CN106022463A - Personalized learning path optimization method based on improved particle swarm optimization algorithm - Google Patents
Personalized learning path optimization method based on improved particle swarm optimization algorithm Download PDFInfo
- Publication number
- CN106022463A CN106022463A CN201610314420.6A CN201610314420A CN106022463A CN 106022463 A CN106022463 A CN 106022463A CN 201610314420 A CN201610314420 A CN 201610314420A CN 106022463 A CN106022463 A CN 106022463A
- Authority
- CN
- China
- Prior art keywords
- particle
- solution
- algorithm
- neighborhood
- learning
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a personalized learning path optimization method based on an improved particle swarm optimization algorithm, comprising the following steps: (1) building a mathematical model of a learning path optimization problem; (2) performing learning path optimization based on an improved particle swarm optimization algorithm; and (3) analyzing the time complexity. According to the invention, by improving a standard particle swarm optimization algorithm, the defect that the standard particle swarm optimization algorithm is easily trapped into local optimum when solving an optimization problem is solved, and the method is advantaged in accuracy and success rate of search.
Description
Technical field
The present invention relates to particle cluster algorithm technical field, more specifically to based on the personalization improving particle cluster algorithm
Learning path optimization method.
Background technology
On-line study system is a kind of knowledge services mode relying on the emerging media such as the Internet to realize learning content transmission.
Under the promotion of information technology, on-line study is increasingly becoming a kind of main way obtaining knowledge.Although on-line study system is amassed
Tire out substantial amounts of education resource, but learner has often been difficult to be quickly found out suitable learning path and study from the resource of magnanimity
Content.Therefore, the intellectuality of on-line study system, individual character have been melted into the study hotspot of Chinese scholars.Intelligent learning system grinds
The key issue studied carefully is the optimization of learning path, system can according to the learning target of learner and mastery of knowledge's degree,
Suitable resource is carried out recombining contents, forms optimization learning path, enable learner to be rapidly completed learning target.
Particle swarm optimization algorithm (Particle Swarm Optimization, PSO) PSO algorithm is by simulation flock of birds
Migrate, during looking for food, a kind of evolutionary computation method proposed with clustering behavior, this algorithm definite conception, be easily achieved,
Solve complicated optimization problem aspect and have been achieved for preferable effect.But, owing to learning path optimization problem is an allusion quotation
The discrete type combinatorial optimization problem of type, traditional PSO algorithm is difficult to process sequence constraint relation therein.And traditional PSO
Algorithm lacks the dynamic regulation of speed, is easily trapped into local optimum, causes convergence precision low and is difficult to convergence.Therefore, it is difficult to it is logical
Cross traditional PSO algorithm and form optimized learning path.
A kind of also can the solution based on Modified particle swarm optimization algorithm of learning path optimization design improving particle cluster algorithm is learned
The method practising the optimization problem in path, is the necessity in intelligent learning system research.
This patent is for similar prior art or product problem to be solved:
At present, in order to obtain optimization learning path in intelligent learning system, mostly use population (PSO) algorithm, two
System population (BPSO) algorithm and standard genetic (GA) algorithm etc..Although particle cluster algorithm (PSO) and genetic algorithm (GA) power
Figure simulates the adaptability of individual population on the basis of natural characteristic, uses certain transformation rule search volume to solve, passes through
Randomized optimization process Population Regeneration and search optimum point.But they can not solve higher-dimension challenge well, often meets
To Premature Convergence and the shortcoming of constringency performance difference, it is impossible to ensure to converge to optimum point.Therefore, these algorithms can not efficiently and
Obtain the Optimal Learning path in intelligent learning system accurately.
Have for similar prior art this patent problem to be solved:
1, propose the mathematical model establishing method of learning path optimization problem, make typical combinatorial optimization problem；
2, in particle cluster algorithm, merge Neighborhood-region-search algorithm and taboo strategy, make up standard particle group's algorithm and ask in solution optimization
The defect of local optimum is easily sunk into during topic；
3, search accuracy and search success rate that this algorithm is applied in on-line study system are promoted.
Summary of the invention
For solving above-mentioned technical problem, the present invention provides based on the individualized learning path optimization side improving particle cluster algorithm
Method, this algorithm is to solve the defect of the local optimum in optimization learning path based on PSO algorithm in intelligent learning system,
Improve search accuracy and search success rate.
In order to realize above-mentioned technical purpose, the present invention adopts the following technical scheme that: based on the individual character improving particle cluster algorithm
Chemistry practises method for optimizing route, and its computational methods are as follows:
1) mathematical model of learning path optimization problem is set up:
2) based on the learning path optimization improving particle cluster algorithm:
On the basis of standard particle group's algorithm, Neighborhood-region-search algorithm based on taboo is incorporated into the searching process of population
In, define a kind of modified particle swarm optiziation TB-PSO；TB-PSO includes real number coding method, neighbour based on taboo
Domain search and the step of TB-PSO algorithm；
3) complexity analysis time:
In PSO algorithm, if the quantity of particle is N in population, the iterations of population is M1, each particle completes once
Operation time required for iteration is T1, then it can be calculated that the standard PSO short-cut counting method completes to optimize the total computing needing to expend
Time is N × M1×T1；
In TB-PSO algorithm, it is located atM1The number of times reaching neighborhood search condition in secondary iteration is M2, each neighborhood search performs
Largest extension number of times be K, the operation time that often performing neighborhood search needs is T2, then it can be calculated that TB-PSO algorithm is complete
Becoming to optimize the required total operation time expended is N × M1×T1(1+M2(K×T2)/M1)。
Further, the position vector of the particle that 6 education resources of step 2 are constituted is x={3.30,0.34,3.65,
4.75,5.28,2.73}, calculate result, r in this particle solution2It is not selected, and the study that remaining 5 education resources are corresponding
Path is { r6, r1, r3, r4, r5, i.e. the learning sequence of resource is 6-1-3-4；Particle solution x={3.30,0.34,3.65,
4.75,5.28, the 2.73} variablees converted are x61=1, x13=1, x34=1, remaining xijIt is 0.
Further, the object function of the learning path optimization problem problem of step 1 is:
(1) learning difficulty: the learning difficulty of all education resources on recommendation learning path is with mastery of knowledge's level of learner
Gap；
(2) study spends: recommend the study of all education resources on learning path to spend；
(3) degree of association: recommend all education resources on learning path with the degree of association gap of object knowledge point；
The constraints of learning path optimization problem problem is:
Ensure that and have on each object knowledge point to be learned and an education resource can only be selected.
Further, the real number coding method of step 2 uses real number coding method to change into discrete by continuous print real solution
Integer solution, particle solution { x1, x2, x3···xnBe made up of integer and fractional part, it is expressed as xi=(Ii, Di), wherein, Ii
Indicate whether to choose this education resource, work as IiWhen being 0, illustrate not select this resource；The size order solved represents education resource
Path selection order；The discrete solution drawn is converted in mathematical model the variable of correspondence again.
Further, the neighborhood search based on taboo of step 2, its algorithm committed step is as follows:
1) neighborhood extending
X is n-dimensional vector { x1, x2,x3…xn, if δ is arbitrary positive number, then open interval (xi-δ, xi+ δ) it is xiA neighborhood,
It is denoted as U(xi, δ)；The X δ neighborhood that open interval (X-δ, X+ δ) is X in all dimensions, is denoted as U(X, δ)；xiδ neighborhood remove
After central point, referred to as xiRemove heart δ neighborhood U(xi, δ)；
The current employing mode that solves is carried out 1 δ neighborhood extending, in neighborhood, generates m candidate solution at random, by current solution with m
The fitness of candidate solution compares, and evaluates the adaptive value of these candidate solutions, and therefrom selects optimal candidate solution；Without searching
Rope, to qualified solution, is further continued for carrying out 1 δ neighborhood, until reaching condition or reaching to extend number of times k；
2) taboo strategy setting
After neighborhood is carried out δ neighborhood extending in a manner described, the optimum solution from candidate solution, it is used for substituting current solution and best
Solve, then this solution is added in taboo list；
3) aspiration criterion
If fitness corresponding to current taboo object is better than history optimal solution, then ignore its taboo attribute, still as
Current selection, this way can effectively prevent from losing optimal solution, improve the Searching efficiency of algorithm.
Further, the TB-PSO algorithm steps of step 2 is specific as follows:
A: initialize population, randomly generate position and the speed of all particles
If particle solution space variable X is { x1, x2,x3…xn, wherein n is the quantity of education resource, at xiInterval (0,
N) interior stochastic generation one initializes particle and speed；
The fitness evaluation of B: particle
First, according to Code Mapping method, the variable X of continuous is mapped as discrete solution；
Secondly, the premise of the individual optimal solution and globally optimal solution that calculate population determines that fitness function, at object functionf 1、f 2、f 3On the basis of constraints, formed comprehensive fitness function evaluate the individual optimal solution of population with
Globally optimal solution, wherein: 0 < a < 1；,,Represent the weight of 3 object functions respectively； It is that constraints is brought in object function as penalty；
C: the speed of more new particle and position
Particle uses the more new formula of PSO algorithm at the speed of solution space and location updating:
Wherein,For inertia weight；r1,r2For aceleration pulse；Rand() it is interval [0,1] upper equally distributed random number；PtWith
GtIt is respectively self desired positions pbest and overall situation desired positions gbest of t particle；Pbest is that particle is crossed from status
Desired positions, and gbest is the overall desired positions corresponding to particle, it is the desired positions that whole colony is experienced；WithPosition And Velocity for moment t；
D: judge whether to meet iterations: if it is satisfied, export optimal solution；If be unsatisfactory for, perform E；
E: if globally optimal solution reaches to start the condition of Local Search, perform neighborhood search based on taboo；Otherwise, jump to
B。
The technical characterstic of the present invention and effect be: wants to be obtained by algorithm the learning path of optimum, first, should set up
Practise the mathematical model of routing problem；Secondly, by the improvement to particle cluster algorithm, by Neighborhood-region-search algorithm and taboo strategy
Introduce wherein, obtain Optimal Learning path；Finally, being proved by the application test result in on-line study system, this algorithm exists
In search accuracy and search success rate, aspect has certain advantage；
By solving population (PSO) algorithm of optimal value, binary system population (BPSO) algorithm and standard genetic with existing
(GA) algorithm compares, and this algorithm, by the improvement to standard particle group's algorithm, solves standard particle group's algorithm excellent in solution
Change problem is the defect easily sinking into local optimum, also has advantage in terms of search accuracy and search success rate simultaneously.
Accompanying drawing explanation
Fig. 1 is knowledge point of the present invention ordering relation figure.
Fig. 2 is present invention Hybrid Particle Swarm based on Memetic framework flow process.
10 knowledge point ordering relations of Fig. 3 present invention " space and figure " part.
Fig. 4 is the searching process contrast of 3 kinds of algorithms of the embodiment of the present invention 2.
Fig. 5 is the optimizing number of success contrast of 3 kinds of algorithms of the embodiment of the present invention 2.
Detailed description of the invention
In order to make the purpose of the present invention, technical scheme and advantage clearer, below in conjunction with drawings and Examples, right
The present invention is further elaborated.Should be appreciated that specific embodiment described herein only in order to explain the present invention, and
It is not used in the restriction present invention.
Embodiment 1
The present invention provides based on the individualized learning method for optimizing route improving particle cluster algorithm, and its optimization method is as follows:
The first step: set up the mathematical model of learning path optimization problem
In on-line study system, in order to recommend personalized education resource and path to learner, on the one hand need to consider
To ' Current Knowledge Regarding and the learning cost of learner self, on the other hand need the learning sequence in view of object knowledge point.
Existing document[5,8,9]In, when scholars set up the mathematical model of learning path optimization problem, it is contemplated that conceptual dependency degree,
The factors such as learning difficulty, learners' knowledge level, but not using ordering relation as the restriction relation of essence in view of mathematical modulo
In type, have impact on the practical function that on-line study system resource is recommended to a certain extent.
Therefore, the optimization of learning path, need to set up on the basis of domain knowledge structure figure.The target that user is to be learned
Ordering relation between knowledge point and knowledge point can constitute a knowledge point structure figure, each knowledge point is corresponding several
The education resource that difficulty is different, as shown in Figure 1.The process of user learning coordinates measurement needs according to user's existing study shape
State, in conjunction with the ordering relation between knowledge point structure figure, it is achieved the personalized generation of learning path, forms coverage goal knowledge point
Optimization path.Learning path optimized algorithm needs learning target based on user, according to knowledge point topological structure and study
Person's state, calculates the education resource assembled scheme of optimum, to meet the individual demand of learner, and to follow all kinds of
Constraints.
Learning path optimization problem can be described as: all knowledge points that known course comprises and the education resource of correspondence,
The learning target of learner and mastery of knowledge's level.Learner is from the beginning of some education resource, by the order between knowledge point
The relation education resource that learning target knowledge point is corresponding successively, and each knowledge point can only learn once, completes all mesh
After the study of mark knowledge point, terminate learning process.This problem solving result is one group of learning path being made up of education resource, and
Path is made to meet condition: (1) education resource difficulty is most suitable；(2) least cost is always learnt；(3) the same mesh of education resource chosen
Mark knowledge point degree of association is maximum.
The parameter of this problem is set to:
r i :Representing i-th education resource, 1N, N are the sum of education resource；
k p: representing pth knowledge point to be learned, 1pM, M are the sum of knowledge point to be learned；
c ip: represent i-th education resource with the degree of association of pth knowledge point, 1iN, 1p, 0c ip1；
s(r i,r j): represent that the study between i-th education resource and jth education resource spends, namely learning cost.
If education resourcer iWith education resourcer jCorresponding knowledge point is ordering relation in knowledge structure graph, then between them
Study spend be designated as 1；If backward relation, then learning cost is penalty constant PL。
d i: represent i-th education resourcer iLearning difficulty；
a j: represent the learner grasp level to jth knowledge point, wherein 0a j1；
The variable of this problem is set to:
If variable xijRepresent education resource recommendation paths, 1iN, 1N:
If variable yiThe selection situation of expression education resource:
The object function of this problem is:
(1) learning difficulty: the learning difficulty of all education resources on recommendation learning path is with mastery of knowledge's level of learner
Gap；
(4) study spends: recommend the study of all education resources on learning path to spend；
(5) degree of association: recommend all education resources on learning path with the degree of association gap of object knowledge point.
The constraints of this problem is:
Formula 4 ensure that to be had on each object knowledge point to be learned and can only select an education resource.
Second step: based on the learning path optimization improving particle cluster algorithm
In terms of learning path optimization, this patent is suggested plans on the basis of standard particle group's algorithm, by neighbour based on taboo
Domain search algorithm is incorporated in the searching process of population, defines a kind of modified particle swarm optiziation TB-PSO.This calculation
Method, to being absorbed in the individuality of locally optimal solution in particle populations, shield by taboo strategy, and searching in its neighborhood can
The more excellent solution of energy, thus increase the multiformity of algorithm solution space.
Wherein real number coding method
Owing to particle cluster algorithm is a kind of continuous optimized algorithm, the span of particle is real number, and this patent is suggested plans
The value of built mathematical model is discrete integer, and therefore, this patent is suggested plans employing real number coding method by continuous print in fact
Number solution changes into discrete integer solution.Particle solution { x1, x2, x3···xnBe made up of integer and fractional part, can represent
For xi=(Ii, Di).Wherein, IiIndicate whether to choose this education resource, work as IiWhen being 0, illustrate not select this resource；The size solved
Sequence list shows the Path selection order of education resource.
The position vector arranging the particle that 6 education resources are constituted is x={3.30,0.34,3.65,4.75,5.28,
2.73}
Can calculate, r in this particle solution2It is not selected, and learning path corresponding to remaining 5 education resources is { r6,
r1, r3, r4, r5, i.e. the learning sequence of resource is 6-1-3-4.
Finally, the discrete solution drawn is converted in mathematical model the variable of correspondence.
Wherein particle solution x={3.30,0.34,3.65,4.75,5.28, the 2.73} variable converted is x61=1, x13=1, x34=
1, remaining xijIt is 0.
Neighborhood search based on taboo
TABU search arranges taboo strategy by employing and limits search and be absorbed in local best points, is a kind of overall excellent algorithm.This
Literary composition, during accelerating local optimal searching by neighborhood search, introduces taboo strategy and avoids particle roundabout at local best points
Search, and then ensure diversified have efficient search to realize global optimization, algorithm committed step is as follows:
Neighborhood extending
X is n-dimensional vector { x1, x2,x3…xn, if δ is arbitrary positive number, then open interval (xi-δ, xi+ δ) it is xiA neighborhood,
It is denoted as U(xi, δ)；The X δ neighborhood that open interval (X-δ, X+ δ) is X in all dimensions, is denoted as U(X, δ).xiδ neighborhood remove
After central point, referred to as xiRemove heart δ neighborhood U (xi, δ).
Use aforesaid way to carry out 1 δ neighborhood extending current solution, in neighborhood, generate m candidate solution at random, will be current
Solution compares with the fitness of m candidate solution, evaluates the adaptive value of these candidate solutions, and therefrom selects optimal candidate solution；As
Fruit does not search qualified solution, is further continued for carrying out 1 δ neighborhood, until reaching condition or reaching to extend number of times k.
Taboo strategy setting
After the mode of neighborhood (1) by definition is carried out δ neighborhood extending, the optimum solution from candidate solution, be used for substituting current solve and
Preferably solve, then this solution is added in taboo list.
Aspiration criterion
If the fitness that current taboo object is corresponding is better than history optimal solution, then ignores its taboo attribute, still made
Selecting for current, this way can effectively prevent from losing optimal solution, improves the Searching efficiency of algorithm.
Algorithm steps
The step of TB-PSO algorithm is as in figure 2 it is shown, specific as follows:
Setp1: initialize population, randomly generate position and the speed of all particles
If particle solution space variable X is { x1, x2,x3…xn, wherein n is the quantity of education resource, at xiInterval (0,
N) interior stochastic generation one initializes particle and speed.
: the fitness evaluation of particle
First, according to the Code Mapping method of above-mentioned introduction, the variable X of continuous is mapped as discrete solution；
Secondly, the premise of the individual optimal solution and globally optimal solution that calculate population determines that fitness function, has given
3 object functions of this problem are gone outf 1、f 2、f 3And constraints, therefore, here at 3 object functions and constraint bar
On the basis of part, form a comprehensive fitness function and evaluate individual optimal solution and the globally optimal solution of population.
Wherein:
0<a<1；
,,Represent the weight of 3 object functions respectively；
It is that constraints is brought in object function as penalty.
: the speed of more new particle and position
Particle uses the more new formula of PSO algorithm at the speed of solution space and location updating:
Wherein,For inertia weight；r1,r2For aceleration pulse；Rand() it is interval [0,1] upper equally distributed random number；PtWith
GtIt is respectively self desired positions pbest and overall situation desired positions gbest of t particle.Pbest is that particle is crossed from status
Desired positions, and gbest is the overall desired positions corresponding to particle, it is the desired positions that whole colony is experienced.WithPosition And Velocity for moment t.
: judge whether to meet iterations: if it is satisfied, export optimal solution；If be unsatisfactory for, perform Step5
When the solution of particle reaches default operational precision or iterations, terminate search；Otherwise will jump to the 5th step start
Perform local neighborhood based on taboo search.
: if globally optimal solution reaches to start the condition of Local Search, performs neighborhood search based on taboo；Otherwise, jump
Forward Step2 to
If the globally optimal solution of population differs less after successive ignition, this algorithm is thought that population is doubtful and has been absorbed in office
Portion is optimum, has reached to start the condition of Local Search, therefore, will use the method for the 3.2nd trifle, use local based on taboo
Search.If same current solution being continuously extended into after k time set in advance than current solution more excellent solution being still not found, then calculate
Method terminates.Otherwise, algorithm will jump to the 2nd step.
3rd step: analysis time complexity
In PSO algorithm, if the quantity of particle is N in population, the iterations of population is M1, each particle completes once
Operation time required for iteration is T1, then it can be calculated that the standard PSO short-cut counting method completes to optimize the total computing needing to expend
Time is N × M1×T1。
In TB-PSO algorithm, it is located atM1The number of times reaching neighborhood search condition in secondary iteration is M2, each neighborhood search
The largest extension number of times performed is K, and the operation time often performing neighborhood search needs is T2, then it can be calculated that TB-
It is N × M that PSO algorithm completes to optimize total operation time of required consuming1×T1(1+M2(K×T2)/M1)。
Embodiment 2
Have chosen 10 knowledge points of Junior Mathematics as object knowledge point, under each knowledge point, have 5 difficulty are different
Practising resource, ordering relation between each knowledge point, the difficulty of education resource, education resource are with the degree of association of knowledge point, this student
Grasp level etc. to knowledge point pair is all known conditions, and it notes at foot.
(1) 10 knowledge points of selection Junior Mathematics " space and figure " part are as experimental data, these 10 knowledge points
It is respectively as follows: 1. straight line and line segment；2. the tolerance of angle and expression；The most parallel and vertical；4. the character of triangle；The most parallel four limits
The character of shape；6. circular character；7. congruent triangles；8. similar triangles；9. Pythagorean theorem；10. round area.
Herein with { k1, k2..., k10Representing above-mentioned knowledge point successively, its knowledge point learning sequence figure is as shown in Figure 2.
Such as, only at study k1After " straight line and line segment ", k could be learnt2" tolerance of angle and expression ", Jin Ercai
K can be learnt4, k5, k6Deng k2Sub-knowledge point.
(2), in learner is arranged on the interval of 0 ~ 1 to the Grasping level of these 10 knowledge points, wherein 0 expression is not slapped completely
Holding, 1 expression is grasped completely.The value of this Setup Experiments be followed successively by 0.9,0.5,0.6,0.3,0.4,0.4,0.2,0.15,0.15,
0.1}。
(3) in these 10 knowledge points, the resource that 5 difficulty that each knowledge point is corresponding are different, have 50 study moneys
Source, r1~r50, difficulty and the resource of each knowledge point correspondence resource are as shown in table 1 with the degree of association details of this knowledge point.
Wherein, in table, preceding paragraph is the difficulty of resource, consequent for resource with the degree of association of knowledge point:
Table 1 knowledge point and education resource association attributes table
In order to verify the performance of improvement particle cluster algorithm TB-PSO in this paper, we by carried algorithm herein with binary system grain
Subgroup (BPSO) algorithm contrasts with standard genetic (GA) algorithm.Binary particle swarm algorithm BPSO is in PSO algorithm
On a kind of discrete binary version PSO algorithm of extending out, this algorithm in the model proposed by the most one-dimensional of particle and particle
The solution of itself is defined to 1 or 0.When the position of particle is updated position, speed is set a threshold value, when the speed of particle
When degree is higher than this threshold value, the position of particle takes 1, otherwise takes 0.Binary system PSO is the most much like with genetic algorithm, but experiment
Result shows, in most of test functions, binary system PSO is faster than genetic algorithm speed[7].Standard GA algorithm is that one is passed through
The method of simulation natural evolution process searches optimal solution.After the generation made new advances of often evolving, according to adaptation individual in Problem Areas
Degree size selects individuality, and is combined intersecting and variation by means of the genetic operator of natural genetics, produce represent new
The population of disaggregation.Optimum individual in its population can be as problem approximate optimal solution in last reign of a dynasty.
The platform that this emulation experiment uses is Windows XP, and Matlab 7.0, the hardware configuration of computer used is
CPU:Pentium E6500 2.93GHz；Internal memory: 2G RAM；Hard disk: 500G.
Interpretation
The searching process relative analysis of three kinds of algorithms
The main checking of this experiment improves particle cluster algorithm TB-PSO carrying compared with other two kinds of algorithms in terms of algorithm operational performance
Rising, Fig. 4 reflects the optimizing curve of the globally optimal solution that each algorithm iteration number of times obtains, wherein vertical coordinate in single experiment with it
For the fitness after each iteration, abscissa is iterations, and the iterations of three algorithms is all set to 100 times.Need explanation
, 50 iteration of having of TB-PSO algorithm are the optimization processes of standard PSO, and other 50 iteration are to carry out with TABU search again
The process optimized.
As seen from Figure 4, standard PSO convergence rate is too fast, easily sink into local owing to still suffering from for binary system PSO algorithm
Excellent shortcoming, effect is the most general.TB-PSO method in this paper, by introducing TABU search, i.e. adds in the motion of particle
Enter random disturbance, add " climbing the mountain " ability of algorithm, so that algorithm performance is introducing the laggard one-step optimization of TABU search, have
Overcome PSO algorithm and be prone to be absorbed in the problem of locally optimal solution to effect, compare binary system PSO algorithm and achieve and preferably optimize effect
Really, the occurrence number of target solution is added.And standard GA algorithm is all far above due to its required number of particles and iterations
PSO algorithm, causes under the present conditions, and its search performance is poor.
The optimization number of success relative analysis of three kinds of algorithms
In order to verify above-mentioned 3 kinds of algorithms optimizing performance the parameter such as population, greatest iteration number is homogeneous while, herein according to grain
Gradually progressively increasing of subgroup quantity, is provided with 11 groups of experiments altogether, each number of particles difference arranged between experiment of organizing, but group interior 3
The number of particles planting algorithm is identical.The iterations of each algorithm is 200 times, if algorithm obtains after reaching iterations
The fitness arrived is specifying below threshold value, just thinks that this experiment algorithm optimizing is successful.Often group experiment is repeated 10 times, in test
The each algorithm of middle statistics can the successful number of times of optimizing.Fig. 5 illustrates 3 kinds of algorithms optimizing number of success pair in each group of experiment
Ratio.
It is found that when population quantity is on the low side, three kinds of algorithm effects are the most bad.But the quantity in population reaches
After more than 60, gradually becoming abundant due to number of particles, the search effect of improvement particle cluster algorithm TB-PSO in this paper is also
Being obviously improved, it optimizes performance also above binary system PSO algorithm and standard GA algorithm.
The foregoing is only presently preferred embodiments of the present invention, not in order to limit the present invention, all essences in the present invention
Any amendment, equivalent and the improvement etc. made within god and principle, should be included within the scope of the present invention.
Claims (6)
1. based on the individualized learning method for optimizing route improving particle cluster algorithm, it is characterised in that: its computational methods are as follows:
1) mathematical model of learning path optimization problem is set up:
2) based on the learning path optimization improving particle cluster algorithm:
On the basis of standard particle group's algorithm, Neighborhood-region-search algorithm based on taboo is incorporated into the searching process of population
In, define a kind of modified particle swarm optiziation TB-PSO；TB-PSO includes real number coding method, neighbour based on taboo
Domain search and the step of TB-PSO algorithm；
3) complexity analysis time:
In PSO algorithm, if the quantity of particle is N in population, the iterations of population is M1, each particle completes once
Operation time required for iteration is T1, then it can be calculated that the standard PSO short-cut counting method completes to optimize the total computing needing to expend
Time is N × M1×T1；
In TB-PSO algorithm, it is located atM1The number of times reaching neighborhood search condition in secondary iteration is M2, each neighborhood search performs
Largest extension number of times is K, and the operation time often performing neighborhood search needs is T2, then it can be calculated that TB-PSO algorithm is complete
Becoming to optimize the required total operation time expended is N × M1×T1(1+M2(K×T2)/M1)。
Individualized learning method for optimizing route based on improvement particle cluster algorithm the most according to claim 1, its feature exists
In: the position vector of the particle that 6 education resources of step 2 are constituted is x={3.30,0.34,3.65,4.75,5.28,2.73},
Calculate result, r in this particle solution2It is not selected, and learning path corresponding to remaining 5 education resources is { r6, r1,
r3, r4, r5, i.e. the learning sequence of resource is 6-1-3-4；Particle solution x={3.30,0.34,3.65,4.75,5.28,2.73}
The variable converted is x61=1, x13=1, x34=1, remaining xijIt is 0.
Individualized learning method for optimizing route based on improvement particle cluster algorithm the most according to claim 1, its feature exists
In: the object function of the learning path optimization problem problem of step 1 is:
(1) learning difficulty: the learning difficulty of all education resources on recommendation learning path is with mastery of knowledge's level of learner
Gap；
Study spends: recommend the study of all education resources on learning path to spend；
Degree of association: recommend all education resources on learning path with the degree of association gap of object knowledge point；
The constraints of learning path optimization problem problem is:
Ensure that and have on each object knowledge point to be learned and an education resource can only be selected.
Individualized learning method for optimizing route based on improvement particle cluster algorithm the most according to claim 1, its feature exists
In: the real number coding method of step 2 uses real number coding method that continuous print real solution changes into discrete integer solution, particle solution
{x1, x2, x3···xnBe made up of integer and fractional part, it is expressed as xi=(Ii, Di), wherein, IiIndicate whether to choose this
Education resource, works as IiWhen being 0, illustrate not select this resource；The size order solved represents the Path selection order of education resource；Again
The discrete solution drawn is converted in mathematical model the variable of correspondence.
Individualized learning method for optimizing route based on improvement particle cluster algorithm the most according to claim 1, its feature exists
In: the neighborhood search based on taboo of step 2, its algorithm committed step is as follows:
1) neighborhood extending
X is n-dimensional vector { x1, x2,x3…xn, if δ is arbitrary positive number, then open interval (xi-δ, xi+ δ) it is xiA neighborhood, note
Make U(xi, δ)；The X δ neighborhood that open interval (X-δ, X+ δ) is X in all dimensions, is denoted as U(X, δ)；xiδ neighborhood remove
After heart point, referred to as xiRemove heart δ neighborhood U(xi, δ)；
The current employing mode that solves is carried out 1 δ neighborhood extending, in neighborhood, generates m candidate solution at random, by current solution with m
The fitness of candidate solution compares, and evaluates the adaptive value of these candidate solutions, and therefrom selects optimal candidate solution；Without searching
Rope, to qualified solution, is further continued for carrying out 1 δ neighborhood, until reaching condition or reaching to extend number of times k；
2) taboo strategy setting
After neighborhood is carried out δ neighborhood extending in a manner described, the optimum solution from candidate solution, it is used for substituting current solution and best
Solve, then this solution is added in taboo list；
3) aspiration criterion
If fitness corresponding to current taboo object is better than history optimal solution, then ignore its taboo attribute, still as
Current selection, this way can effectively prevent from losing optimal solution, improve the Searching efficiency of algorithm.
Individualized learning method for optimizing route based on improvement particle cluster algorithm the most according to claim 1, its feature exists
In: the TB-PSO algorithm steps of step 2 is specific as follows:
A: initialize population, randomly generate position and the speed of all particles
If particle solution space variable X is { x1, x2,x3…xn, wherein n is the quantity of education resource, at xiInterval (0,
N) interior stochastic generation one initializes particle and speed；
The fitness evaluation of B: particle
First, according to Code Mapping method, the variable X of continuous is mapped as discrete solution；
Secondly, the premise of the individual optimal solution and globally optimal solution that calculate population determines that fitness function, at object functionf 1、f 2、f 3On the basis of constraints, formed comprehensive fitness function evaluate the individual optimal solution of population with
Globally optimal solution, wherein: 0 < a < 1；,,Represent the weight of 3 object functions respectively； It is that constraints is brought in object function as penalty；
C: the speed of more new particle and position
Particle uses the more new formula of PSO algorithm at the speed of solution space and location updating:
Wherein,For inertia weight；r1,r2For aceleration pulse；Rand() it is interval [0,1] upper equally distributed random number；PtWith
GtIt is respectively self desired positions pbest and overall situation desired positions gbest of t particle；Pbest is that particle is crossed from status
Desired positions, and gbest is the overall desired positions corresponding to particle, it is the desired positions that whole colony is experienced；WithPosition And Velocity for moment t；
D: judge whether to meet iterations: if it is satisfied, export optimal solution；If be unsatisfactory for, perform E；
E: if globally optimal solution reaches to start the condition of Local Search, perform neighborhood search based on taboo；Otherwise, jump to
B。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610314420.6A CN106022463B (en) | 2016-05-13 | 2016-05-13 | Based on the individualized learning method for optimizing route for improving particle cluster algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610314420.6A CN106022463B (en) | 2016-05-13 | 2016-05-13 | Based on the individualized learning method for optimizing route for improving particle cluster algorithm |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106022463A true CN106022463A (en) | 2016-10-12 |
| CN106022463B CN106022463B (en) | 2018-11-13 |
Family
ID=57100318
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610314420.6A Active CN106022463B (en) | 2016-05-13 | 2016-05-13 | Based on the individualized learning method for optimizing route for improving particle cluster algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106022463B (en) |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106596462A (en) * | 2016-12-19 | 2017-04-26 | 中国矿业大学 | Paper sheet quantitative detection method based on terahertz phase shifting characteristics and particle swarm algorithm |
| CN106886843A (en) * | 2017-03-24 | 2017-06-23 | 哈尔滨工业大学 | Based on the digital microcurrent-controlled failure of chip detection method and system of improving particle cluster algorithm |
| CN107230174A (en) * | 2017-06-13 | 2017-10-03 | 深圳市鹰硕技术有限公司 | A kind of network online interaction learning system and method |
| CN107346458A (en) * | 2017-05-10 | 2017-11-14 | 西北工业大学 | A kind of unmanned plane paths planning method and its device |
| CN107358325A (en) * | 2017-07-19 | 2017-11-17 | 武汉轻工大学 | A kind of Location Selection of Logistics Distribution Center method, apparatus and computer-readable recording medium |
| CN107944623A (en) * | 2017-11-22 | 2018-04-20 | 哈尔滨工业大学 | A kind of optimization method and its application based on saccharomycete budding breeding |
| CN107992051A (en) * | 2017-12-26 | 2018-05-04 | 江南大学 | Unmanned vehicle paths planning method based on improved multi-objective particle swarm algorithm |
| CN108172047A (en) * | 2018-01-19 | 2018-06-15 | 上海理工大学 | A kind of network on-line study individualized resource real-time recommendation method |
| CN109508779A (en) * | 2018-11-09 | 2019-03-22 | 重庆化工职业学院 | A kind of energy-saving control method of municipal road lamp |
| CN110569425A (en) * | 2018-06-06 | 2019-12-13 | 微软技术许可有限责任公司 | generating customized learning paths |
| CN111581728A (en) * | 2020-05-13 | 2020-08-25 | 山东女子学院 | Adaptive value estimation method based on adaptive information granulation |
| CN112529155A (en) * | 2020-12-07 | 2021-03-19 | 华中师范大学 | Dynamic knowledge mastering modeling method, modeling system, storage medium and processing terminal |
| CN112907188A (en) * | 2021-03-12 | 2021-06-04 | 北京化工大学 | Shared bicycle carrying optimization method based on adaptive neighborhood search algorithm |
| CN113012449A (en) * | 2021-03-11 | 2021-06-22 | 华南理工大学 | Smart city signal lamp timing optimization method based on multi-sample learning particle swarm |
| CN117119469A (en) * | 2023-10-18 | 2023-11-24 | 国网江苏省电力有限公司信息通信分公司 | Electric power wireless private network base station address selection method and device based on improved particle swarm optimization |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101859100A (en) * | 2010-06-18 | 2010-10-13 | 杭州电子科技大学 | Improved particle swarm optimization method based on streamline production scheduling of fuzzy due date |
| CN102880667B (en) * | 2012-09-04 | 2015-05-13 | 北京航空航天大学 | Test task scheduling method based on critical paths and tabu search |
| CN103678649A (en) * | 2013-12-20 | 2014-03-26 | 上海电机学院 | Traffic path searching system and method based on cloud self-adaptation particle swarm optimization |
-
2016
- 2016-05-13 CN CN201610314420.6A patent/CN106022463B/en active Active
Cited By (25)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106596462A (en) * | 2016-12-19 | 2017-04-26 | 中国矿业大学 | Paper sheet quantitative detection method based on terahertz phase shifting characteristics and particle swarm algorithm |
| CN106596462B (en) * | 2016-12-19 | 2019-09-10 | 中国矿业大学 | Page quantification detection method based on Terahertz phase shift characteristics and particle swarm algorithm |
| CN106886843B (en) * | 2017-03-24 | 2020-08-07 | 哈尔滨工业大学 | Digital microfluidic chip fault detection method and system based on improved particle swarm optimization |
| CN106886843A (en) * | 2017-03-24 | 2017-06-23 | 哈尔滨工业大学 | Based on the digital microcurrent-controlled failure of chip detection method and system of improving particle cluster algorithm |
| CN107346458A (en) * | 2017-05-10 | 2017-11-14 | 西北工业大学 | A kind of unmanned plane paths planning method and its device |
| CN107230174A (en) * | 2017-06-13 | 2017-10-03 | 深圳市鹰硕技术有限公司 | A kind of network online interaction learning system and method |
| CN107358325A (en) * | 2017-07-19 | 2017-11-17 | 武汉轻工大学 | A kind of Location Selection of Logistics Distribution Center method, apparatus and computer-readable recording medium |
| CN107358325B (en) * | 2017-07-19 | 2021-01-22 | 武汉轻工大学 | Logistics distribution center site selection method and device and computer readable storage medium |
| CN107944623A (en) * | 2017-11-22 | 2018-04-20 | 哈尔滨工业大学 | A kind of optimization method and its application based on saccharomycete budding breeding |
| CN107944623B (en) * | 2017-11-22 | 2021-08-31 | 哈尔滨工业大学 | Airplane fleet retention rate optimization method based on yeast budding propagation optimization |
| CN107992051A (en) * | 2017-12-26 | 2018-05-04 | 江南大学 | Unmanned vehicle paths planning method based on improved multi-objective particle swarm algorithm |
| CN107992051B (en) * | 2017-12-26 | 2020-10-23 | 江南大学 | Unmanned vehicle route planning method based on improved multi-target particle swarm algorithm |
| CN108172047B (en) * | 2018-01-19 | 2019-11-01 | 上海理工大学 | A kind of network on-line study individualized resource real-time recommendation method |
| CN108172047A (en) * | 2018-01-19 | 2018-06-15 | 上海理工大学 | A kind of network on-line study individualized resource real-time recommendation method |
| CN110569425A (en) * | 2018-06-06 | 2019-12-13 | 微软技术许可有限责任公司 | generating customized learning paths |
| CN109508779A (en) * | 2018-11-09 | 2019-03-22 | 重庆化工职业学院 | A kind of energy-saving control method of municipal road lamp |
| CN109508779B (en) * | 2018-11-09 | 2023-10-13 | 重庆化工职业学院 | Energy-saving control method for municipal street lamp |
| CN111581728A (en) * | 2020-05-13 | 2020-08-25 | 山东女子学院 | Adaptive value estimation method based on adaptive information granulation |
| CN112529155A (en) * | 2020-12-07 | 2021-03-19 | 华中师范大学 | Dynamic knowledge mastering modeling method, modeling system, storage medium and processing terminal |
| CN112529155B (en) * | 2020-12-07 | 2024-04-05 | 华中师范大学 | Dynamic knowledge mastering modeling method, modeling system, storage medium and processing terminal |
| CN113012449B (en) * | 2021-03-11 | 2022-03-29 | 华南理工大学 | Smart city signal lamp timing optimization method based on multi-sample learning particle swarm |
| CN113012449A (en) * | 2021-03-11 | 2021-06-22 | 华南理工大学 | Smart city signal lamp timing optimization method based on multi-sample learning particle swarm |
| CN112907188A (en) * | 2021-03-12 | 2021-06-04 | 北京化工大学 | Shared bicycle carrying optimization method based on adaptive neighborhood search algorithm |
| CN117119469A (en) * | 2023-10-18 | 2023-11-24 | 国网江苏省电力有限公司信息通信分公司 | Electric power wireless private network base station address selection method and device based on improved particle swarm optimization |
| CN117119469B (en) * | 2023-10-18 | 2024-01-05 | 国网江苏省电力有限公司信息通信分公司 | Electric power wireless private network base station address selection method and device based on improved particle swarm optimization |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106022463B (en) | 2018-11-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106022463A (en) | Personalized learning path optimization method based on improved particle swarm optimization algorithm | |
| Chen et al. | Distributed individuals for multiple peaks: A novel differential evolution for multimodal optimization problems | |
| CN106205126B (en) | Large-scale Traffic Network congestion prediction technique and device based on convolutional neural networks | |
| CN104636801B (en) | A kind of prediction transmission line of electricity audible noise method based on Optimized BP Neural Network | |
| CN106600065B (en) | Method and system for extracting and splicing personalized learning path based on directed hypergraph | |
| CN110032782A (en) | A kind of City-level intelligent traffic signal control system and method | |
| CN109945881A (en) | A kind of method for planning path for mobile robot of ant group algorithm | |
| CN106228183A (en) | A kind of semi-supervised learning sorting technique and device | |
| Lin et al. | Machine learning templates for QCD factorization in the search for physics beyond the standard model | |
| CN106934722A (en) | Multi-objective community detection method based on k node updates Yu similarity matrix | |
| CN111709186A (en) | Integrated estimation method for health state of retired power lithium battery | |
| CN107705556A (en) | A kind of traffic flow forecasting method combined based on SVMs and BP neural network | |
| CN104200096B (en) | Arrester grading ring optimization based on differential evolution algorithm and BP neural network | |
| CN110444022A (en) | The construction method and device of traffic flow data analysis model | |
| CN110807509A (en) | Depth knowledge tracking method based on Bayesian neural network | |
| CN106777449A (en) | Distribution Network Reconfiguration based on binary particle swarm algorithm | |
| CN107528649A (en) | A kind of Forecasting Methodology of the cognitive radio networks frequency spectrum based on Bagging | |
| WO2022257190A1 (en) | Quantum walk-based multi-feature simulation method for behavior trajectory sequences | |
| CN103984721A (en) | Personalized book searching method based on interactive evolutionary optimization | |
| CN104778338A (en) | Optimization method for set value of low-energy-consumption molecular distillation process | |
| CN105160097B (en) | A kind of three value FPRM circuit area optimization methods of utilization Population Migration Algorithm | |
| CN109447312A (en) | Route planning method, device, electronic equipment and readable storage medium storing program for executing | |
| CN107807919A (en) | A kind of method for carrying out microblog emotional classification prediction using random walk network is circulated | |
| CN106326188A (en) | Task division system based on backward learning of radius particle swarm optimization and method thereof | |
| CN109840551A (en) | A method of the optimization random forest parameter for machine learning model training |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |