CN109284860A - A kind of prediction technique based on orthogonal reversed cup ascidian optimization algorithm - Google Patents

A kind of prediction technique based on orthogonal reversed cup ascidian optimization algorithm Download PDF

Info

Publication number
CN109284860A
CN109284860A CN201810989716.7A CN201810989716A CN109284860A CN 109284860 A CN109284860 A CN 109284860A CN 201810989716 A CN201810989716 A CN 201810989716A CN 109284860 A CN109284860 A CN 109284860A
Authority
CN
China
Prior art keywords
ascidian
cup
cup ascidian
value
individual
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.)
Pending
Application number
CN201810989716.7A
Other languages
Chinese (zh)
Inventor
焦珊
陈慧灵
徐粤婷
罗杰
张谦
陈昊
赵学华
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Wenzhou University
Original Assignee
Wenzhou University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Wenzhou University filed Critical Wenzhou University
Priority to CN201810989716.7A priority Critical patent/CN109284860A/en
Publication of CN109284860A publication Critical patent/CN109284860A/en
Pending legal-status Critical Current

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]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Strategic Management (AREA)
  • Human Resources & Organizations (AREA)
  • Economics (AREA)
  • Computational Linguistics (AREA)
  • Development Economics (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Game Theory and Decision Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Marketing (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The present invention provides a kind of prediction technique based on orthogonal reversed cup ascidian optimization algorithm, including is loaded into data set and carries out standard processing to sample data；Penalty coefficient and core width are optimized using orthogonal reversed cup ascidian optimization algorithm, and construct prediction supporting vector machine model.Implement the present invention, it is wide to optimize penalty factor and core by orthogonal study and backward learning strategy on the basis of traditional cup ascidian optimization algorithm, it can effectively reduce a possibility that falling into local optimum in its searching process, be all significantly improved in the quality and convergence efficiency of solution.

Description

A kind of prediction technique based on orthogonal reversed cup ascidian optimization algorithm
Technical field
The present invention relates to field of computer technology more particularly to a kind of predictions based on orthogonal reversed cup ascidian optimization algorithm Method.
Background technique
Grid search and gradient decline are the most common two kinds of parameter optimization methods of support vector machines (SVM) at present.Grid Search is a kind of exhaustive search method, it is empty to specified parameter generally by reasonable section bound and interval steps are arranged Between divided, then to each grid node represent parameter combination be trained and predict, will be taken in these prediction results It is worth optimal parameter of the highest one group of parameter as final SVM model.Although this exhaustive-search method is to a certain extent It can guarantee to obtain parameter combination optimal in given parameters space, increase however as parameter space, search efficiency can be significantly It reduces, reasonable section is especially set and interval steps value is often extremely difficult, thus greatly reduces its feasibility, and Model is also easily trapped into local optimum very much.One major defect of gradient descent method is that it is very sensitive to initial value, When initial parameter setting is very remote from optimal solution, model is easy to converge to locally optimal solution.
In recent years, the searching algorithm based on meta-heuristic receives academic and work by its unique global optimizing ability The extensive concern of industry, they are generally considered to have than traditional optimization method there is a greater chance that finding globally optimal solution.At present Itd is proposed it is a variety of based on the SVM training algorithm of meta-heuristic algorithm come processing parameter optimization problem.
Because SVM is in specific application, performance is mainly influenced by kernel function and free parameter.Common core letter in SVM Number generally comprises linear kernel function, Polynomial kernel function, radial base (RBF) kernel function and sigmoid kernel function etc..Actually answering In, the SVM based on RBF kernel function is selected under normal circumstances.RBF core SVM relates generally to two important parameter C and γ.C is Penalty factor, it is used to control the degree to error sample punishment, plays and balance between controlled training error and model complexity Effect；C value is smaller, then also smaller to the punishment for judging sample in data by accident, so that training error becomes larger, therefore structure risk Become larger.On the contrary, C value is bigger, and it is bigger to the degree of restraint of error sample, although will lead to model in this way to the mistake of training data Sentence that rate is very low, but whole generalization ability is very poor, is easy to appear " over-fitting " phenomenon.Its selection is usually by specifically asking Depending on topic, and the quantity of noise in data is depended on, at least there is a suitable C value in each sample space, so that SVM Generalization ability is most strong.The core that second parameter γ is represented in RBF kernel function is wide, it determines the width of kernel function, directly affects The performance of SVM.Because being one-to-one between kernel function, mapping function and feature space three, it is determined that kernel function is also Mean that mapping function and feature space has impliedly been determined.The change of nuclear parameter impliedly changes mapping function, and then sample The complexity of eigen spatial distribution is also changed.For a particular problem, if γ acquirement is inappropriate, SVM is difficult to obtain expected learning effect.γ value is too small to will lead to over-fitting, and the too big discriminant function that can make SVM of γ value is excessively Gently.So penalty factor and the wide γ of core affect the Optimal Separating Hyperplane of SVM from different angles.In practical applications, they Value is excessive or the too small Generalization Capability that can all make SVM is deteriorated.
Compared to trellis search method, cup ascidian optimization algorithm has better ability of searching optimum, certainly by simulation The operational capabilities and foraging behavior of marine organisms cup ascidian in right boundary obtain the approximate optimal solution of problem, cup ascidian optimization algorithm It is strong with global optimizing ability, the high feature of low optimization accuracy.But the algorithm be directed to challenge when, can also fall into local pole A possibility that value.
Summary of the invention
The technical problem to be solved by the embodiment of the invention is that providing a kind of based on orthogonal reversed cup ascidian optimization algorithm Method to construct prediction model, on the basis of traditional cup ascidian optimization algorithm by it is orthogonal study and backward learning strategy come Optimize penalty factor and core is wide, can effectively reduce a possibility that falling into local optimum in its searching process, compared with other optimization algorithms It is all significantly improved in the quality and convergence efficiency of solution.
In order to solve the above-mentioned technical problem, the embodiment of the invention provides one kind to be based on orthogonal reversed cup ascidian optimization algorithm Method to construct prediction model, the described method comprises the following steps:
Step S1: parameter initialization；Wherein, the parameter of initialization includes: maximum number of iterations T, cup sea in cup ascidian chain Search space [the C of number N, C of sheathmin,Cmax] and γ search space [γmin,γmax]；
Step S2: cup ascidian population position initialization: being randomly generated the position of N number of cup ascidian, wherein i-th cup ascidian Position is Xi=(xi1,xi2), i=1,2 ... ..., N；Wherein, xi1Indicate C value of cup ascidian i at current location, xi2Indicate cup γ value of the ascidian i at current location；
Step S3: backward learning is carried out to initialization population.Firstly, to cup ascidian individual X being randomly generatediIt carries out general Backward learning obtains its reversed solution X according to formula (1)i'=(xi1',xi2'), thus obtain 2N cup ascidian；Then, to 2N Cup ascidian calculates its fitness fi, which is C the and γ value based on the current location cup ascidian individual i, is rolled over and is handed over internal K The accuracy ACC that authentication policy calculates support vector machines is pitched, and using the value as the fitness f of cup ascidian iiValue, will adapt to Spend fiFitness value is come as new population by descending sequence, cup ascidian individual for taking fitness value to come top N The fitness of one cup ascidian saves as vector FoodFitness, and using its position as the position of food, saves as vector FoodPosition；Finally, calculating the bat ACC for obtaining and obtaining based on K folding cross validation according to formula (2)；
xij'=λ (aj+bj)-xij(1)；
In formula (1) and (2), aj=min (x1j,x2j,...,xNj), bj=max (x1j,x2j,...,xNj), xij'∈[aj, bj], λ ∈ (0,1) is random number；acckIndicate each broken number according to the upper accuracy for calculating and obtaining；
Step S4: it takes with after C and γ coding mode identical in step S3, cross validation policy calculation is rolled over internal K The fitness of each cup ascidian；
Step S5: the position of cup ascidian individual i is updated according to formula (3)-(5).Firstly, according to formula (3) undated parameter c1；Then, the individual in cup ascidian chain is divided into two parts of leader and follower, cup ascidian of first half is considered as leader Person simultaneously carries out location updating according to formula (4), and latter half individual is then used as follower and carries out position more according to formula (5) Newly；
In formula (3)-(5), t is current iteration number, and T is maximum number of iterations；FoodPositionjFor food position, I.e. current optimal solution, ubjAnd lbjDifferent values according to j are respectively the maximum of C and γ, minimum value, c2、c3It is then random number；
Step S6: orthogonal study is carried out to cup ascidian.Firstly, the fitness value of cup ascidian i is calculated according to step S3, if should Fitness value is greater than FoodFitness, then is constructed on the basis of cup ascidian i and FoodPosition according to formula (6) new Cup ascidian individual TSalpPosition；Then, to the study orthogonal with cup ascidian i progress of new cup ascidian of construction.In orthogonal In habit, number of levels Q is initialized first and because of prime number F, then generates orthogonal arrage LM(QF), M candidate is generated according to formula (7) Then body carries out factor level analysis construction optimal level factor combination, is selected according to fitness optimal in M+1 candidate solution BestPosition is solved, optimal solution and its fitness value BestFitness are returned to, by the adaptation of BestFitness and cup ascidian i Degree is made comparisons, if BestFitness is greater than the fitness value of cup ascidian i, by XiReplace with BestPosition；
TSalpPosition=Xi+r1(FoodPosition-Xi)+r2(FoodPosition-Xk) (6)；
In formula (6)-(7), r1、r2For random number, XkIt is to be different from XiRandom individual；L (i, j) is orthogonal arrage LM(QF) in Value；M is the number for new cup ascidian that orthogonal experiment generates, referred to as candidate solution；
Step S7: backward learning is carried out again to updated cup ascidian.According to step S3, to cup sea behind update position Sheath carries out backward learning and obtains N number of reversed solution, similarly sorts from large to small to the fitness value of 2N cup ascidian individual, before taking Individual is assigned to FoodPosition as the individual in cup ascidian chain, and by cup ascidian position that fitness value ranked first；
Step S8: judge whether to reach maximum number of iterations T, rapid S9 is performed the next step if having reached, is otherwise jumped to Step S4；
Step S9: the position FoodPosition of food, i.e., optimal penalty coefficient C and the wide γ value of core are exported；
Step S10: by the step S9 optimal penalty coefficient C obtained the and wide γ of core, for constructing optimal classification function formula (8) to predict supporting vector machine model；
In formula (8), K (xi,xj)=exp (- r | | xi-xj||2), αiFor Lagrange coefficient, b is threshold value, xiIt is to be tested Sample (i=1,2 ... ..., n), n are individual of sample number, yiIndicate label corresponding with training sample, yj(j=1,2 ..., N) value is 1 and -1, wherein 1 indicates that current individual of sample is positive class sample, -1 current individual of sample of expression is negative class sample.
The implementation of the embodiments of the present invention has the following beneficial effects:
Present invention fusion obtains one kind efficiently accurately based on orthogonal reversed cup ascidian optimization algorithm and support vector machines Intelligent classification, prediction model cross cup ascidian optimization algorithm Support Vector Machines Optimized based on orthogonal study and backward learning strategy Key parameter, that is, penalty coefficient C and the wide γ of core, introduces orthogonal study and backward learning in optimization process, so that algorithm Optimal value can be fast and effeciently found, and backward learning increases multifarious feature and falls into local extremum to prevent, and can have Effect reduces a possibility that falling into local optimum in its searching process, and quality and convergence efficiency compared with other optimization algorithms in solution have It is apparent to improve.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this Some embodiments of invention, for those of ordinary skill in the art, without any creative labor, according to These attached drawings obtain other attached drawings and still fall within scope of the invention.
Fig. 1 is the method provided in an embodiment of the present invention for constructing prediction model based on orthogonal reversed cup ascidian optimization algorithm Flow chart.
Specific embodiment
To make the object, technical solutions and advantages of the present invention clearer, the present invention is made into one below in conjunction with attached drawing Step ground detailed description.
As shown in Figure 1, one kind of proposition is constructed based on orthogonal reversed cup ascidian optimization algorithm in the embodiment of the present invention The method of prediction model, the described method comprises the following steps:
Step S1: parameter initialization；Wherein, the parameter of initialization includes: maximum number of iterations T, cup sea in cup ascidian chain Search space [the C of number N, C of sheathmin,Cmax] and γ search space [γmin,γmax]；
Step S2: cup ascidian population position initialization: being randomly generated the position of N number of cup ascidian, wherein i-th cup ascidian Position is Xi=(xi1,xi2), i=1,2 ... ..., N；Wherein, xi1Indicate C value of cup ascidian i at current location, xi2Indicate cup γ value of the ascidian i at current location；
Step S3: backward learning is carried out to initialization population.Firstly, to cup ascidian individual X being randomly generatediIt carries out general Backward learning obtains its reversed solution X according to formula (1)i'=(xi1',xi2'), thus obtain 2N cup ascidian；Then, to 2N Cup ascidian calculates its fitness fi, which is C the and γ value based on the current location cup ascidian individual i, is rolled over and is handed over internal K The accuracy ACC that authentication policy calculates support vector machines is pitched, and using the value as the fitness f of cup ascidian iiValue, will adapt to Spend fiFitness value is come as new population by descending sequence, cup ascidian individual for taking fitness value to come top N The fitness of one cup ascidian saves as vector FoodFitness, and using its position as the position of food, saves as vector FoodPosition；Finally, calculating the bat ACC for obtaining and obtaining based on K folding cross validation according to formula (2)；
xij'=λ (aj+bj)-xij(1)；
In formula (1) and (2), aj=min (x1j,x2j,...,xNj), bj=max (x1j,x2j,...,xNj), xij'∈[aj, bj], λ ∈ (0,1) is random number；acckIndicate each broken number according to the upper accuracy for calculating and obtaining；
Step S4: it takes with after C and γ coding mode identical in step S3, cross validation policy calculation is rolled over internal K The fitness of each cup ascidian；
Step S5: the position of cup ascidian individual i is updated according to formula (3)-(5).Firstly, according to formula (3) undated parameter c1；Then, the individual in cup ascidian chain is divided into two parts of leader and follower, cup ascidian of first half is considered as leader Person simultaneously carries out location updating according to formula (4), and latter half individual is then used as follower and carries out position more according to formula (5) Newly；
In formula (3)-(5), t is current iteration number, and T is maximum number of iterations；FoodPositionjFor food position, I.e. current optimal solution, ubjAnd lbjDifferent values according to j are respectively the maximum of C and γ, minimum value, c2、c3It is then random number；
Step S6: orthogonal study is carried out to cup ascidian.Firstly, the fitness value of cup ascidian i is calculated according to step S3, if should Fitness value is greater than FoodFitness, then is constructed on the basis of cup ascidian i and FoodPosition according to formula (6) new Cup ascidian individual TSalpPosition；Then, to the study orthogonal with cup ascidian i progress of new cup ascidian of construction.In orthogonal In habit, number of levels Q is initialized first and because of prime number F, then generates orthogonal arrage LM(QF), M candidate is generated according to formula (7) Then body carries out factor level analysis construction optimal level factor combination, is selected according to fitness optimal in M+1 candidate solution BestPosition is solved, optimal solution and its fitness value BestFitness are returned to, by the adaptation of BestFitness and cup ascidian i Degree is made comparisons, if BestFitness is greater than the fitness value of cup ascidian i, by XiReplace with BestPosition；
TSalpPosition=Xi+r1(FoodPosition-Xi)+r2(FoodPosition-Xk) (6)；
In formula (6)-(7), r1、r2For random number, XkIt is to be different from XiRandom individual；L (i, j) is orthogonal arrage LM(QF) in Value；M is the number for new cup ascidian that orthogonal experiment generates, referred to as candidate solution；
Step S7: backward learning is carried out again to updated cup ascidian.According to step S3, to cup sea behind update position Sheath carries out backward learning and obtains N number of reversed solution, similarly sorts from large to small to the fitness value of 2N cup ascidian individual, before taking Individual is assigned to FoodPosition as the individual in cup ascidian chain, and by cup ascidian position that fitness value ranked first；
Step S8: judge whether to reach maximum number of iterations T, rapid S9 is performed the next step if having reached, is otherwise jumped to Step S4；
Step S9: the position FoodPosition of food, i.e., optimal penalty coefficient C and the wide γ value of core are exported
Step S10: by the step S9 optimal penalty coefficient C obtained the and wide γ of core, for constructing optimal classification function formula (8) to predict supporting vector machine model；
In formula (8), K (xi,xj)=exp (- r | | xi-xj||2), αiFor Lagrange coefficient, b is threshold value, xiIt is to be tested Sample (i=1,2 ... ..., n), n are individual of sample number, yiIndicate label corresponding with training sample, yj(j=1,2 ..., N) value is 1 and -1, wherein 1 indicates that current individual of sample is positive class sample, -1 current individual of sample of expression is negative class sample.
In embodiments of the present invention, answering to the method for constructing prediction model based on orthogonal reversed cup ascidian optimization algorithm It is described further with scene:
Using breast cancer data as sample data, sample set indicates in this way: (xi,yi), i=1,2 ..., 699, In ' xi' indicate the feature vectors of 9 dimensions, y be value for 1 or -1 sample label, ' 1 ' to represent the sample be to suffer from breast cancer, ' -1 ' generation The table patient is healthy.
Firstly, will be standardized to each characteristic attribute value of experiment sample data, formula is utilized Sample data is standardized, wherein SiThe feature original value of attribute in representative sample, Si' it is SiIt is obtained by formula Value after standardization, SminIndicate the minimum value in corresponding sample data, SmaxIndicate the maximum value in corresponding sample data；
Then, the punishment system based on orthogonal study and cup ascidian optimization algorithm Support Vector Machines Optimized of backward learning is utilized The number C and wide γ of core, and (sample for importing model is subjected to K folding cutting, often in internal optimize using K folding Crossover Strategy Once all using K-1 therein roll over as training data, and training while use orthogonal reversed cup ascidian optimization algorithm for The critical parameter of two of them optimizes, it is expected that obtain optimal intelligent classification model, after model construction is good, with residue Data as test data, the performance of the intelligent decision making model of building is assessed).It is simply that for not Same intelligent classification decision problem, it would be desirable to which reality is gone using orthogonal reversed cup ascidian optimization algorithm with ability of searching optimum It now constructs for the optimal categorised decision model of problems, certainly as previously discussed: the penalty coefficient C and wide γ of core is to this The performance of model has important influence, that is to say, that the quality of the two parameters will directly affect the performance of decision model Quality goes to complete the selection to the two parameters, to a certain extent so herein proposing orthogonal reversed cup ascidian optimization algorithm Improve convergence speed of the algorithm and precision.
Input training sample (xi,yi), and become according to the problem of Lagrange dual problem optimization:
It then, (is radial base core letter to C and γ using orthogonal reversed cup ascidian optimization algorithm for above optimization problem The number wide K (x of parameter corei,xj)=exp (- γ | | xi-xj||2)) optimize, and solve the value of optimal solution are as follows: a*=(a1 *, a2 *,...,a* 699)T；
Then there is following solution:So final optimal separating hyper plane function are as follows:
The implementation of the embodiments of the present invention has the following beneficial effects:
Present invention fusion obtains one kind efficiently accurately based on orthogonal reversed cup ascidian optimization algorithm and support vector machines Intelligent classification, prediction model optimize supporting vector by cup ascidian optimization algorithm based on orthogonal study and backward learning strategy The key parameter of the machine, that is, penalty coefficient C and wide γ of core, introduces orthogonal study and backward learning, so that algorithm in optimization process Optimal value can be fast and effeciently found, and backward learning increases multifarious feature and falls into local extremum to prevent, and can have Effect reduces a possibility that falling into local optimum in its searching process, and quality and convergence efficiency compared with other optimization algorithms in solution have It is apparent to improve.
Those of ordinary skill in the art will appreciate that implement the method for the above embodiments be can be with Relevant hardware is instructed to complete by program, the program can be stored in a computer readable storage medium, The storage medium, such as ROM/RAM, disk, CD.
Above disclosed is only a preferred embodiment of the present invention, cannot limit the power of the present invention with this certainly Sharp range, therefore equivalent changes made in accordance with the claims of the present invention, are still within the scope of the present invention.

Claims (1)

1. a kind of prediction technique based on orthogonal reversed cup ascidian optimization algorithm, which is characterized in that the method includes following steps It is rapid:
Step S1: parameter initialization；Wherein, the parameter of initialization includes: maximum number of iterations T, cup ascidian in cup ascidian chain Search space [the C of number N, Cmin,Cmax] and γ search space [γmin,γmax]；
Step S2: cup ascidian population position initialization: being randomly generated the position of N number of cup ascidian, wherein the position of i-th of cup ascidian For Xi=(xi1,xi2), i=1,2 ... ..., N；Wherein, xi1Indicate C value of cup ascidian i at current location, xi2Indicate cup ascidian γ value of the i at current location；
Step S3: backward learning is carried out to initialization population, firstly, to cup ascidian individual X being randomly generatediIt carries out general reversed Study obtains its reversed solution X according to formula (1)i'=(xi1',xi2'), thus obtain 2N cup ascidian；Then, to 2N cup sea Sheath calculates its fitness fi, which is C the and γ value based on the current location cup ascidian individual i, is intersected with internal K folding and is tested The accuracy ACC of policy calculation support vector machines is demonstrate,proved, and using the value as the fitness f of cup ascidian iiValue, by fitness fiBy Small sequence is arrived greatly, and fitness value is come first cup as new population by cup ascidian individual for taking fitness value to come top N The fitness of ascidian saves as vector FoodFitness, and using its position as the position of food, saves as vector FoodPosition；Finally, calculating the bat ACC for obtaining and obtaining based on K folding cross validation according to formula (2)；
xij'=λ (aj+bj)-xij(1)；
In formula (1) and (2), aj=min (x1j,x2j,...,xNj), bj=max (x1j,x2j,...,xNj), xij'∈[aj,bj], λ ∈ (0,1) is random number；acckIndicate each broken number according to the upper accuracy for calculating and obtaining；
Step S4: it takes with after C and γ coding mode identical in step S3, it is each that cross validation policy calculation is rolled over internal K The fitness of a cup ascidian；
Step S5: the position of cup ascidian individual i is updated according to formula (3)-(5), firstly, according to formula (3) undated parameter c1；So Afterwards, the individual in cup ascidian chain is divided into two parts of leader and follower, cup ascidian of first half is considered as leader simultaneously Location updating is carried out according to formula (4), latter half individual is then used as follower and carries out location updating according to formula (5)；
In formula (3)-(5), t is current iteration number, and T is maximum number of iterations；FoodPositionjFor food position, i.e., currently Optimal solution, ubjAnd lbjDifferent values according to j are respectively the maximum of C and γ, minimum value, c2、c3It is then random number；
Step S6: orthogonal study is carried out to cup ascidian, firstly, the fitness value of cup ascidian i is calculated according to step S3, if the adaptation Angle value is greater than FoodFitness, then constructs new cup on the basis of cup ascidian i and FoodPosition according to formula (6) Ascidian individual TSalpPosition；Then, to the study orthogonal with cup ascidian i progress of new cup ascidian of construction, in orthogonal study In, number of levels Q is initialized first and because of prime number F, then generates orthogonal arrage LM(QF), M candidate individual is generated according to formula (7), Then factor level analysis construction optimal level factor combination is carried out, the optimal solution in M+1 candidate solution is selected according to fitness BestPosition returns to optimal solution and its fitness value BestFitness, by the fitness of BestFitness and cup ascidian i It makes comparisons, if BestFitness is greater than the fitness value of cup ascidian i, by XiReplace with BestPosition；
TSalpPosition=Xi+r1(FoodPosition-Xi)+r2(FoodPosition-Xk) (6)；
In formula (6)-(7), r1、r2For random number, XkIt is to be different from XiRandom individual；L (i, j) is orthogonal arrage LM(QF) in Value；M is the number for new cup ascidian that orthogonal experiment generates, referred to as candidate solution；
Step S7: carrying out backward learning to updated cup ascidian again, according to step S3, to update cup ascidian behind position into Row backward learning obtains N number of reversed solution, similarly sorts from large to small to the fitness value of 2N cup ascidian individual, takes top n Individual is assigned to FoodPosition as the individual in cup ascidian chain, and by cup ascidian position that fitness value ranked first；
Step S8: judge whether to reach maximum number of iterations T, rapid S9 is performed the next step if having reached, otherwise jumps to step S4；
Step S9: the position FoodPosition of food, i.e., optimal penalty coefficient C and the wide γ value of core are exported；
Step S10: by the step S9 optimal penalty coefficient C obtained the and wide γ of core, for construct optimal classification function formula (8) with Predict supporting vector machine model；
In formula (8), K (xi,xj)=exp (- r | | xi-xj||2), αiFor Lagrange coefficient, b is threshold value, xiFor sample to be tested (i=1,2 ... ..., n), n are individual of sample number, yiIndicate label corresponding with training sample, yj(j=1,2 ..., n) it takes Value is 1 and -1, wherein 1 indicates that current individual of sample is positive class sample, -1 current individual of sample of expression is negative class sample.
CN201810989716.7A 2018-08-28 2018-08-28 A kind of prediction technique based on orthogonal reversed cup ascidian optimization algorithm Pending CN109284860A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810989716.7A CN109284860A (en) 2018-08-28 2018-08-28 A kind of prediction technique based on orthogonal reversed cup ascidian optimization algorithm

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810989716.7A CN109284860A (en) 2018-08-28 2018-08-28 A kind of prediction technique based on orthogonal reversed cup ascidian optimization algorithm

Publications (1)

Publication Number Publication Date
CN109284860A true CN109284860A (en) 2019-01-29

Family

ID=65183562

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810989716.7A Pending CN109284860A (en) 2018-08-28 2018-08-28 A kind of prediction technique based on orthogonal reversed cup ascidian optimization algorithm

Country Status (1)

Country Link
CN (1) CN109284860A (en)

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111047528A (en) * 2019-11-27 2020-04-21 天津大学 High-spectrum image unmixing method based on goblet sea squirt group
CN111126549A (en) * 2019-12-24 2020-05-08 昆明理工大学 Double-star spectrum fitting method based on strategy improved goblet and sea squirt intelligent algorithm
CN111383288A (en) * 2020-03-02 2020-07-07 东北林业大学 Goblet ascidian-adaptive differential evolution hybrid camera internal parameter optimization algorithm
CN112085059A (en) * 2020-08-06 2020-12-15 温州大学 Breast cancer image feature selection method based on improved sine and cosine optimization algorithm
CN112115969A (en) * 2020-08-11 2020-12-22 温州大学 Method and device for optimizing FKNN model parameters based on variant goblet sea squirt group algorithm
CN112215259A (en) * 2020-09-17 2021-01-12 温州大学 Gene selection method and apparatus
CN114398762A (en) * 2021-12-22 2022-04-26 淮阴工学院 Fitting method and device of high-precision energy storage element model based on goblet sea squirt algorithm
CN115014617A (en) * 2022-06-21 2022-09-06 福州大学 Cable force synchronous monitoring method for cable-stayed bridge inhaul cable based on ground-based radar

Cited By (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111047528A (en) * 2019-11-27 2020-04-21 天津大学 High-spectrum image unmixing method based on goblet sea squirt group
CN111047528B (en) * 2019-11-27 2023-07-07 天津大学 Hyperspectral image unmixing method based on goblet sea squirt group
CN111126549A (en) * 2019-12-24 2020-05-08 昆明理工大学 Double-star spectrum fitting method based on strategy improved goblet and sea squirt intelligent algorithm
CN111383288A (en) * 2020-03-02 2020-07-07 东北林业大学 Goblet ascidian-adaptive differential evolution hybrid camera internal parameter optimization algorithm
CN111383288B (en) * 2020-03-02 2022-11-11 东北林业大学 Goblet ascidian-adaptive differential evolution hybrid camera internal parameter optimization algorithm
CN112085059B (en) * 2020-08-06 2023-10-20 温州大学 Breast cancer image feature selection method based on improved sine and cosine optimization algorithm
CN112085059A (en) * 2020-08-06 2020-12-15 温州大学 Breast cancer image feature selection method based on improved sine and cosine optimization algorithm
CN112115969A (en) * 2020-08-11 2020-12-22 温州大学 Method and device for optimizing FKNN model parameters based on variant goblet sea squirt group algorithm
CN112115969B (en) * 2020-08-11 2023-11-17 温州大学 Method and device for optimizing FKNN model parameters based on variant sea squirt swarm algorithm
CN112215259A (en) * 2020-09-17 2021-01-12 温州大学 Gene selection method and apparatus
CN112215259B (en) * 2020-09-17 2023-12-01 温州大学 Gene selection method and apparatus
CN114398762A (en) * 2021-12-22 2022-04-26 淮阴工学院 Fitting method and device of high-precision energy storage element model based on goblet sea squirt algorithm
CN115014617B (en) * 2022-06-21 2023-04-07 福州大学 Cable-stayed bridge cable force synchronous monitoring method based on ground radar
CN115014617A (en) * 2022-06-21 2022-09-06 福州大学 Cable force synchronous monitoring method for cable-stayed bridge inhaul cable based on ground-based radar

Similar Documents

Publication Publication Date Title
CN109284860A (en) A kind of prediction technique based on orthogonal reversed cup ascidian optimization algorithm
CN110070909B (en) Deep learning-based multi-feature fusion protein function prediction method
CN105488528B (en) Neural network image classification method based on improving expert inquiry method
CN111612125B (en) Novel HTM time pool method and system for online learning
CN111047563B (en) Neural network construction method applied to medical ultrasonic image
CN106355192A (en) Support vector machine method based on chaos and grey wolf optimization
CN114841257B (en) Small sample target detection method based on self-supervision comparison constraint
CN109934330A (en) The method of prediction model is constructed based on the drosophila optimization algorithm of diversified population
KR20210032140A (en) Method and apparatus for performing pruning of neural network
CN110738362A (en) method for constructing prediction model based on improved multivariate cosmic algorithm
US11514315B2 (en) Deep neural network training method and apparatus, and computer device
CN107392919A (en) Gray threshold acquisition methods, image partition method based on self-adapted genetic algorithm
CN110222838B (en) Document sorting method and device, electronic equipment and storage medium
CN113360670B (en) Knowledge graph completion method and system based on fact context
CN110222751A (en) A method of prediction model is constructed based on orthogonal sine and cosine algorithm on multiple populations
CN111079074A (en) Method for constructing prediction model based on improved sine and cosine algorithm
CN115422369B (en) Knowledge graph completion method and device based on improved TextRank
Zhang et al. Evolving neural network classifiers and feature subset using artificial fish swarm
CN110069817A (en) A method of prediction model is constructed based on California gray whale optimization algorithm is improved
CN110991494A (en) Method for constructing prediction model based on improved moth optimization algorithm
US11461656B2 (en) Genetic programming for partial layers of a deep learning model
CN109409522B (en) Biological network reasoning algorithm based on ensemble learning
CN110751257A (en) Method for constructing prediction model based on hunger game search algorithm
CN109948675A (en) The method for constructing prediction model based on outpost's mechanism drosophila optimization algorithm on multiple populations
CN115858725A (en) Method and system for screening text noise based on unsupervised graph neural network

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination