CN104598770B - Wheat aphid quantitative forecasting technique and system based on human evolution's gene expression programming - Google Patents
Wheat aphid quantitative forecasting technique and system based on human evolution's gene expression programming Download PDFInfo
- Publication number
- CN104598770B CN104598770B CN201510014836.1A CN201510014836A CN104598770B CN 104598770 B CN104598770 B CN 104598770B CN 201510014836 A CN201510014836 A CN 201510014836A CN 104598770 B CN104598770 B CN 104598770B
- Authority
- CN
- China
- Prior art keywords
- population
- individual
- gene
- module
- individuality
- 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.)
- Expired - Fee Related
Links
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses wheat aphid quantitative forecasting technique and system based on human evolution's gene expression programming, reparation is carried out in worst individual or infeasible individual using " roguing " operation to generation in natural evolution be translated into feasible individual, it is the Fineness gene fragment in Fineness gene storehouse is spread in population using " increasing excellent " operation, so as to effectively increase the quality of individuality in population, convergence of algorithm speed is accelerated;Intervened using population, for Premature convergence caused by being lacked due to population diversity during natural evolution, by the poor individuality for replacing equivalent amount with the very big mirror image individuality of the diversity factor of the new feasible individual for randomly generating and Mirroring Mapping generation in population, to increase the diversity of population, so as to effectively improve the global optimizing ability of algorithm.
Description
Technical field
The present invention relates to a kind of wheat Wheat Aphid Population quantitative forecasting technique and system, it is specifically related to based on human evolution's base
Because of the wheat aphid quantitative forecasting technique and system of expression formula programming.
Background technology
Small grain aphid is the Occurrence of each wheat belt of China, not only nibbles wheat nutrition, influence photosynthesis,
But also cereal virus are propagated, cause the wheat significantly underproduction and quality decline, the production of serious threat wheat.Carry out wheat long tube
The prediction that aphid population occurs be generally acknowledge both at home and abroad prevented and treated and reduced Pesticide use to small grain aphid disaster and had
Effect measure.However, small grain aphid Population breeding speed is high, generation overlap, and various small grain aphids are in different ecological
The area of wheat mixes generation, but the pests occurrence rule of different small grain aphid species, ecological habit, host range and harmful characteristics etc. are all
Difference, this is not only difficult to Accurate Prediction forecast, and huge difficulty is also brought to preventing and controlling.Therefore, one is set up
The small grain aphid Occurrence forecast mechanism of efficiently and accurately is covered, huge shadow will be produced to the validity of small grain aphid preventing and controlling
Ring, largely can also alleviate the situation that national aphid occurring area expands year by year under global warming, be people group
Crowd retrieves necessary economic loss.
At present, structure and Forecasting Methodology on small grain aphid generation model have three classes:The first is statistics prediction
Method, it is, based on probability and multiple-factor linear correlation principle, the screening of the certain limit interior prediction factor to be carried out using mathematical measure
And modeling.Statistics predicted method principle and method are simple, are easily mastered, but prediction effect is unstable.Second is information prediction,
Including management information system, expert system and GIS-Geographic Information System etc..The model structure that the method builds is complicated, parameter it is numerous and
It is assumed that constant, causes theoretical model larger with realistic model difference, reduce predictablity rate.The third is artificial intelligence side
Method, including evolution algorithm, swarm intelligence algorithm, artificial neural network, SVMs, gray theory etc., such method due to
Not strong to experience dependence, algorithm has the advantages that adaptivity, global optimizing, non-linear, the model accuracy of foundation is higher,
Generalization ability is strong, is current cutting edge technology and developing direction.But artificial intelligence technology is applied to Model of The Dynamic of Insect Populations
Foundation research and application relatively lag behind and go deep into not enough, generally there is local convergence, fitting excessive, single etc. no using algorithm
Foot.
Evolutionary computation is the random search techniques that a kind of groups is oriented to, and is very suitable for functional relation and pinpoints the problems.Gene
Expression formula programming is a kind of new evolution algorithm, and it is not having any priori, is not knowing about mechanism, only experiment inside the matters
Accurate formula can be excavated in the case of data, it is to avoid the blindness of selected type function in advance when traditional algorithm is modeled
Property, there is extraordinary mapping ability for complexity, multi input, uncertain nonlinear problem, set up in model and surveyed and function
Excavation aspect has been achieved for good achievement.Current gene expression programming has been successfully applied to numerous Practical Project necks
In domain, but the research in agricultural pest prediction field is just at the early-stage.
Gene expression programming can be instructed using the conventional climatic environment data in each department, crops and disease pest data
Practice modeling, corresponding Forecasting Pests constructed in the case where need not completely understand each factor internal correlation relation,
Worked using the insect pest preventing and controlling that the model can be crops and make guidances, it is to avoid because prevention is not enough and caused by the underproduction or excessive
The fund that prevention is caused is wasted and environmental pollution (such as excessively spraying insecticide).But find in actual applications, traditional base
When being applied to small grain aphid Prediction to population numbers because of expression formula programmed algorithm, exist that convergence rate is slow, be easily trapped into part
Optimal deficiency, so as to cause, precision of prediction is low, model generalization ability.Therefore, how gene expression programming is improved
Global optimizing ability, to accelerate convergence rate, the accuracy of the precision and its prediction that model is set up in raising be current to be badly in need of solving
Problem.
The content of the invention
For defect present in above-mentioned prior art or deficiency, it is an object of the present invention to provide being based on human evolution
The wheat aphid quantitative forecasting technique and system of gene expression programming.
To achieve these goals, the present invention uses following technical scheme:
Wheat aphid quantitative forecasting technique based on human evolution's gene expression programming, comprises the following steps:
Step 1:Gathering N number of time point investigates the field meteorological condition and population dynamic of small grain aphid, obtains
For the field investigation data that small grain aphid population quantity model is set up, such as daily maximum temperature, daily mean temperature, daily precipitation
Amount, wheat growing stage, small grain aphid natural enemy quantity, and small grain aphid total amount, as the sample data of modeling, are designated as
Matrix A;
Step 2:Initiation parameter is set, and produces initialization population, using initialization population as current population pop;Institute
The parameter stated includes population scale sizepop, maximum iteration maxgen, functor set F, full stop set T, mirror image letter
Number symbol set M, connector between mrna length, chromogene number, chromogene, fall string length, insert string length, variation generally
Rate, fall string probability, insert string probability, recombination probability, Fineness gene storehouse scale m, increase excellent probability, algorithm of tournament selection contest scale p;
Step 3:The adaptation of each individuality in the current population pop that the matrix A calculation procedure 2 obtained using step 1 is obtained
Degree, and preserve optimum individual maxpop and its fitness;
Step 4:According to each the individual fitness in the current population pop that step 3 is obtained, Fineness gene storehouse is set up
GBL and characteristic information storehouse BBL inferior;
Step 5:Judge that whether current iterations, more than maximum iteration maxgen, is to go to step 11;Otherwise
Proceed as follows:The individual natural evolution operation for performing gene expression programming in current population pop is obtained in first
Between population pop1;Calculate the individual fitness of the described first middle population pop1, at the same record wherein infeasible individual and its
The position in the individuality of gene inferior, is saved in matrix badgen;
It is further comprising the steps of:
Step 6:The functor collection F and full stop collection T obtained using step 2, in the matrix badgen obtained to step 5
Individuality in the corresponding first middle population pop1 of infeasible individual carries out " roguing " operation, by the first middle population pop1
Infeasible individual becomes feasible individual, obtains the second middle population pop2;
Step 7:The Fineness gene storehouse GBL obtained using step 4, the second middle population pop2 obtained to step 6 is performed
" increasing excellent " during individuality is intervened operates, and forms the 3rd middle population pop3;
Step 8:The functor collection F and full stop collection T obtained using step 2, the 3rd middle population that calculation procedure 7 is obtained
The comentropy H of pop3;According to comentropy H and the comparative result of the information entropy threshold of setting, the 3rd among obtained to step 7
Group pop3 performs population intervention operation, obtains the 4th middle population pop4;
Step 9:Merge the 4th middle population pop4 and Fineness gene storehouse GBL, and according to the individual adaptation degree after merging
Size, the Fineness gene storehouse GBL after being updated;
Step 10:According to current optimum individual maxpop and its fitness, the 4th middle population pop4 obtained to step 8
Carry out the algorithm of tournament selection with elitism strategy to operate, obtain the 5th middle population pop5;Iterations plus 1, and by the 5th
Middle population pop5 goes to step 5 as current population pop;
Step 11:The fitness of individuality in current population pop is calculated, the maximum individual conduct of fitness value currently kind is chosen
Optimum individual bestpop in group pop, calculates the effective length of each gene of the optimum individual bestpop, according to each
The effective length of gene travels through each gene from right to left, and decoding obtains the mathematic(al) representation of each gene, then with the phase of step 3
The individual expression formula of optimum individual bestpop is obtained with means;By the individual expression formula of described optimum individual bestpop,
According to the real-time measurement values of each parameter in expression formula, the predicted value of small grain aphid population quantity can be obtained.
Further, the step 3 specifically includes following steps:
Step 3.1:The effective length of each gene of each individuality in the current population pop that calculation procedure 2 is obtained;
Step 3.2:According to the effective length of each gene of each individuality, each gene position is traveled through from right to left, draw
The mathematic(al) representation of each gene;
Step 3.3:Each gene of s-th individuality that will be obtained in step 3.2 with connector "+" between chromogene
Expression formula is coupled together, and obtains s-th expression formula of individuality, and the data correspondence in sample data matrix A is substituted into s-th
The expression formula of body, calculates predicted values of the individual s in the expression formula of time point jS-th individuality is similarly obtained in all times
The predicted value of the expression formula of point;
Step 3.4:Each individual fitness is calculated using the multiple correlation coefficient method shown in formula (1), and preserves optimal
Body maxpop and its fitness;Fitness computing formula is:
Fitness=R2=1-SSE/SST (1) is to be adapted to using a square conduct for the multiple correlation coefficient in statistics
Degree, wherein, R is multiple correlation coefficient,
Wherein, N is the time point sum that the small grain aphid total amount of observation is chosen, yjIt is j-th time stored in matrix A
The small grain aphid amount that point is observed,It is yjAverage value.SSE is residual sum of squares (RSS), and SST is total sum of squares of deviations.It is optimal
Individual criterion is:The maximum individuality of fitness in all individualities.
Further, the step 5 specifically includes following steps:
Step 5.1:Judge that whether current iterations, more than maximum iteration maxgen, is to go to step 11;It is no
Then perform step 5.2;
Step 5.2:It is individual to each in current population pop according to the probability of each genetic operator set in step 2
Successively perform gene expression programming natural evolution operation, that is, make a variation, fall string, IS insert string, RIS insert string, single-point restructuring, 2 points
Restructuring, the operation of genetic recombination, obtain the first middle population pop1;
Step 5.3:According to each individual fitness in the first middle population pop1 of formula (1) calculating, while record is every
Position of the individual gene position inferior in individuality;
Step 5.4:The gene bit matrix inferior of infeasible individual is set up using the gene position inferior in step 5.3
badgen。
Further, the step 6 specifically includes following steps:
Step 6.1:Each gene location institute inferior of all infeasible individuals is right in the matrix badgen obtained to step 5
A functor or a full stop are randomly selected in the symbol answered, the functor collection F obtained from step 2 and full stop collection T, it is right
Symbol at described gene location carries out man-to-man replacement;
Step 6.2:The fitness of each infeasible individual after being replaced in calculation procedure 6.1, if what is obtained is individual
The value of fitness is not underrange (NaN), bears infinite (- Inf) or plural (A+Bi/A-Bi), is then replaced successfully;Otherwise, continue
Step 6.1 is performed, untill being replaced successfully;Obtain the second middle population pop2.
Further, the step 7 specifically includes following steps:
Step 7.1:Each in the second middle population pop2 obtained for step 6 is individual, if randomly generate one
Number between individual 0-1 then carries out " increasing excellent " operation of step 7.2 less than excellent probability is increased to the individuality;Otherwise, not to the individuality
" increasing excellent " operation;
Step 7.2:An individual is randomly selected in the Fineness gene storehouse GBL obtained from step 4, is randomly selected from individuality
One section of gene of random size, with this section of gene replace described in step 7.1 individuality in a certain section of gene;And after calculating replacement
The individual fitness;Specifically " increase excellent " process as follows:
pop2(i,m1:M2)=GBL (j, m1:m2) (2)
Wherein, i is i-th individuality in the second middle population pop2, and j is j-th high-quality in the GBL of Fineness gene storehouse
Body, m1:M2 is one section of gene of m1 to m2 in certain individuality;
Step 7.3:If replace after individual adaptation degree fitness1 for underrange (NaN), bear infinite (- Inf) or
Plural (A+Bi/A-Bi), and more than original individual fitness fitness, i.e. fitness1 (i) ≠ NaN/-inf/A+Bi/A-
Bi and fitness1 (i)>Fitness (i), then be replaced successfully;Otherwise, do not replace;
Step 7.4:After the completion of " increasing excellent " operation, new middle population pop3 is formed.
Further, the step 8 specifically includes following steps:
Step 8.1:The comentropy H of the 3rd middle population pop3 that calculation procedure 7 is obtained, is herein using comentropy as kind
The diversity index of group;
Step 8.2:Evolutionary process is divided into three phases, the three phases correspondence according to maximum iteration maxgen
Information entropy threshold be followed successively by T1, T2, T3, wherein T1>T2>T3;
Step 8.3:If the comentropy H of the 3rd middle population pop3 that step 8.1 is calculated is less than current iteration time
Place stage corresponding information entropy threshold is counted, then population pop3 middle to the 3rd carries out population intervention.
Further, the step 8.1 is comprised the following steps that:
Step 8.1.1:The functor collection F and full stop collection T obtained using step 2, count i-th functor or termination
Accord with the number of times C occurred on all individual same gene position j of the 3rd middle population pop3ij;
Step 8.1.2:I-th functor or full stop are obtained in all individual same gene positions of the population pop3
The probability P occurred on upper jij, computing formula is as follows:
Step 8.1.3:Calculate the comentropy of the 3rd middle population pop3;The specific formula for calculation of comentropy is as follows:
Wherein, L is sum i.e. each individual total length of each individual gene location, and S is functor and full stop
Sum.
Further, the step 8.3 is comprised the following steps that:
Step 8.3.1:Each individuality of population pop3 middle to the 3rd calculates individual adaptation degree, and by fitness to individuality
Carry out ascending order arrangement;
Step 8.3.2:The quantity a of feasible individual that definition is randomly generated, individual of the mirror image replacement operation for carrying out
Number b;
Step 8.3.3:Randomly generate a feasible individualities and replace the worst preceding a individuality of fitness after sequence;
Step 8.3.4:The common b individuality of (a+1)~(a+b) after to sequence carries out mirror image replacement operation, i.e., according to
The correspondence position of functor set F and mirror image function symbol set M, replaces with all functors of each individuality and the functor
The mirror image function symbol of correspondence position;
Step 8.3.5:Fitness is recalculated respectively to a+b after replacement individuality, if feasible individual, i.e.,
Fitness1 (i) ≠ NaN/-inf/A+Bi/A-Bi, then be replaced successfully;Otherwise, do not replace;Obtain the 4th middle population pop4.
Further, the step 10 is comprised the following steps that:
Step 10.1:According to the algorithm of tournament selection contest scale p set in step 2, every time from the 4th that step 8 is obtained
Between population pop4 individuality in randomly select p it is individual, using the individuality of fitness maximum in this p individuality as the 5th among
Group pop5;Then the above-mentioned steps in step 10.1 are repeated sizepop-1 times again, and the fitness that will be obtained every time is most
Big individuality is added sequentially the 5th middle population, finally gives the 5th middle population pop5 that individual number is sizepop;
Step 10.2:Each individual fitness in the 5th middle population pop5 is calculated, the individual of fitness value maximum is chosen
Body is used as the optimum individual bestpop in the 5th middle population pop5, if the fitness of the optimum individual bestpop is more than
The fitness of current optimum individual maxpop, then replace current optimum individual maxpop with bestpop;Otherwise, described will work as
Preceding optimum individual maxpop replaces the individuality of the fitness minimum in the 5th middle population pop5;Iterations plus 1, and will
5th middle population pop5 is used as current population pop;Go to step 5.
Wheat aphid quantitative forecast system based on human evolution's gene expression programming, including the module 1 being sequentially connected:Modeling
Sample data acquisition module, module 2:Initialization of population module, module 3:Current population's fitness computing module, module 4:High-quality
Gene pool (GBL) and characteristic information storehouse (BBL) inferior set up module and module 5:First middle population pop1 and matrix
Badgen sets up module,
Also include module 6:Second middle population pop2 sets up module, module 7:3rd middle population pop3 set up module,
Module 8:4th middle population pop4 sets up module, module 9:Fineness gene storehouse GBL update module, module 10:5th among
Group pop5 sets up module and module 11:Current population optimum individual is chosen and decoder module;
Described module 1, for realizing following functions:
Gathering N number of time point investigates the field meteorological condition and population dynamic of small grain aphid, obtains for small
The field investigation data that grain aphid population quantity model is set up, such as daily maximum temperature, daily mean temperature, intra day ward, wheat
Growing stage, small grain aphid natural enemy quantity, and small grain aphid total amount, as the sample data of modeling, are designated as matrix A;
Described module 2, for realizing following functions:
Initiation parameter is set, and produces initialization population, using initialization population as current population pop;Described ginseng
Number includes population scale sizepop, maximum iteration maxgen, functor set F, full stop set T, mirror image function symbol collection
Close M, connector between mrna length, chromogene number, chromogene, fall string length, insert string length, mutation probability, fall
String probability, insert string probability, recombination probability, Fineness gene storehouse scale m, increase excellent probability, algorithm of tournament selection contest scale p;
Described module 3, for realizing following functions:
Each individual fitness in the current population pop that the matrix A computing module 2 obtained using module 1 is obtained, and protect
Deposit optimum individual maxpop and its fitness;
Described module 4, for realizing following functions:
According to each the individual fitness in the current population pop that module 3 is obtained, set up Fineness gene storehouse (GBL) and
Characteristic information storehouse (BBL) inferior;
Described module 5, for realizing following functions:
Whether current iterations is judged more than maximum iteration maxgen, is then revolving die block 11;Otherwise carry out as
Lower operation:First middle population is obtained to the individual natural evolution operation for performing gene expression programming in current population pop
pop1;The individual fitness of the described first middle population pop1 is calculated, while record wherein infeasible individual and its base inferior
The position in the individuality of cause, is saved in matrix badgen;
Described module 6, for realizing following functions:
The functor collection F and full stop collection T obtained using module 2, it is infeasible in the matrix badgen obtained to module 5
Individuality in individual corresponding first middle population pop1 carries out " roguing " operation, will be infeasible in the first middle population pop1
Individuality becomes feasible individual, obtains the second middle population pop2;
Described module 7, for realizing following functions:
The Fineness gene storehouse GBL obtained using module 4, the second middle population pop2 obtained to module 6 performs a soma
" increasing excellent " operation in pre-, forms the 3rd middle population pop3;
Described module 8, for realizing following functions:
The functor collection F and full stop collection T obtained using module 2, the 3rd middle population pop3's that computing module 7 is obtained
Comentropy H;According to comentropy H and the comparative result of the information entropy threshold of setting, the 3rd middle population pop3 obtained to module 7
Population intervention operation is performed, the 4th middle population pop4 is obtained;
Described module 9, for realizing following functions:
Merge the 4th middle population pop4 and Fineness gene storehouse GBL, and according to the size of the individual adaptation degree after merging, obtain
Fineness gene storehouse GBL after to renewal;
Described module 10, for realizing following functions:
According to current optimum individual maxpop and its fitness, the 4th middle population pop4 obtained to module 8 carries out band
The algorithm of tournament selection for having elitism strategy is operated, and obtains the 5th middle population pop5;Iterations plus 1, and the 5th among
The contemporary population pop of group's conduct, into module 5;
Described module 11, for realizing following functions:
The fitness of individuality in current population pop is calculated, the maximum individuality of fitness value is chosen as in current population pop
Optimum individual bestpop, the effective length of each gene of the optimum individual bestpop is calculated, according to each gene
Effective length travels through each gene from right to left, and decoding obtains the mathematic(al) representation of each gene, then with the same approach of module 3
Obtain the individual expression formula of optimum individual bestpop;By the individual expression formula of described optimum individual bestpop, according to table
The real-time measurement values of each parameter up in formula, can obtain the predicted value of small grain aphid population quantity.
Compared with prior art, the present invention has advantages below:
1st, the present invention is repaired using " roguing " operation to the worst individual or infeasible individual that are produced in natural evolution
Feasible individual is translated into, the Fineness gene fragment in Fineness gene storehouse is expanded in population using " increasing excellent " operation
Dissipate, so as to effectively increase the quality of individuality in population, accelerate convergence of algorithm speed.
2nd, the present invention is intervened using population, is received due to precocious caused by population diversity missing for during natural evolution
Phenomenon is held back, by individual with the very big mirror image of the diversity factor of the new feasible individual for randomly generating and Mirroring Mapping generation in population
Body replaces the poor individuality of equivalent amount, to increase the diversity of population, so as to effectively improve the global optimizing ability of algorithm.
3rd, the present invention uses the algorithm of tournament selection method with elitism strategy to be further ensured that algorithm in theory can be with general
Rate 1 converges to globally optimal solution.
4th, the present invention effectively overcomes traditional gene expression to compile by improving individual quality, the diversity of regulation population
Stagnation, the precocious phenomenon even degenerated occurred in journey, with the global optimizing energy accelerated convergence of algorithm speed, improve algorithm
Power.Using the gene expression programming of the simulation human evolution for proposing with small grain aphid quantity Key Influential Factors, such as
Daily maximum temperature, daily mean temperature, intra day ward, wheat growing stage, small grain aphid natural enemy quantity are small as input variable
Grain aphid quantity is output variable, the Mathematical Modeling of small grain aphid total amount is automatically excavated, so as to predict wheat long tube
The quantity of aphid.
Brief description of the drawings
Fig. 1 is flow chart of the method for the present invention;
Technical scheme is explained further and is illustrated with reference to the accompanying drawings and examples.
Specific embodiment
Embodiment:
Defer to above-mentioned technical proposal, the wheat aphid quantitative forecast side based on human evolution's gene expression programming of the invention
Method, specifically includes following steps:
Step 1:Gathering N number of time point investigates the field meteorological condition and population dynamic of small grain aphid, and N can set
13~21 are set to, the field investigation data set up for small grain aphid population quantity model are obtained.Choose field investigation data
Middle Key Influential Factors:Daily maximum temperature, daily mean temperature, intra day ward, wheat growing stage, small grain aphid natural enemy number
Amount, and small grain aphid total amount, as the sample data of modeling, are designated as matrix A, and matrix A is the matrix of the row of N rows 6;If matrix
The i-th row in A is designated as Ai, i=1~N, its value is i-th above-mentioned 6 sample data at time point, i.e.,:Daily maximum temperature (is taken the photograph
Family name's degree), daily mean temperature (degree Celsius), intra day ward (millimeter), wheat growing stage, small grain aphid natural enemy quantity and small
Grain aphid total amount (head/stem).
Step 2:Initiation parameter is set, and produces initialization population, using initialization population as current population pop;Institute
The parameter stated includes that population scale sizepop (taking 30-200), maximum iteration maxgen (take 100-10000), functor
Set F, full stop set T, mirror image function symbol set M, connector between mrna length, chromogene number, chromogene,
String length, insert string length, mutation probability, fall string probability, insert string probability, recombination probability, Fineness gene storehouse scale m and (take less
In 0.5*sizepop), increase excellent probability (taking 0.1-0.5), algorithm of tournament selection contest scale p (taking 1-30);
In the present embodiment, it is assumed that population scale sizepop=70;Maximum iteration maxgen=1000;Functor collection
Close F=+,-, * ,/, sin, cos, tan, cot, ln, log, exp, Tx, Q, S, abs, wherein+represent plus ,-represent subtract, * generations
Table multiplies ,/represent and remove, sin represents SIN function, and cos represents cosine function, and tan represents tan, and cot represents cotangent,
It is the index at bottom that Tx is represented with 10, and log represents denary logarithm, and ln represents natural logrithm, and exp is represented with natural constant e
It is the index at bottom, Q is represented and extracted square root, S representatives square, and abs represents absolute value;Full stop set T={ X1, X2, X3, X4, X5 },
Wherein X1 represents max. daily temperature, and X2 represents daily mean temperature, and X3 represents intra day ward, and X4 represents wheat growing stage, and X5 is represented
Small grain aphid natural enemy quantity;Mirror image function symbol set M=-,+,/, *, cos, sin, cot, tan, exp, Tx, ln, log, S,
Q、abs};Mrna length=13;Chromogene number=6;Connector is "+" between chromogene;String length=4;Insert
String length=3;Mutation probability=0.3;Go here and there probability=0.1;Insert string probability=0.1;Recombination probability=0.2;Fineness gene storehouse
Scale m=0.2*sizepop;Increase excellent probability=0.2;Algorithm of tournament selection contest scale p=3.
Step 3:The adaptation of each individuality in the current population pop that the matrix A calculation procedure 2 obtained using step 1 is obtained
Degree, and preserve optimum individual maxpop and its fitness.Comprise the following steps that:
Step 3.1:The effective length of each gene of each individuality in the current population pop that calculation procedure 2 is obtained;
Step 3.2:According to the effective length of each gene of each individuality, each gene position is traveled through from right to left, draw
The mathematic(al) representation of each gene;(Datong District, 2 kinds of coding/decoding methods of Chen Qiao cloud gene expression programmings are thanked herein with reference to document
[J] computer engineering, 2008,34 (23):The first decoding process in 210-211.).
Step 3.3:Each gene of s-th individuality that will be obtained in step 3.2 with connector "+" between chromogene
Expression formula is coupled together, wherein, s=1~(sizepop=70) obtains s-th expression formula of individuality, and by sample data square
Data correspondence in battle array A substitutes into s-th expression formula of individuality, calculates predicted values of the individual s in the expression formula of time point j
Similarly obtain predicted value of s-th individuality in the expression formula at all time points;
Step 3.4:Each individual fitness is calculated using the multiple correlation coefficient method shown in formula (1), and preserves optimal
Body maxpop and its fitness;Fitness computing formula is:
Fitness=R2=1-SSE/SST (1) is to be adapted to using a square conduct for the multiple correlation coefficient in statistics
Degree, wherein, R is multiple correlation coefficient,
Wherein, N is the time point sum that the small grain aphid total amount of observation is chosen, yjIt is j-th time stored in matrix A
The small grain aphid total amount that point is observed,It is yjAverage value;Represent predicted values of the individuality s in the expression formula of time point j;
SSE is residual sum of squares (RSS), and SST is total sum of squares of deviations;Fitness R2Value is bigger, and expression model-fitting degree is higher.Optimum individual
Criterion is:The maximum individuality of fitness in all individualities.
Step 4:According to each the individual fitness in the current population pop that step 3 is obtained, Fineness gene storehouse is set up
And characteristic information storehouse inferior (BBL) (GBL).Comprise the following steps that:
Step 4.1:Using Fitness=sort (fitness, ' descend ') ask in function pair step 3 per each and every one
The fitness fitness of body carries out descending sort;
Step 4.2:According to the scale m in the Fineness gene storehouse set in step 2, fitness maximum in selection population pop
First m individual, sets up Fineness gene storehouse (GBL);
Step 4.3:It is underrange (NaN), the individuality for bearing infinite (- Inf) or plural (A+Bi/A-Bi) by fitness, makees
It is infeasible individual;Three kinds of characteristic informations inferior are underrange (NaN), bear infinite (- Inf) and plural (A+Bi/A-Bi), is deposited
Enter characteristic information storehouse (BBL) inferior.
Step 5:Judge that whether current iterations, more than maximum iteration maxgen, is to go to step 11;Otherwise
Proceed as follows:The individual natural evolution operation for performing gene expression programming in current population pop is obtained in first
Between population pop1;Calculate the individual fitness of the described first middle population pop1, at the same record wherein infeasible individual and its
The position in the individuality of gene inferior, is saved in matrix badgen.Comprise the following steps that:
Step 5.1:Judge that whether current iterations, more than maximum iteration maxgen, is to go to step 11;It is no
Then perform step 5.2;
Step 5.2:It is individual to each in current population pop according to the probability of each genetic operator set in step 2
Successively perform gene expression programming natural evolution operation, that is, make a variation, fall string, IS insert string, RIS insert string, single-point restructuring, 2 points
Restructuring, the operation of genetic recombination, obtain the first middle population pop1.Concrete operations are:Operated for certain natural evolution, first
The random number between one 0~1 is produced, judges the random number whether less than this kind of natural evolution operation set in step 2
Correspondence probability, is to carry out the operation to individuality;The operation is not carried out otherwise;For example, before carrying out mutation operation to certain individuality,
The random number between one 0~1 is generated first, it is right if the random number is less than the corresponding mutation probability 0.3 of mutation operation
The individuality carries out mutation operation;Mutation operation is not carried out otherwise, and other operations are carried out successively;
Step 5.3:According to each individual fitness in the first middle population pop1 of formula (1) calculating, while record is every
Position of the individual gene position inferior in individuality.
The determination of gene position inferior:According to the effective length of each gene of each individuality, each base is traveled through from right to left
Because of position, for each gene position, (Datong District, 2 kinds of coding/decoding method [J] meters of Chen Qiao cloud gene expression programmings are thanked according to document
Calculation machine engineering, 2008,34 (23):The decoding process be given in 210-211.) finds one or two operand of the gene position,
Operand is calculated using the functor (functor is the functor that functor is concentrated) in the gene position, if calculating knot
The value of fruit is Nan, Inf or plural number, then the gene position is exactly a gene position inferior.For example, it is assumed that c-th individuality in the population
D positions on functor be "/", functor "/" is binary operator, and two operands to d calculate, if
2nd operand is 0, and the value obtained after being calculated the two operands using functor "/" is Inf, then d is exactly
One gene position inferior.
Step 5.4:The gene bit matrix inferior of infeasible individual is set up using the gene position inferior in step 5.3
badgen.Its line number represents the individual sequence number of the first middle population pop1;Each value in matrix in certain row represents be located at successively
Each gene inferior location in the individuality in the individuality of the row;
For example, in the matrix badgen being provided below, sizepop refers to the individual total of the first middle population pop1
Number, because the sizepop individual contained poor quality gene digit is 7, so matrix column number is 7 row, it is bad in other individualities
Inadequate 7 of matter gene digit, it is assigned to 0.The value of each row is respectively in the first row:3rd, 6,11,16,20, represent in the row
The 3rd, 6,11,16,20 of body are gene position inferior.For another example, sizepop-1 rows are all 0 in matrix, represent row correspondence
Individuality without poor quality gene position.
Step 6:The functor collection F and full stop collection T obtained using step 2, in the matrix badgen obtained to step 5
Individuality in the corresponding first middle population pop1 of infeasible individual carries out " roguing " operation, by the first middle population pop1
Infeasible individual becomes feasible individual, obtains the second middle population pop2.Comprise the following steps that:
Step 6.1:Each gene location institute inferior of all infeasible individuals is right in the matrix badgen obtained to step 5
A functor or a full stop are randomly selected in the symbol answered, the functor collection F obtained from step 2 and full stop collection T, it is right
Symbol at described gene location carries out man-to-man replacement;
Step 6.2:The fitness of each infeasible individual after being replaced in calculation procedure 6.1, if what is obtained is individual
The value of fitness is not underrange (NaN), bears infinite (- Inf) or plural (A+Bi/A-Bi), is then replaced successfully;Otherwise, continue
Step 6.1 is performed, untill being replaced successfully;Obtain the second middle population pop2;
Step 7:The Fineness gene storehouse GBL obtained using step 4, the second middle population pop2 obtained to step 6 is performed
" increasing excellent " during individuality is intervened operates, and forms the 3rd middle population pop3.Comprise the following steps that:
Step 7.1:Each in the second middle population pop2 obtained for step 6 is individual, if randomly generate one
Number between individual 0-1 then carries out " increasing excellent " operation of step 7.2 less than excellent probability is increased to the individuality;Otherwise, not to the individuality
" increasing excellent " operation;
Step 7.2:An individual is randomly selected in the Fineness gene storehouse GBL obtained from step 4, is randomly selected from individuality
One section of gene of random size, with this section of gene replace described in step 7.1 individuality in a certain section of gene;And after calculating replacement
The individual fitness;Specifically " increase excellent " process as follows:
pop2(i,m1:M2)=GBL (j, m1:m2) (2)
Wherein, i is i-th individuality in the second middle population pop2, and j is j-th high-quality in the GBL of Fineness gene storehouse
Body, m1:M2 is one section of gene of m1 to m2 in certain individuality;
Step 7.3:If the individual fitness fitness1 after replacing not is underrange (NaN), bears infinite (- Inf)
Or it is plural (A+Bi/A-Bi), and more than the original fitness fitness of the individuality, i.e. fitness1 (i) ≠ NaN/-inf/A
+ Bi/A-Bi and fitness1 (i)>Fitness (i), then be replaced successfully;Do not replace otherwise;
Step 7.4:After the completion of " increasing excellent " operation, the 3rd middle population pop3 is formed.
Step 8:The functor collection F and full stop collection T obtained using step 2, the 3rd middle population that calculation procedure 7 is obtained
The comentropy H of pop3;According to comentropy H and the comparative result of the information entropy threshold of setting, the 3rd among obtained to step 7
Group pop3 performs population intervention operation, obtains the 4th middle population pop4;Comprise the following steps that:
Step 8.1:The comentropy H of the 3rd middle population pop3 that calculation procedure 7 is obtained, is herein using comentropy as kind
The diversity index of group.Comprise the following steps that;
Step 8.1.1:The functor collection F and full stop collection T obtained using step 2, count i-th functor or termination
Accord with the number of times C occurred on all individual same gene position j of the 3rd middle population pop3ij;In this example i be 1~
20, refer to the sum of functor and full stop in functor collection F and full stop collection T;
Step 8.1.2:Obtain all individual same bases of i-th functor or full stop in the 3rd middle population pop3
Because of the probability P occurred on j on positionij, computing formula is as follows:
Step 8.1.3:Calculate the comentropy of the 3rd middle population pop3;The specific formula for calculation of comentropy is as follows:
Wherein, L is sum i.e. each individual total length of each individual gene location in the 3rd middle population pop3,
S is the sum of functor and full stop.
Step 8.2:Evolutionary process is divided into three phases (for example, according to maximum iteration maxgen:1- the 1/th
3*maxgen times, 1/3*maxgen+1- the 2/3*maxgen times, 2/3*maxgen+1- the maxgen times), described three
Stage, corresponding information entropy threshold was followed successively by T1, T2, T3, wherein T1>T2>T3;
Step 8.3:If the comentropy H of the 3rd middle population pop3 that step 8.1.3 is calculated is less than current iteration
The corresponding information entropy threshold of stage where number of times, then population pop3 middle to the 3rd carry out population intervention;It is specific that population is intervened
Step is as follows:
Step 8.3.1:Each individuality of population pop3 middle to the 3rd calculates individual adaptation degree, and by fitness to individuality
Carry out ascending order arrangement;
Step 8.3.2:The quantity a (taking a=0.1*sizepop) of the feasible individual that definition is randomly generated, the mirror image for carrying out
The individual number b (taking b=0.2*sizepop) of replacement operation;
Step 8.3.3:Randomly generate a feasible individualities and replace the worst preceding a individuality of fitness after sequence;
Step 8.3.4:The common b individuality of (a+1)~(a+b) after to sequence carries out mirror image replacement operation, i.e., according to
The correspondence position of functor set F and mirror image function symbol set M, replaces with all functors of each individuality and the functor
The mirror image function symbol of correspondence position;Such as F=+,-, * ,/, sin, cos, tan, cot, ln, log, exp, Tx, Q, S, abs, M
=-,+,/, *, cos, sin, cot, tan, exp, Tx, ln, log, S, Q, abs, then functor set F and mirror image function accord with collection
Close the functor in M correspond, mirror image each other;
Step 8.3.5:Fitness is recalculated respectively to a+b after replacement individuality, if feasible individual, i.e.,
Fitness1 (i) ≠ NaN/-inf/A+Bi/A-Bi, then be replaced successfully;Otherwise, do not replace;Obtain the 4th middle population pop4.
Step 9:Merge the 4th middle population pop4 and Fineness gene storehouse GBL, and according to the individual adaptation degree after merging
Size, the Fineness gene storehouse GBL after being updated.Comprise the following steps that:
Step 9.1:The Fineness gene storehouse GBL that 4th middle population pop4 and step 4 are obtained merges, i.e.,:Union=
Union (fitness of Fitness, GBL), union () function is pooled function;
Step 9.2:The all individual fitness after merging are calculated, and carries out descending sort, i.e.,:Fitness1=sort
(Union, ' descend '), take the individual of preceding m fitness and update Fineness gene storehouse GBL, the Fineness gene storehouse after being updated
GBL。
Step 10:According to current optimum individual maxpop and its fitness, the 4th middle population pop4 obtained to step 8
Carry out the algorithm of tournament selection with elitism strategy to operate, obtain the 5th middle population pop5.Comprise the following steps that:
Step 10.1:According to the algorithm of tournament selection contest scale p set in step 2, every time from the 4th that step 8 is obtained
Between population pop4 individuality in randomly select p it is individual, using the individuality of fitness maximum in this p individuality as the 5th among
Group pop5;Then the above-mentioned steps in step 10.1 are repeated sizepop-1 times again, and the fitness that will be obtained every time is most
Big individuality is added sequentially the 5th middle population, finally gives the 5th middle population pop5 that individual number is sizepop;
Step 10.2:Each individual fitness in the 5th middle population pop5 is calculated, the individual of fitness value maximum is chosen
Body is used as the optimum individual bestpop in the 5th middle population pop5, if the fitness of the optimum individual bestpop is more than
The fitness of current optimum individual maxpop, then replace current optimum individual maxpop with bestpop;Otherwise, described will work as
Preceding optimum individual maxpop replaces the individuality of the fitness minimum in the 5th middle population pop5;Iterations plus 1, and will
5th middle population pop5 is used as current population pop;Go to step 5.
Step 11:The fitness of individuality in current population pop is calculated, the maximum individual conduct of fitness value currently kind is chosen
Optimum individual bestpop in group pop, calculates the effective length of each gene of the optimum individual bestpop, according to each
The effective length of gene travels through each gene from right to left, and decoding obtains the mathematic(al) representation of each gene, then with step 3.3
Same approach obtains the individual expression formula of optimum individual bestpop.
By the individual expression formula of described optimum individual bestpop, according to the real-time measurement values of each parameter in expression formula, can
To obtain the predicted value of small grain aphid population quantity.
Wheat aphid quantitative forecast system based on human evolution's gene expression programming, including the module 1 being sequentially connected:Modeling
Sample data acquisition module, module 2:Initialization of population module, module 3:Current population's fitness computing module, module 4:High-quality
Gene pool (GBL) and characteristic information storehouse (BBL) inferior set up module and module 5:First middle population pop1 and matrix
Badgen sets up module,
Also include module 6:Second middle population pop2 sets up module, module 7:3rd middle population pop3 set up module,
Module 8:4th middle population pop4 sets up module, module 9:Fineness gene storehouse GBL update module, module 10:5th among
Group pop5 sets up module and module 11:Current population optimum individual is chosen and decoder module;
Described module 1, for realizing following functions:
Gathering N number of time point investigates the field meteorological condition and population dynamic of small grain aphid, obtains for small
The field investigation data that grain aphid population quantity model is set up, such as daily maximum temperature, daily mean temperature, intra day ward, wheat
Growing stage, small grain aphid natural enemy quantity, and small grain aphid total amount, as the sample data of modeling, are designated as matrix A;
Described module 2, for realizing following functions:
Initiation parameter is set, and produces initialization population, using initialization population as current population pop;Described ginseng
Number includes population scale sizepop, maximum iteration maxgen, functor set F, full stop set T, mirror image function symbol collection
Close M, connector between mrna length, chromogene number, chromogene, fall string length, insert string length, mutation probability, fall
String probability, insert string probability, recombination probability, Fineness gene storehouse scale m, increase excellent probability, algorithm of tournament selection contest scale p;
Described module 3, for realizing following functions:
Each individual fitness in the current population pop that the matrix A computing module 2 obtained using module 1 is obtained, and protect
Deposit optimum individual maxpop and its fitness;
Described module 3 includes submodule 31, submodule 32, submodule 33 and submodule 34;Wherein,
Submodule 31 is used to realize following functions:
The effective length of each gene of each individuality in the current population pop that computing module 2 is obtained.
Submodule 32 is used to realize following functions:
According to the effective length of each gene of each individuality, each gene position is traveled through from right to left, draw each gene
Mathematic(al) representation.
Submodule 33 is used to realize following functions:
The expression formula of each gene of s-th individuality that will be obtained in submodule 2 with connector "+" between chromogene connects
Pick up and, obtain s-th expression formula of individuality, and the data correspondence in sample data matrix A is substituted into s-th expression of individuality
Formula, calculates predicted values of the individual s in the expression formula of time point jSimilarly obtain expression of s-th individuality at all time points
The predicted value of formula;
Submodule 34 is used to realize following functions:
Each individual fitness is calculated using the multiple correlation coefficient method shown in formula (1), and preserves optimum individual maxpop
And its fitness;Fitness computing formula is:
Fitness=R2=1-SSE/SST (1) is to be adapted to using a square conduct for the multiple correlation coefficient in statistics
Degree, wherein, R is multiple correlation coefficient,
Wherein, N is the time point sum that the small grain aphid total amount of observation is chosen, yjIt is j-th time stored in matrix A
The small grain aphid amount that point is observed,It is yjAverage value.SSE is residual sum of squares (RSS), and SST is total sum of squares of deviations.It is optimal
Individual criterion is:The maximum individuality of fitness in all individualities.
Described module 4, for realizing following functions:
According to each the individual fitness in the current population pop that module 3 is obtained, set up Fineness gene storehouse (GBL) and
Characteristic information storehouse (BBL) inferior;
Described module 4 includes submodule 41, submodule 42, submodule 43 and submodule 44;Wherein,
Submodule 41 is used to realize following functions:
Using Fitness=sort (fitness, ' descend ') adaptation of each individuality asked in function pair module 3
Degree fitness carries out descending sort.
Submodule 42 is used to realize following functions:
According to the scale m in the Fineness gene storehouse set in module 2, choose fitness maximum in population pop preceding m
Body, sets up Fineness gene storehouse (GBL);
Submodule 43 is used to realize following functions:
It is underrange (NaN), the individuality for bearing infinite (- Inf) or plural (A+Bi/A-Bi) by fitness, as infeasible
It is individual;Three kinds of characteristic informations inferior are underrange (NaN), bear infinite (- Inf) and plural (A+Bi/A-Bi), spy inferior is stored in
Levy information bank (BBL).
Described module 5, for realizing following functions:
Whether current iterations is judged more than maximum iteration maxgen, is then revolving die block 11;Otherwise carry out as
Lower operation:First middle population is obtained to the individual natural evolution operation for performing gene expression programming in current population pop
pop1;The individual fitness of the described first middle population pop1 is calculated, while record wherein infeasible individual and its base inferior
The position in the individuality of cause, is saved in matrix badgen;
Described step 5 includes submodule 51, submodule 52, submodule 53 and submodule 54, wherein,
Submodule 51 is used to realize following functions:
Judge that whether current iterations, more than maximum iteration maxgen, is then to enter module 11;Otherwise enter
Submodule 52.
Submodule 52 is used to realize following functions:
According to the probability of each genetic operator set in module 2, each individuality in current population pop is performed successively
Gene expression programming natural evolution operation, that is, make a variation, fall string, IS insert string, RIS insert string, single-point restructuring, 2 points restructuring, bases
Because of the operation for recombinating, the first middle population pop1 is obtained.
Submodule 53 is used to realize following functions:
According to each individual fitness in the first middle population pop1 of formula (1) calculating, while recording each individuality
Gene position inferior;
Submodule 54 is used to realize following functions:
The gene bit matrix badgen inferior of infeasible individual is set up using the gene position inferior in submodule 53.
Described module 6, for realizing following functions:
The functor collection F and full stop collection T obtained using module 2, it is infeasible in the matrix badgen obtained to module 5
Individuality in individual corresponding first middle population pop1 carries out " roguing " operation, will be infeasible in the first middle population pop1
Individuality becomes feasible individual, obtains the second middle population pop2;
Described module 6 includes submodule 61 and submodule 62, wherein,
Submodule 61 is used to realize following functions:
Symbol in the matrix badgen obtained to module 5 corresponding to each gene location inferior of all infeasible individuals
Number, functor collection F that slave module 2 is obtained and a functor or a full stop are randomly selected in full stop collection T, to described
Symbol at gene location carries out man-to-man replacement.
Submodule 62 is used to realize following functions:
The fitness of each infeasible individual after being replaced in calculating sub module 61, if the individual fitness for obtaining
Value is not underrange (NaN), bears infinite (- Inf) or plural (A+Bi/A-Bi), is then replaced successfully;Otherwise, submodule is gone successively to
Block 61, untill being replaced successfully;Obtain the second middle population pop2.
Described module 7, for realizing following functions:
The Fineness gene storehouse GBL obtained using module 4, the second middle population pop2 obtained to module 6 performs a soma
" increasing excellent " operation in pre-, forms the 3rd middle population pop3;
Described module 7 includes submodule 71, submodule 72, submodule 73 and submodule 74, wherein,
Submodule 71 is used to realize following functions:
Each in the second middle population pop2 obtained for module 6 is individual, if between the 0-1 for randomly generating
Number less than excellent probability is increased, then " increasing excellent " operation of module 72 is carried out to the individuality;Otherwise, operation " is not increased excellent " to the individuality.
Submodule 72 is used to realize following functions:
An individual is randomly selected in the Fineness gene storehouse GBL that slave module 4 is obtained, one section is randomly selected from individuality at random
The gene of size, a certain section of gene in the individuality of module 71 is replaced with this section of gene;And calculate the suitable of the individuality after replacing
Response;Specifically " increase excellent " process as follows:
pop2(i,m1:M2)=GBL (j, m1:m2) (2)
Wherein, i is i-th individuality in the second middle population pop2, and j is j-th high-quality in the GBL of Fineness gene storehouse
Body, m1:M2 is one section of gene of m1 to m2 in certain individuality;
Submodule 73 is used to realize following functions:
If the individual adaptation degree fitness1 after replacing not is underrange (NaN), negative infinite (- Inf) or plural number (A+
Bi/A-Bi), and more than original individual fitness fitness, i.e. fitness1 (i) ≠ NaN/-inf/A+Bi/A-Bi and
fitness1(i)>Fitness (i), then be replaced successfully;Otherwise, do not replace;
Submodule 74 is used to realize following functions:
After the completion of " increasing excellent " operation, new middle population pop3 is formed.
Described module 8, for realizing following functions:
The functor collection F and full stop collection T obtained using module 2, the 3rd middle population pop3's that computing module 7 is obtained
Comentropy H;According to comentropy H and the comparative result of the information entropy threshold of setting, the 3rd middle population pop3 obtained to module 7
Population intervention operation is performed, the 4th middle population pop4 is obtained;
Described module 8 includes submodule 81, submodule 82 and submodule 83, wherein,
Submodule 81 is used to realize following functions:
The comentropy H of the 3rd middle population pop3 that computing module 7 is obtained, is herein using comentropy as the various of population
Property index.
Submodule 82 is used to realize following functions:
Evolutionary process is divided into three phases according to maximum iteration maxgen, the corresponding comentropy of the three phases
Threshold value is followed successively by T1, T2, T3, wherein T1>T2>T3.
Submodule 83 is used to realize following functions:
If the comentropy H of the 3rd middle population pop3 that submodule 81 is calculated is less than rank where current iteration number of times
The corresponding information entropy threshold of section, then population pop3 middle to the 3rd carry out population intervention, form the 4th middle population pop4.
Described module 9, for realizing following functions:
Merge the 4th middle population pop4 and Fineness gene storehouse GBL, and according to the size of the individual adaptation degree after merging, obtain
Fineness gene storehouse GBL after to renewal;
Described module 10, for realizing following functions:
According to current optimum individual maxpop and its fitness, the 4th middle population pop4 obtained to module 8 carries out band
The algorithm of tournament selection for having elitism strategy is operated, and obtains the 5th middle population pop5;Go to step 5;
Described module 10 includes submodule 101 and submodule 102, wherein,
Submodule 101 is used to realize following functions:
According to the algorithm of tournament selection contest scale p set in module 2, the 4th middle population that each slave module 8 is obtained
P individuality is randomly selected in the individuality of pop4, using the maximum individuality of fitness in this p individuality as the 5th middle population
pop5;Then repeat into sizepop-1 times in module 101, and the maximum individuality of the fitness that will be obtained every time sequentially adds
Enter the 5th middle population, finally give the 5th middle population pop5 that individual number is sizepop.
Submodule 102 is used to realize following functions:
Each individual fitness in the 5th middle population pop5 is calculated, the maximum individuality of fitness value is chosen as the 5th
Optimum individual bestpop in middle population pop5, if the fitness of the optimum individual bestpop is more than current optimal
The fitness of body maxpop, then replace current optimum individual maxpop with bestpop;Otherwise, by the current optimum individual
Maxpop replaces the individuality of the fitness minimum in the 5th middle population pop5;Iterations plus 1, and by the 5th among
Group pop5 is used as current population pop;Into module 5.
Described module 11, for realizing following functions:
The fitness of individuality in current population pop is calculated, the maximum individuality of fitness value is chosen as in current population pop
Optimum individual bestpop, the effective length of each gene of the optimum individual bestpop is calculated, according to each gene
Effective length travels through each gene from right to left, and decoding obtains the mathematic(al) representation of each gene, then with the same approach of module 3
Obtain the individual expression formula of optimum individual bestpop;By the individual expression formula of described optimum individual bestpop, according to table
The real-time measurement values of each parameter up in formula, can obtain the predicted value of small grain aphid population quantity.
Claims (10)
1. the wheat aphid quantitative forecasting technique based on human evolution's gene expression programming, comprises the following steps:
Step 1:Gathering N number of time point investigates the field meteorological condition and population dynamic of small grain aphid, and obtaining is used for
The field investigation data that small grain aphid population quantity model is set up, by daily maximum temperature, daily mean temperature, intra day ward, small
Wheat growing stage, small grain aphid natural enemy quantity, and small grain aphid total amount, as the sample data of modeling, are designated as matrix
A;
Step 2:Initiation parameter is set, and produces initialization population, using initialization population as current population pop;Described
Parameter includes population scale sizepop, maximum iteration maxgen, functor set F, full stop set T, mirror image function symbol
Connector between set M, mrna length, chromogene number, chromogene, fall string length, insert string length, mutation probability,
String probability, insert string probability, recombination probability, Fineness gene storehouse scale m, increase excellent probability, algorithm of tournament selection contest scale p;
Step 3:Each individual fitness in the current population pop that the matrix A calculation procedure 2 obtained using step 1 is obtained, and
Preserve optimum individual maxpop and its fitness;
Step 4:According to each the individual fitness in the current population pop that step 3 is obtained, set up Fineness gene storehouse GBL and
Characteristic information storehouse BBL inferior;
Step 5:Judge that whether current iterations, more than maximum iteration maxgen, is to go to step 11;Otherwise carry out
Following operation:First among is obtained to the individual natural evolution operation for performing gene expression programming in current population pop
Group pop1;The individual fitness of the described first middle population pop1 is calculated, while record wherein infeasible individual and its poor quality
The position in the individuality of gene, is saved in matrix badgen;
Characterized in that, further comprising the steps of:
Step 6:The functor collection F and full stop collection T obtained using step 2, in the matrix badgen obtained to step 5 can not
Individuality in the individual corresponding first middle population pop1 of row carries out " roguing " operation, by the first middle population pop1 can not
Row individuality becomes feasible individual, obtains the second middle population pop2;
Step 7:The Fineness gene storehouse GBL obtained using step 4, the second middle population pop2 obtained to step 6 performs individuality
" increasing excellent " operation in intervention, forms the 3rd middle population pop3;
Step 8:The functor collection F and full stop collection T obtained using step 2, the 3rd middle population pop3 that calculation procedure 7 is obtained
Comentropy H;According to comentropy H and the comparative result of the information entropy threshold of setting, the 3rd middle population obtained to step 7
Pop3 performs population intervention operation, obtains the 4th middle population pop4;
Step 9:Merge the 4th middle population pop4 and Fineness gene storehouse GBL, and according to the size of the individual adaptation degree after merging,
Fineness gene storehouse GBL after being updated;
Step 10:According to current optimum individual maxpop and its fitness, the 4th middle population pop4 obtained to step 8 is carried out
Algorithm of tournament selection operation with elitism strategy, obtains the 5th middle population pop5;Iterations plus 1, and by the middle of the 5th
Population pop5 goes to step 5 as current population pop;
Step 11:The fitness of individuality in current population pop is calculated, the maximum individuality of fitness value is chosen as current population
Optimum individual bestpop in pop, calculates the effective length of each gene of the optimum individual bestpop, according to each base
The effective length of cause travels through each gene from right to left, and decoding obtains the mathematic(al) representation of each gene, then identical with step 3
Means obtain the individual expression formula of optimum individual bestpop;By the individual expression formula of described optimum individual bestpop, root
According to the real-time measurement values of each parameter in expression formula, the predicted value of small grain aphid population quantity can be obtained.
2. the wheat aphid quantitative forecasting technique of human evolution's gene expression programming is based on as claimed in claim 1, and its feature exists
In the step 3 specifically includes following steps:
Step 3.1:The effective length of each gene of each individuality in the current population pop that calculation procedure 2 is obtained;
Step 3.2:According to the effective length of each gene of each individuality, each gene position is traveled through from right to left, draw each
The mathematic(al) representation of gene;
Step 3.3:The expression of each gene of s-th individuality that will be obtained in step 3.2 with connector "+" between chromogene
Formula is coupled together, and obtains s-th expression formula of individuality, and the data in sample data matrix A are corresponded into s-th individuality of substitution
Expression formula, calculates predicted values of the individual s in the expression formula of time point j, s-th individuality is similarly obtained at all time points
The predicted value of expression formula;
Step 3.4:Each individual fitness is calculated using the multiple correlation coefficient method shown in formula (1), and preserves optimum individual
Maxpop and its fitness;Fitness computing formula is:
Fitness=R2=1-SSE/SST (1)
I.e. using the multiple correlation coefficient in statistics square as fitness, wherein, R is multiple correlation coefficient,
Wherein, N is the time point sum that the small grain aphid total amount of observation is chosen, yjFor j-th time point stored in matrix A is seen
The small grain aphid amount for measuring,It is yjAverage value;SSE is residual sum of squares (RSS), and SST is total sum of squares of deviations;Optimum individual
Criterion be:The maximum individuality of fitness in all individualities.
3. the wheat aphid quantitative forecasting technique of human evolution's gene expression programming is based on as claimed in claim 2, and its feature exists
In the step 5 specifically includes following steps:
Step 5.1:Judge that whether current iterations, more than maximum iteration maxgen, is to go to step 11;Otherwise hold
Row step 5.2;
Step 5.2:The probability of each genetic operator set in foundation step 2, to each individuality in current population pop successively
Perform gene expression programming natural evolution operation, that is, make a variation, fall string, IS insert string, RIS insert string, single-point restructuring, two point weights
Group, the operation of genetic recombination, obtain the first middle population pop1;
Step 5.3:According to each individual fitness in the first middle population pop1 of formula (1) calculating, while record is per each and every one
Position of the gene position inferior of body in individuality;
Step 5.4:The gene bit matrix badgen inferior of infeasible individual is set up using the gene position inferior in step 5.3.
4. the wheat aphid quantitative forecasting technique of human evolution's gene expression programming is based on as claimed in claim 1, and its feature exists
In the step 6 specifically includes following steps:
Step 6.1:In the matrix badgen obtained to step 5 corresponding to each gene location inferior of all infeasible individuals
A functor or a full stop are randomly selected in symbol, the functor collection F obtained from step 2 and full stop collection T, to described
Gene location at symbol carry out man-to-man replacement;
Step 6.2:The fitness of each infeasible individual after being replaced in calculation procedure 6.1, if the individual adaptation for obtaining
The value of degree is not underrange (NaN), bears infinite (- Inf) or plural (A+Bi/A-Bi), is then replaced successfully;Otherwise, continue executing with
Step 6.1, untill being replaced successfully;Obtain the second middle population pop2.
5. the wheat aphid quantitative forecasting technique of human evolution's gene expression programming is based on as claimed in claim 1, and its feature exists
In the step 7 specifically includes following steps:
Step 7.1:Each in the second middle population pop2 obtained for step 6 is individual, if the 0-1 for randomly generating
Between number less than excellent probability is increased, then " increasing excellent " operation of step 7.2 is carried out to the individuality;Otherwise, not to the individuality " increasing excellent "
Operation;
Step 7.2:An individual is randomly selected in the Fineness gene storehouse GBL obtained from step 4, one section is randomly selected from individuality
The gene of random size, a certain section of gene in the individuality for replacing described in step 7.1 with this section of gene;And calculate being somebody's turn to do after replacing
Individual fitness;Specifically " increase excellent " process as follows:
pop2(i,m1:M2)=GBL (j, m1:m2) (2)
Wherein, i is i-th individuality in the second middle population pop2, and j is that j-th high-quality in the GBL of Fineness gene storehouse is individual,
m1:M2 is one section of gene of m1 to m2 in certain individuality;
Step 7.3:If the individual adaptation degree fitness1 after replacing not is underrange (NaN), bears infinite (- Inf) or plural number
(A+Bi/A-Bi), and more than original individual fitness fitness, i.e. fitness1 (i) ≠ NaN/-inf/A+Bi/A-Bi and
fitness1(i)>Fitness (i), then be replaced successfully;Otherwise, do not replace;
Step 7.4:After the completion of " increasing excellent " operation, new middle population pop3 is formed.
6. the wheat aphid quantitative forecasting technique of human evolution's gene expression programming is based on as claimed in claim 1, and its feature exists
In the step 8 specifically includes following steps:
Step 8.1:The comentropy H of the 3rd middle population pop3 that calculation procedure 7 is obtained, is herein using comentropy as population
Diversity index;
Step 8.2:Evolutionary process is divided into three phases according to maximum iteration maxgen, the corresponding letter of the three phases
Breath entropy threshold is followed successively by T1, T2, T3, wherein T1>T2>T3;
Step 8.3:If the comentropy H of the 3rd middle population pop3 that step 8.1 is calculated is less than current iteration number of times institute
In stage corresponding information entropy threshold, then population pop3 middle to the 3rd carries out population intervention.
7. the wheat aphid quantitative forecasting technique of human evolution's gene expression programming is based on as claimed in claim 6, and its feature exists
In comprising the following steps that for, the step 8.1:
Step 8.1.1:The functor collection F and full stop collection T obtained using step 2, are counted i-th functor or full stop exist
The number of times C occurred on all individual same gene position j of the 3rd middle population pop3ij;
Step 8.1.2:Obtain i-th functor or the full stop j on all individual same gene positions of the population pop3
The probability P of upper appearanceij, computing formula is as follows:
Step 8.1.3:Calculate the comentropy of the 3rd middle population pop3;The specific formula for calculation of comentropy is as follows:
Wherein, L is sum i.e. each individual total length of each individual gene location, and S is the total of functor and full stop
Number.
8. the wheat aphid quantitative forecasting technique of human evolution's gene expression programming is based on as claimed in claim 6, and its feature exists
In comprising the following steps that for, the step 8.3:
Step 8.3.1:Each individuality of population pop3 middle to the 3rd calculates individual adaptation degree, and individuality is carried out by fitness
Ascending order is arranged;
Step 8.3.2:The quantity a, the individual number b of the mirror image replacement operation for carrying out of the feasible individual that definition is randomly generated;
Step 8.3.3:Randomly generate a feasible individualities and replace the worst preceding a individuality of fitness after sequence;
Step 8.3.4:The common b individuality of (a+1)~(a+b) after to sequence carries out mirror image replacement operation, i.e., according to function
The correspondence position of symbol set F and mirror image function symbol set M, replaces with all functors of each individuality corresponding with the functor
The mirror image function symbol of position;
Step 8.3.5:Fitness is recalculated respectively to a+b individuality after replacement, if feasible individual, i.e. fitness1
I () ≠ NaN/-inf/A+Bi/A-Bi, then be replaced successfully;Otherwise, do not replace;Obtain the 4th middle population pop4.
9. the wheat aphid quantitative forecasting technique of human evolution's gene expression programming is based on as claimed in claim 1, and its feature exists
In comprising the following steps that for, the step 10:
Step 10.1:According to the algorithm of tournament selection contest scale p set in step 2, the 4th among for being obtained from step 8 every time
Randomly selected in the individuality of group pop4 p it is individual, using the individuality of fitness maximum in this p individuality as the 5th middle population
pop5;Then the above-mentioned steps in step 10.1 are repeated sizepop-1 times again, and the fitness that will be obtained every time is maximum
Individuality be added sequentially the 5th middle population, finally give individual number be sizepop the 5th middle population pop5;
Step 10.2:Each individual fitness in the 5th middle population pop5 is calculated, the maximum individual work of fitness value is chosen
It is the optimum individual bestpop in the 5th middle population pop5, if the fitness of the optimum individual bestpop is more than current
The fitness of optimum individual maxpop, then replace current optimum individual maxpop with bestpop;Otherwise, by it is described it is current most
Excellent individual maxpop replaces the individuality of the fitness minimum in the 5th middle population pop5;Iterations plus 1, and by the 5th
Middle population pop5 is used as current population pop;Go to step 5.
10. the wheat aphid quantitative forecast system based on human evolution's gene expression programming, including the module 1 being sequentially connected:Modeling
Sample data acquisition module, module 2:Initialization of population module, module 3:Current population's fitness computing module, module 4:High-quality
Gene pool (GBL) and characteristic information storehouse (BBL) inferior set up module and module 5:First middle population pop1 and matrix
Badgen sets up module,
Characterized in that, also including module 6:Second middle population pop2 sets up module, module 7:3rd middle population pop3 builds
Formwork erection block, module 8:4th middle population pop4 sets up module, module 9:Fineness gene storehouse GBL update module, module 10:5th
Middle population pop5 sets up module and module 11:Current population optimum individual is chosen and decoder module;
Described module 1, for realizing following functions:
Gathering N number of time point investigates the field meteorological condition and population dynamic of small grain aphid, obtains long for wheat
The field investigation data that pipe aphid population quantity model is set up, by daily maximum temperature, daily mean temperature, intra day ward, wheat fertility
Stage, small grain aphid natural enemy quantity, and small grain aphid total amount, as the sample data of modeling, are designated as matrix A;
Described module 2, for realizing following functions:
Initiation parameter is set, and produces initialization population, using initialization population as current population pop;Described parameter bag
Include population scale sizepop, maximum iteration maxgen, functor set F, full stop set T, mirror image function symbol set M,
Connector between mrna length, chromogene number, chromogene, fall string length, insert string length, mutation probability, fall to go here and there it is general
Rate, insert string probability, recombination probability, Fineness gene storehouse scale m, increase excellent probability, algorithm of tournament selection contest scale p;
Described module 3, for realizing following functions:
Each individual fitness in the current population pop that the matrix A computing module 2 obtained using module 1 is obtained, and preserve most
Excellent individual maxpop and its fitness;
Described module 4, for realizing following functions:
According to each the individual fitness in the current population pop that module 3 is obtained, Fineness gene storehouse (GBL) and poor quality are set up
Characteristic information storehouse (BBL);
Described module 5, for realizing following functions:
Whether current iterations is judged more than maximum iteration maxgen, is then revolving die block 11;Otherwise grasped as follows
Make:First middle population pop1 is obtained to the individual natural evolution operation for performing gene expression programming in current population pop;
Calculate the individual fitness of the described first middle population pop1, at the same record wherein infeasible individual and its gene inferior
Position in the individuality, is saved in matrix badgen;
Described module 6, for realizing following functions:
The functor collection F and full stop collection T obtained using module 2, the infeasible individual in the matrix badgen obtained to module 5
Individuality in corresponding first middle population pop1 carries out " roguing " operation, by the infeasible individual in the first middle population pop1
Become feasible individual, obtain the second middle population pop2;
Described module 7, for realizing following functions:
The Fineness gene storehouse GBL obtained using module 4, the second middle population pop2 obtained to module 6 was performed in individual the intervention
" increase excellent " operation, form the 3rd middle population pop3;
Described module 8, for realizing following functions:
The functor collection F and full stop collection T obtained using module 2, the information of the 3rd middle population pop3 that computing module 7 is obtained
Entropy H;According to comentropy H and the comparative result of the information entropy threshold of setting, the 3rd middle population pop3 obtained to module 7 is performed
Population intervention operation, obtains the 4th middle population pop4;
Described module 9, for realizing following functions:
Merge the 4th middle population pop4 and Fineness gene storehouse GBL, and according to the size of the individual adaptation degree after merging, obtain more
Fineness gene storehouse GBL after new;
Described module 10, for realizing following functions:
According to current optimum individual maxpop and its fitness, the 4th middle population pop4 obtained to module 8 is carried out with essence
The algorithm of tournament selection operation of English strategy, obtains the 5th middle population pop5;Iterations plus 1, and the 5th middle population is made
It is contemporary population pop, into module 5;
Described module 11, for realizing following functions:
Calculate individual fitness in current population pop, choose fitness value it is maximum it is individual as in current population pop most
Excellent individual bestpop, calculates the effective length of each gene of the optimum individual bestpop, according to the effective of each gene
Length travels through each gene from right to left, and decoding obtains the mathematic(al) representation of each gene, then is obtained with the same approach of module 3
The individual expression formula of optimum individual bestpop;By the individual expression formula of described optimum individual bestpop, according to expression formula
In each parameter real-time measurement values, the predicted value of small grain aphid population quantity can be obtained.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510014836.1A CN104598770B (en) | 2015-01-12 | 2015-01-12 | Wheat aphid quantitative forecasting technique and system based on human evolution's gene expression programming |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510014836.1A CN104598770B (en) | 2015-01-12 | 2015-01-12 | Wheat aphid quantitative forecasting technique and system based on human evolution's gene expression programming |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104598770A CN104598770A (en) | 2015-05-06 |
CN104598770B true CN104598770B (en) | 2017-06-09 |
Family
ID=53124548
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510014836.1A Expired - Fee Related CN104598770B (en) | 2015-01-12 | 2015-01-12 | Wheat aphid quantitative forecasting technique and system based on human evolution's gene expression programming |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104598770B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106919645A (en) * | 2017-01-17 | 2017-07-04 | 广西师范学院 | The sight spot meteorological element Intelligent fine Forecasting Methodology at the big scenic spot of complex landform |
CN116862941B (en) * | 2023-09-05 | 2023-11-14 | 天津市城市规划设计研究总院有限公司 | Image edge detection method and system based on improved gene expression programming |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1643509A (en) * | 2002-03-18 | 2005-07-20 | 摩托罗拉公司 | Gene expression programming algorithm |
-
2015
- 2015-01-12 CN CN201510014836.1A patent/CN104598770B/en not_active Expired - Fee Related
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1643509A (en) * | 2002-03-18 | 2005-07-20 | 摩托罗拉公司 | Gene expression programming algorithm |
Non-Patent Citations (6)
Title |
---|
"Co-evolutionary Gene Expression Programming and Its Application in Wheat Aphid Population Forecast Modelling";Chaoxue Wang et al;《ICSI 2014》;20141231;275-283 * |
"Double System Gene Expression Programming and It"s Application in Function Finding Problems";Chaoxue Wang et al;《International Conference on Mechatronics,Control and Electronic Engineering》;20140930;357-361 * |
"一种新型GEP解码方法";王晓 等;《计算机工程与应用》;20120229;第48卷(第3期);43-45 * |
"基于协同进化基因表达式编程的函数发行研究";王超学 等;《计算机工程与应用》;20130930;第49卷(第17期);53-57 * |
"基因表达式编程核心技术研究";左劼;《中国优秀博硕士学位论文全文数据库(博士)-信息科技辑》;20050615;第2005年卷(第2期);I138-16 * |
"基因表达式编程的改进及其在麦芽种群预测模型构建中的应用";张星;《中国优秀硕士学位论文全文数据库-农业科技辑》;20140515;第2014年卷(第5期);D046-32 * |
Also Published As
Publication number | Publication date |
---|---|
CN104598770A (en) | 2015-05-06 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Madani et al. | A Monte-Carlo game theoretic approach for multi-criteria decision making under uncertainty | |
Uribe‐Convers et al. | Shifts in diversification rates linked to biogeographic movement into new areas: An example of a recent radiation in the Andes | |
WO2019033055A1 (en) | Development of complex agricultural simulation models from limited datasets | |
EP3127054A1 (en) | A computer implemented method of deriving performance from a local model | |
CN107292436A (en) | Blue-green alga bloom Forecasting Methodology based on nonlinear kinetics temporal model | |
Nabavi-Pelesaraei et al. | Modeling and optimization of energy inputs and greenhouse gas emissions for eggplant production using artificial neural network and multi-objective genetic algorithm | |
Miao et al. | Automation of leaf counting in maize and sorghum using deep learning | |
Ranjani et al. | Crop yield prediction using machine learning algorithm | |
CN114492191A (en) | Heat station equipment residual life evaluation method based on DBN-SVR | |
CN115829162A (en) | Crop yield prediction method, device, electronic device and medium | |
CN116993000A (en) | Fruit tree pest control prediction method based on random forest and cellular automaton | |
CN104598770B (en) | Wheat aphid quantitative forecasting technique and system based on human evolution's gene expression programming | |
Torres-Martínez et al. | Evolution of specialization in a plant-microbial mutualism is explained by the oscillation theory of speciation | |
CN105913326B (en) | Constrain the Crop growing stage model Cultivar parameter optimization method of sex knowledge and the tactful Genetic Algorithm Fusion of elite individual | |
Al-Hajj et al. | Estimating solar irradiance using genetic programming technique and meteorological records. | |
CN117575094A (en) | Crop yield prediction and optimization method and device based on digital twin | |
CN115994459B (en) | Method, system and equipment for optimizing construction time sequence of mass power grid infrastructure projects | |
Jabjone et al. | Decision support system using artificial neural network to predict rice production in Phimai district, Thailand | |
Utama | The Optimization of the 3-d structure of plants, using functional-structural plant models. case study of Rice (Oryza sativa L.) in Indonesia | |
Marinković et al. | Data mining approach for predictive modeling of agricultural yield data | |
Marndi et al. | Crop production estimation using deep learning technique | |
CN113222288B (en) | Classified evolution and prediction method of village and town community space development map | |
CN112837739B (en) | Hierarchical feature phylogenetic model based on self-encoder and Monte Carlo tree | |
CN109858127B (en) | Blue algae bloom prediction method based on recursive time sequence deep confidence network | |
Watts et al. | New Zealand fern distributions from the last glacial maximum to 2070: a dynamic tale of migration and community turnover |
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 | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20170609 Termination date: 20200112 |
|
CF01 | Termination of patent right due to non-payment of annual fee |