candidate's subgraph, set interestDis=5;

6) judge in newly-generated k-CM whether have two candidate's subgraph k-cm _iwith k-cm _jmeet | k-CV _i∩ k-CV _j|=k-1; If exist, return to 4) continue; Otherwise, if k-CM is empty, return to candidate's subgraph of relative support sup (cm, TG) _ r maximum in (k-1)-CM; If k-CM is not empty, return to candidate's subgraph of relative support maximum in k-CM.

Further, the mining algorithm of foaming figure and two foaming figure is described as shown in table 1:

Table 1 bubbles and schemes and two foaming figure mining algorithms

4.2expandCM algorithm

The major function of expandCM algorithm is: when r≤Rthreshold, excavate each first stage of taking turns by the both candidate nodes collection r-VS of epicycle at foaming figure, candidate limit collection r-ES joins in candidate's sub collective drawing k-CM of last round of generation and generates new k+1 node candidate sub collective drawing (k+1)-CM, corresponding to the 5th step in table 1.The arthmetic statement of expandCM is as shown in table 2:

Table 2expandCM algorithm

4.3 candidate's subgraph expansion merge algorithm

The major function of combine algorithm is: in test-and-generation, by two k node candidate subgraph k-cm in k-CM _i(DV, CV _i, CE _i) and k-cm _j(DV, CV _j, CE _j) merging generation 1 k+1 node candidate subgraph (k+1)-cm (DV, CV, CE).The combine algorithmic procedure is as shown in table 3:

Table 3combine algorithm

4.4 support computing method

The function of support computing method computeSupport (TG, k-cm) is: calculate the support of study transaction graph collection TG to certain candidate's subgraph k-cm (DV, CV, CE).The key that support is calculated is certain study transaction graph tg (TV, TE) and k-cm learning state equivalence whether in judgement TG.The algorithm flow of computeSupport (TG, k-cm) is in Table 5:

Table 5 support computational algorithm

4.5 relative support computing method

Along with the growth of learning system learning person quantity and the continuity of learning time, the study transaction graph quantity in study affairs atlas TG can constantly increase, and TG also can be the trend of growth to the support of each candidate's subgraph k-cm.Support threshold value in constraint condition also needs constantly to adjust so, for follow-up foaming figure has brought larger complexity with the mining algorithm design of two foaming figure.For this reason, the present invention adopts " support relatively " to TG, the support of each candidate's subgraph k-cm to be measured.Support computational algorithm computeSupRatio (TG, k-cm) is as shown in table 6 relatively:

Table 6 is the support computational algorithm relatively

4.6 foaming constraint diagram condition

By learner's daily record is analyzed and pre-service, build the learning path network and it is carried out signature analysis and finds, foaming figure should meet constraint condition as shown in table 7:

The constraint condition threshold value of table 7 foaming figure and two foaming figure

5 recommend Result the learner and gather learning log

Overall, the concrete steps that e-learning of the present invention is recommended are as follows:

Step 1: learner's logging in network learning system enters foaming figure and recommends interface with two foaming figure;

Step 2: if the learner wishes that the object knowledge unit of study only has one, and wish to understand other Knowledge Elements that referring-to relation is arranged with this Knowledge Element, input object knowledge unit in foaming figure input frame; If the learner wishes to learn two object knowledge units, and wish to understand what similarities and differences that have of these two object knowledge units, i.e. relativity, input object knowledge unit in the input frame of two foaming figure;

Step 3: according to foaming figure mining algorithm of the present invention, the input parameter using user's input request as algorithm, excavate foaming figure, and form corresponding XML file push to foreground;

Step 4: XML is resolved, finally Result is presented to the user in the mode of scheming;

Step 5: recording user with learning System reciprocal process in the learning log that produces.

Claims (8)

1. the Network Learning Resource recommend method based on the thinking map is characterized in that: comprise following operation:

1) gather the learning behavior daily record produced in learner and the learning System reciprocal process based on Knowledge Map, study duration in the analytic learning user behaviors log, obtain study duration threshold value, and with this threshold value to learning behavior daily record filtered, obtain effective daily record;

2) build learning path network and the set of study transaction graph according to the daily record data after filtering;

3) excavate according to extraneous services request and learning path network, the set of study transaction graph bubble figure or two foaming figure; Excavate foaming figure when request is a Knowledge Element, excavates two foaming figure when request is two Knowledge Elements; First initialization candidate subgraph set when excavating, then in the middle of the learning path networking, finding has limit to be connected with candidate's subgraph and this limit has the node of maximum weights, this node is added to this candidate's subgraph in limit therewith, forms new candidate's subgraph set; Repeat this process, until the node number of candidate's subgraph reaches given threshold value, finally choose foaming figure or the two foaming figure of candidate's subgraph of relative support maximum as Result;

4) Result is recommended to the learner, and carry out gathering learning log in mutual process learner and learning system.

2. the Network Learning Resource recommend method based on the thinking map according to claim 1, is characterized in that, the step of obtaining effective daily record is as follows:

The learning behavior log sheet produced in learner and the learning System reciprocal process based on Knowledge Map is shown { clickInfor _i; ClickInfor=(userId wherein, objectId, timeStamp, clickLength), userId means that the learner identifies, objectId means the corresponding learning object sign of this daily record, and timeStamp means the start time of the corresponding learning process of this daily record, and clickLength means the study duration of the corresponding learning process of this daily record;

At first all click behaviors in the learning behavior daily record are analyzed, are counted the corresponding relation between study duration clickLength and corresponding number of clicks clickCount, the click behavior in the learning behavior daily record is regarded as to the set of two types of clicks:

A kind of user of being and system are carried out the mutual generation of generality, and its number of clicks f (t) obeys Weibull distribution with the rule that stops duration t variation

t>=0;

Another kind is the daily record that learner's true learning process produces, and its number of clicks g (t) is with the rule Gaussian distributed that stops duration t variation

Using clickLength as independent variable t, clickCount regards the value sum of f (t) and g (t) as, and f (t) and g (t) are carried out to matching:

$f\left(t\right)=\left\{\begin{array}{cc}499*t^{0.09}*e^{(-0.12*t^{1.09})}& t>0\\ 0& t\&le;0\end{array}\right.;$

$g\left(t\right)=94.14*e^{(-{((t-38.98)/21.32)}^2);}$

The fiducial interval that is 90% according to g (t) degree of confidence, the click stop duration threshold value of choosing learner's learning behavior is clickLengthThreshold;

Then filter out learning behavior daily record { clickInfor _iin all click duration clickLength _ithe clickInfor of<clickLengthThreshold _i, obtain effective daily record collection { clickInf _i.

3. the Network Learning Resource recommend method based on the thinking map according to claim 1, is characterized in that: according to effect daily record collection { clickInf _ibuild learning path network LM step as follows:

The learning path net list is shown LM (LU, LE, lb (LU), lb (LE)), and wherein, LU means the Knowledge Element set that the learner learnt;

LE means the set of each nonoriented edge on learning path that the learner forms in learning process;

Lb (LU) means all node v in LU _ithe set that forms of mark, to arbitrary node v _i∈ LU, lb (v _i)=(v _i, ccv (v _i), ltv (v _i), lqv (v _i)); Ccv (v _i) mean node v _ithe accumulative total number of clicks, ltv (v _i) mean node v _iaccumulative total stop duration; Lqv (v _i) mean to learn node v _idifferent learners' quantity;

Lb (LE) means all e in LE _ithe set that forms of mark, to any limit e _i∈ LE, lb (e _i)=(e _i, jce (e _i), lte (e _i), lqe (e _i)); Jce (e _i) mean along e _ithe accumulative total number of hops, lte (e _i) mean along e _iaccumulative total study duration, lqe (e _i) mean along e _ithe different learners' of redirect accumulated quantity; (v _i, v _j) and (v _j, v _i) all mean the same limit;

At first add up effective daily record collection { clickInf _iin all learning objects sign objectId _i, form set LU, according to effective daily record collection { clickInf _ithe statistics LU in each node v _icorresponding accumulative total number of clicks ccv (v _i), accumulative total stops duration ltv (v _i), different learners' quantity lqv (v _i), form set lb (LU);

Next adds up effective daily record collection { clickInf _iin each learner userId _kcorresponding daily record collection { clickInf _k1, clickInf _k2..., clickInf _kn, and by limit (objectId _ki, objectId _{k (i+1)}) form in set LE; According to effective daily record collection { clickInf _ithe statistics LE in each limit e _iaccumulative total number of hops jce (e _i), accumulative total is learnt duration lte (e _i), different learners' accumulated quantity lqe (e _i), form set lb (LE).

4. the Network Learning Resource recommend method based on the thinking map according to claim 3, is characterized in that: according to effective daily record collection { clickInf _ibuild study transaction graph set TG step as follows:

Study transaction graph set expression is TG (tg ₁, tg ₂..., tg _m), wherein, tg _ithe non-directed graph that means the learning path formation that the learner once logins, tg _i=(TV _i, TE _i); TV _imean study transaction graph tg _inode set, lU means the Knowledge Element set that the learner learnt, TE _imean study transaction graph tg _ilimit set, (v _i, v _j) and (v _j, v _i) mean identical limit; M means that the learner logins the number of times of study;

At first add up effective daily record collection { clickInf _iin each learner userId _kthe corresponding daily record collection of learning process { clickInf after each login _k1, clickInf _k2..., clickInf _kn, the corresponding study transaction graph of this process tg _i=(TV _i, TE _i) be:

TV _i={ objectId _k1, objectId _k2..., objectId _kn, objectId _knmean n the learning object that k learner learns;

Two learning objects form a limit,

TE _i={(objectId _k1,objectId _k2),(objectId _k2,objectId _k3),...,(objectId _k(n-1),objectId _kn)}；

Study transaction graph set TG is according to effective daily record collection { clickInf _ianalyze all study transaction graph tg draw _ithe set formed.

5. the Network Learning Resource recommend method based on the thinking map according to claim 4, it is characterized in that, excavate with study transaction graph set TG the foaming figure or the two foaming figure that obtain according to learning path network LM and be expressed as cm (DV, CV, CE), wherein, the set that the learning objective Knowledge Element that DV is input forms, excavate foaming figure while in DV, a Knowledge Element only being arranged, if when two Knowledge Elements are arranged in DV excavate two foaming figure; CV is the set of Knowledge Element node in bubble figure or two foaming figure; CE is the set on limit between node in bubble figure or two foaming figure,

the concrete steps of excavating the foaming figure of take centered by learning objective Knowledge Element set DV or two foaming figure according to learning path network LM and study transaction graph set TG as:

1) if element number in DV | DV|=1, initialization 1-CM={ (DV, CV, CE) }, CV=DV,

if | DV|=2, initialization 2-CM={ (DV, CV, CE) }, CV=DV,

wherein, k-CM is the set with candidate's subgraph composition of k node;

2) initialization both candidate nodes set VS={v _i| dis (DV, v _i, LM)≤r}-DV, initialization candidate limit set ES={ (v _i, v _j) | v _i, v _j∈ VS, (v _i, v _j) ∈ LE}, wherein, dis (DV, v _i, LM)=min{dis (v _i, v _j, LM) }, v _j∈ DV; Dis (v _i, v _j, LM) mean v _iwith v _jdistance in LM;

3) each node in VS is added respectively in CV, limit corresponding in ES is added in CE, form new k+1 node candidate sub collective drawing (k+1)-CM, the method that adds ingress and limit is: at first by node v _jadd in CV, then computing node v _jwith each node v in DV _ibetween weight the weight ((v on limit _i, v _j));

Weight ((v _i, v _j))=jce ((v _i, v _j))/[dis (DV, v _i, LM)+dis (DV, v _j, LM)], v _i∈ DV, by weight ((v _i, v _j)) maximum limit adds in CE and form (k+1)-cm=(DV, CV ∪ { v _j, CE ∪ (v _i, v _j)), and k+1 node candidate subgraph (k+1)-cm is added in k+1 node candidate sub collective drawing (k+1)-CM; If to certain node v in VS _j, all weight ((v _i, v _j)) all equal 0, do not add this node, k-cm to mean that certain has candidate's subgraph of k node;

4) to any two the candidate's subgraph k-cm in k-CM _i(DV, k-CV _i, k-CE _i) and k-cm _j(DV, k-CV _j, k-CE _j): if | k-CV _i∩ k-CV _j|=k-1 merges k-cm _iwith k-cm _jfor new (k+1)-cm _h(DV, k-CV _i∪ k-CV _j, (k+1)-CE _h), wherein, (k+1)-CE _hfor k-CV _i∪ k-CV _jthe set that limit in the maximum spanning tree of middle node forms; K-CV _i∪ k-CV _jin any two node v _iwith v _jbetween weight weight ((v _i, v _j))=jce ((v _i, v _j))/[dis (DV, v _i, LM)+dis (DV, v _j, LM)];

5) judgement (k+1)-cm _hwhether meet constraint condition, if meet, by (k+1)-cm _hadd in (k+1)-CM; Otherwise, by (k+1)-cm _hdelete; Being expressed as of constraint condition:

Wherein, | CV| is the number of nodes in bubble figure or two foaming figure cm;

$(\mathrm{ccv}\left(D\right)-\mathrm{ccv}\left(v_i\right))\_r=\frac{\left[\underset{v_k\&Element;\mathrm{DV}}{\mathrm{\&Sigma;}}(\mathrm{ccv}\left(v_k\right)-\mathrm{ccv}\left(v_i\right))\right]}{\left|\mathrm{DV}\right|}*\frac{\left|\mathrm{LU}\right|}{\underset{v_h\&Element;\mathrm{LU}}{\mathrm{\&Sigma;}}\mathrm{ccv}\left(v_h\right)};$

$(\mathrm{deg}\left(D\right)-\mathrm{deg}\left(v_i\right))\_r=\frac{\left[\underset{v_k\&Element;\mathrm{DV}}{\mathrm{\&Sigma;}}(\mathrm{deg}\left(v_k\right)-\mathrm{deg}\left(v_i\right))\right]}{\left|\mathrm{DV}\right|}*\frac{\left|\mathrm{LU}\right|}{\underset{v_h\&Element;\mathrm{LU}}{\mathrm{\&Sigma;}}\mathrm{deg}\left(v_h\right)};$

In formula: deg (v _i) expression node v _ithe degree value; Sup (cm, TG) _ r means the relative support of TG to cm;

Sup (cm, TG) means the support of TG to cm, | and { tg _j| tg _j∈ TG& LearnEqual (cm, tg _j)=1}|;

Study transaction graph set TG={tg ₁, tg ₂..., tg _m;

Tg _i=(TV _i, TE _i) mean the non-directed graph that learning path that the learner once logins forms;

TV _imean study transaction graph tg _inode set,

TE _imean study transaction graph tg _ilimit set,

(v _i, v _j) and (v _j, v _i) mean identical limit;

learnEqual(cm _i,tg _j)=1

s.t.

$\mathrm{for}\&ForAll;v_k\&Element;{\mathrm{CV}}_i,\&Exists;v_h\&Element;{\mathrm{CV}}_i\mathrm{makesdis}(v_k,v_h,{\mathrm{tg}}_j)\&le;\mathrm{interestDis}$

Wherein, cm _i=(DV _i, CV _i, CE _i) be i candidate's subgraph, DV _ithe target study node set that means candidate's subgraph, CV _ifor the node set of candidate's subgraph, CE _ifor the limit set of candidate's subgraph, set interestDis=5;

6) judge in newly-generated k-CM whether have two candidate's subgraph k-cm _iwith k-cm _jmeet | k-CV _i∩ k-CV _j|=k-1; If exist, return to 4) continue; Otherwise, if k-CM is empty, return to candidate's subgraph of relative support sup (cm, TG) _ r maximum in (k-1)-CM; If k-CM is not empty, return to candidate's subgraph of relative support maximum in k-CM.

6. the Network Learning Resource recommend method based on the thinking map as claimed in claim 5, is characterized in that, for the figure that bubbles, in constraint condition, each threshold value is:

$\left\{\begin{array}{c}\mathrm{nodesThreshold}=7\\ \sup \_r\_\mathrm{Threshold}=0.05\\ \mathrm{ccv}\_r\_\mathrm{Threshold}=3\\ \mathrm{deg}\_r\_\mathrm{Threshold}=1.5\end{array}\right.$

For two foaming figure, in constraint condition, each threshold value is:

$\left\{\begin{array}{c}\mathrm{nodesThreshold}=10\\ \sup \_r\_\mathrm{Threshold}=0.075\\ \mathrm{ccv}\_r\_\mathrm{Threshold}=0\\ \mathrm{deg}\_r\_\mathrm{Threshold}=0\end{array}\right..$

7. the Network Learning Resource recommend method based on the thinking map as claimed in claim 5, it is characterized in that, the explicit queries that described request is the learner, the click of learner to certain Knowledge Element in Knowledge Map, or be the mark of field master of instruction for the teaching exercise.

8. the Network Learning Resource recommend method based on the thinking map according to claim 1, it is characterized in that: the Network Learning Resource recommend method based on the thinking map be input as learning objective Knowledge Element set DV, be output as foaming figure or two foaming figure cm, the concrete steps that education resource is recommended are:

1) daily record is carried out to pre-service, analytic learning duration threshold value, and learning log is filtered and obtained effective daily record collection { clickInf _i;

2) according to effective daily record collection { clickInf _ibuild learning path network LM and learn transaction graph set TG;

3) excavate foaming figure or the two foaming figure cm centered by DV;

4) Result recommended to the learner and gathered learning log.

Images