基于psosvm的sci文章

这篇英文文献收录在sci中，采用的PSOSVM进行数据分析
1. Lou et aL. Applied Mathematical Modelling 39(2015)59075916 5909 In this paper, a novel method of phytoplankton abundance prediction and forecast are presented by svr based on pso algorithms. This approach is the use of pso to determine free parameters of SvR, i.e., PSoSVR, which optimize all svr's parameters simultaneously from the training data. Then the monthly phytoplankton abundance data in MSR was used as a case study for the development of the prediction and forecast model to simulate the dynamic change of phytoplankton abundance in Macau Reservoir given a variety of water variables. The measured data from 2001 to 2011 were used to train and test the model. Present study will lead to better understanding of the algal problems in Macau, which will help to develop later guidelines for forecasting the onset of algae blooms in raw water resources The rest of this paper was organized as follows: The study area and data source of msr were first introduced concisely, followed by modeling approaches(SvR and pso)formulation, and the performance indicators used for evaluation in Section 2. Section 3 presented and discussed the modeling results performed by pso and svr respectively, and made com parisons between both methods. The conclusion was drawn finally in Section 4. 2. Materials and methods 2. 1. MSR and water parameters measurement Macau is located at south of China with a subtropical seasonal climate that is greatly influenced by the monsoons. As the biggest reservoir in Macau, Macau Main Storage Reservoir (msr)(11333/12"E longitude and 22 12/12"N latitude)(as shown in Fig. 1)is located in the east of Macau peninsula with the capacity of about 1.9 million m and the water surface area of 0.35 km In recent years, algal blooms and the situation appeared to be worsening were reported(macao Water Co. Ltd, unpublished data). Macau Water Supply Co. Ltd is responsible for waterquality monitoring and management. Locations in the inlet, outlet and mid of the reservoir was selected for sampling. Samples were collected in duplicate monthly from May 2001 to February 2011 at 0.5 m from the water surface. There are 23 water quality parameters, including hydrological, physical, chemical and biological parameters, were monitored monthly and these 23 parameters were measured according to the standard methods [28, 29]. Imported volume, exported volume and water level were recorded by the inlet and outlet flow meters, based on which the hydraulic retention time(hrt) can be calculated The phytoplankton samples were fixed using 5% formaldehyde and transported to laboratory for microscopic counting In order to identify the water parameters that were significantly correlated with phytoplankton abundance, correlation analysis was conducted firstly The criteria for selecting parameters as the inputs to the psosvr models is that correlation coefficients for selected parameters needs to be greater than 0.3. It should be pointed out that in forecast models the param eters are different from those in the prediction models, because the water parameters in previous data were also used in the correlation analysis. In this study the forecast model was based on the last three months data. Including the three month data in this forecast model is to adopt the historical effect of the last year that have similar environmental conditions such as temperature influence the growth of phytoplankton 2. 2. Support Vector Regression (SVR) The svr is to map the feature vector x into a higher dimensional space by using a nonlinear mapping (x). Assume the f(x) takes the linear form f(X,)=(·x)+ba→(),b), Chi s 十 Macau 1. Location of the msr 5910 L. Lou et aL. / Applied Mathematical Modelling 39(2015)59075916 Where x is the independent variables, y is the dependent one. o refers to a vector for regression coefficients, b is model error value. The coefficients(w and b) are estimated by minimizing the following regularized risk function Rm(x)=1∑C1f(xk,m)+C.Sf),x∈A 2) n∑>(ykf(x,x)l+c·S(f) where is a function of parameters a in domain n (0, c)where Su) is the smoothness of functions f and c is the tradeoff coef ficient between the empirical risk and smoothness of functions f. SVM is designed to find the a that minimizes the risk Function Rsvm(o) Ce=lyk f(xk, ae= yk f(xk, a) lYkf(Xk, a)I<& otherwise (3 Ca is called oinsensitive Cost Function, C(y, f(x, a)), is define to evaluate the discrepancy between an observation yk and the response of learning machine f(xk, a). Determining the tradeoff between the training error and the model flatness, n is the number of training samples, a is a prescribed parameter controlling the tolerance to error Both C and 1 are userdetermined parameters. Two positive slack variables s and s*, which represent the distance from actual values to the corresponding boundary values of 1tube, are introduced. Then, (4) is transformed into the following convex optimization problem minimIze 2o2+C∑(1+) y1()·x1)b≤E+ (4 subject to(Oxi)+by;<&+5 By introducing lagrange multipliers and exploiting the optimality constraints the decision function given by (1) has the following explicit Start Set scope of three parameters of SVM and Stopping condition of the whole program Initialize Pso particles and velocities of a population Evaluate the fitness value Update particle best and global best Update particle velocity and position Does it meet the stopping condition Yes Output optimal parameters of SVM Fig. 2. A flow diagram of the proposed nonlinear ensemble forecasting model 1. Lou et aL. Applied Mathematical Modelling 39(2015)59075916 5911 f(x)=(O(x)+b=∑(11)(x)x)+b=∑(1)k(xx)+b, Where, K(xi, xi)is called the kernel function, li and ii are the Socalled Lagrange multipliers. Kernel function is an inner pro duct of two transformed feature vectors and it is written by k(xi xi)=p(xi).p(x). In the literature, there are several kernel functions, namely linear, polynomial, and Gaussian kernels. Generally, using Gaussian function will yield better prediction performance. Thus, in this work, the Gaussian function K(xi, xi)=exp(x;x l2/(202)), is used in the SVR. Where, o2 repre sents the bandwidth of gaussian kernel To build a svr model efficiently we need to select three positive parameters, o, s and C, of a SVr model These parameters must be selected accurately, since they determine the structure of high dimensional feature space and govern the complexity of the final solution. However, structural methods of selecting parameters efficiently are lacking Thus, despite its superior features, SVR is limited in some research and applications, because the user has to define various parameters appropriately. An excessively large value for parameters in SVR leads to overfitting or while a disproportionately small value leads to underfitting [18, 20 Different parameter settings can cause significant differences in performance Therefore, selecting the optimal hyperparameter is an important step in svr design [ 20, 30. 2.3. Particle Swarm Optimization(PSO) In PSO, a particle's neighborhood is the subset of particles with which it is able to communicate. Depending on how the neighborhood is determined, the pso algorithm may embody the gbest and lbest models. In the former, each particle is con nected to every other in the swarm, thus obtaining information from the whole group In the latter a particle is not able to communicate with the entire swarm, but only with some selected particles. The particle Swarm Optimization concept consists of, at each time step, changing the velocity of (accelerating) each par ticle toward its pbest and lbest locations (local version of PSO). Acceleration is weighted by a random term, with separate ran dom numbers being generated for acceleration toward pbest and lbest locations. In the standard PSo model, with M particles in Ddimensional problem space, each particle is denoted as Xi=( xil, Xi,.., Xp,i=1, 2, ., M which is represented as a potential solution the velocity of each particle along with each dimension is denoted as vi=(vil, v2,.. vip The velocity and position of particle i at(t+ 1) iteration are updated by the fol lowing equations vi(t+1)=wvi(t)+Ciri(t)PCu(t)Xi(t)]+C2r2(t)[P, (t)Xu(t)], X(t+1)=X(t)+v(t+1), Table 1 Correlation analysis of prediction and forecast model Parameters Prediction model Forecast model Time lagged(month) t t3 Turbidity 0.03 0.01 0.06 Temperature 0.19 0.19 0.14 PH 0.49 0.42 0.38 0.33 0.08 0.01 0.14 0.21 0.01 0.22 0.28 0.03 0.03 0.14 0.22 Sio 0.3 0.31 0.16 0.04 Alkalinity 0.34 0.3 0.2 0.12 HCO 0.4 0.24 0.39 0.34 0.29 0.22 0.22 0.15 0.1 0.08 0.02 0.03 NH 0.11 0.08 0.25 TN 0.68 0.6 0.53 0.46 254 0.56 0.55 0. 0.47 0.14 0.08 0.02 0.08 0.02 Suspended solid 0.3 OC 0.38 0.33 0.29 0.35 HRT 0.12 0.13 0.16 Water level 0.13 0.05 0.01 0.02 Precipitation 0.09 0.05 0.11 0.06 Phytoplankton abundance 0.82 0.71 0.62 5912 L. Lou et aL. / Applied Mathematical Modelling 39(2015)59075916 where vi is velocity measures for particles; w is inertial weight controlling velocity direction(a measure of how much the particle"trusts"its own exploration ); C1 and C2 are acceleration coefficients; ri and r2 are random numbers uniformly dis tributed between [0, 1]. Thus C,r represents a uniformly distributed random number from 0 to C1. It is a measure of how much a particle"trusts"its neighborhood best velocity, and c2r2 represents a uniformly distributed random number from 0 to C2. Independent from Cirl, it is a measure of how much a particle"trusts"the global velocity; p_ci is the neighborhood best position. ""represents the difference of two positions is the velocity that will transform the second position into the first position p g; is the global best position. " represents the transformation of a position using the velocity(yields a posi tion). x is position measure of for particles Thus, the change of position of each particle from one iteration to another can be computed according to the distance between the current position and its previous best position, and the distance between the current position and the best posi tion of swarm. The particle flies through potential solutions towardsp_cit)andp(t in a navigated way while still exploring new areas by the stochastic mechanism to escape from local optima. Since there was no actual mechanism for controlling the velocity of a particle, it is necessary to impose a maximum value vmax on it. If the velocity exceeds the threshold, it is set equal to vmax, which controls the maximum travel distance at each iteration to avoid this particle flying past good solutions. The PSO algorithm is terminated with a maximal number of generations or the best particle position of the entire swarm cannot be improved further after a sufficiently large number of generations. The pso algorithm has shown its robustness and effi cacy in solving function value optimization problems in real number spaces 2. 4. The proposed hybrid model integrating SVM and pSo The research scheme of the proposed methodology is shown in Fig. 2 First, the population of particles is initialized, each particles is represented as(τ1,τ2,τ3), whereτ1,τ2andτ3 denote the regularization parameters C, 02 and 8, respectively. The initial group is randomly generated, each group having a random position and a random velocity for each dimension. Second, each particle fitness function of pso (fit ness=MAPe=2(yiYi/Yi where nMApe is the mean absolute percentage error; Yi and Yi are the observation data and the modeling results, respectively) for the SvM is evaluated. The each particles fitness in this study is the regression accuracy. If the fitness is better than the particle's best fitness, then the position vector is saved for the particle. If the par ticle's fitness is better than the global best fitness, then the position vector is saved for the global best. Finally the particles velocity and position are updated until the termination condition is satisfied PSOSVR is applied in the proposed svr software reliability prediction model to optimize parameter selection 2.5 Performance indicators The performance of models was evaluated using the following indicators: coefficient of determination(r)that provides the variability measure for the data reproduced in the model Prediction R is a good measure for both comparison and seeing the model's prediction capability The calculation method is also known as the cross validation, in which we exclude the first observation and build the model with the remaining ones, use this model to predict the excluded observation and repeat for all observations. It is a good measure for outofsample accuracy. As this test cannot give the accuracy of the model, other statistical parameters should be reported. Mean absolute error(Mae) and root mean square error(rmse) that measure residual errors, providing a global idea of the difference between the observation and modeling. the indicators were defined as below by Eqs. 812 F (8) F 9) ∑(Y1Y) (10) Table 2 Performance indexes of the prediction and forecast models Prediction model Forecast model Performance index Accuracy performance Generalization Accuracy performance Generalization (Training set) performance(Testing set) (Training set) performance(Testing SVM 31 SVM+ PSo SVM[31 SVM+ PSo SVM 31 SVM+ PSO SVM[31 SVM+PSO R2 0.760 0.764 0.758 0.767 0.863 0.875 0.863 0.876 RMSE 0.307 0.291 0.351 0.343 0.229 0.219 0.264 0.258 MAE 0.243 0.236 0.274 0.267 0.127 0.120 0.226 0.224 1. Lou et aL. Applied Mathematical Modelling 39(2015)59075916 5913 8.5 Observed PSOSVM 87.5 6.5 20012002 2003 2004 2005 2006 20072008 2009 Fig 3 Observed and predicted phytoplankton level for the training and validation data set of the prediction models 8.5 8 7.5 百65 Observed PSOSVM 2009 2010 Fig. 4. Observed and predicted phytoplankton level for the testing data set of the prediction models 8. R2=0.747239 R2=0.734257 y=0.596*X+3.012 y=0.675x+2309 7.5 7.5 65 6.5 6.5 7.5 8.5 6.5 7.5 8.5 Observed Observed ig. 5. PSoSVM results for the training and validation (a)and testing(b) data set of the prediction model MAE= RMSE n∑(YY), 5914 L. Lou et aL. / Applied Mathematical Modelling 39(2015)59075916 8.5 8 PSOSVM 87.5 6.5 6 Year Fig. 6. Observed and predicted phytoplankton level for the training and validation data set of the forecast model 1 that based on the previous 1st, 2nd and 3rd months data where n is the number of data; Yi and Yi are observation data and the mean of observation data, respectively, and Yi is the modeling results(see Table 1). 3. Results and discussion To illustrate this hybrid prediction and forecast methods, the estimation of phytoplankton abundances based on the water parameters and environmental variables were studied in this work. The correlation of log1o phytoplankton abun dances and all variables for prediction model and forecast model were shown in Table 2. It can be seen that parameters with correlation coefficients greater than 0.3 are highlighted in bold, and these highlighted parameters will be kept in the models Because the water parameters in previous data were also used in the correlation analysis, the parameters selected in forecast models are different from those in the prediction models. In the forecast models of svr, phytoplankton abundance(t) is a function in terms of water parameter(t1), water parameter(t2)and water parameters(t3). At here t1, t2 and t3 represent the 1 month, 2 months and 3 months before the time t. So, in this way only 9 parameters used in the prediction models while 23 timelagged parameters selected for the forecast models When the correlation analysis is done it needs to test the models which cover two parts, the accuracy performance and the generalization performance. Accuracy performance is to check how capable the model is to predict the output for the given input set that originally used to train the model For generalization performance is to test if the model is capable of predicting the output for the given input sets that were not used to train the model. It is not good if model that is memorizing the inputs instead of generalized learning, so in order avoid this problem both performance checks should be covered In this work, 50 runs were conducted to average the performance indexes for Svrbased models. So for the predication models in the application of Svr in this work, after the correlation analysis, as the independent variables, there are 9 parameters including pH, SiO 2 are selected, and phytoplankton abundance are selected as the induced variable to set a target. And then, the data process for both the prediction and forecast model follows the same steps as illus trated in 31. After the data were processed, the performance of prediction and forecast models were shown in Table 2. The results indicated that the svr were successful in the prediction and forecast phytoplankton abundance in msr, with the r greater than 0.87 for both training and testing data sets. Compared to the prediction model with the r2 of 0.764(0.767) 8.5 8 8 6.5 Observed PSOSVM 2009 2010 2011 Yea Fig. 7. Observed and predicted phytoplankton level for the testing data set of the forecast model 1 that based on the previous 1 st, 2nd and 3rd months data 1. Lou et aL. Applied Mathematical Modelling 39(2015)59075916 5915 8.5 8.5 R4=0.874729 R2=0.876301 y=0.818*×+1.290 8 0.869*x+0.809 7.5 6.5 a 6.5 7.5 8.5 Observed 7.5 Observed Fig 8. PSoSVM results for the training and validation(a)and testing(b)data set of the forecast model 1 that based on the previous 1st, 2nd and 3rd months data RMSE of 0. 291(0.343), and MAE of 0. 236(0.267)for training(testing), the forecast model had better performance with the r2 of 0. 875(0.876), RMSE of 0. 219(0.258), and MAe of 0. 120(0.224)for training(testing), suggesting that the historical water parameters including the phytoplankton abundance has effect on the prediction, which can improve the prediction powers These results further confirmed the historical effects on the model accuracy and generalization performance, and also implied that take the latest 3 months data as memorizing learning can improve the prediction power in the forecast model Besides, further compared with our previous study [31] for the prediction and forecast of phytoplankton abundance using SVR only with R of 0.863, the present study integrating SVR and PSo as the optimization algorithm have better prediction power with r, RMse and mae The observed data versus the modeling data were shown in Fig. 5 (prediction model), and Fig 8(forecast model 1), and the observed and modeling phytoplankton abundance change over time were listed in Figs. 3 and 4 (prediction model) and Figs 6 and 7(forecast model 1). These results confirmed that integrating SVM and pSo can handle well the nonlinear rela tionship between water parameters and phytoplankton abundance From the modeling point of view, a disadvantage of learning machine method is the mechanisms of the inner signal processing are not unknown. However, it has provided enough information for msr to prevent the algal bloom problems, for controlling the algal bloom problem occurrence, the engineers can only control the predicted or forecast phytoplankton by adjusting the operational variables, such as reduc ing the nitrogen and phosphorus concentrations without understanding the complete mechanisms and the relationships among the variables. The forecast model integrating PSO and SvR showed high values of RMSe and R and simple algorithm which is highly probably due to the more complete algal bloom causative variables involved, thus providing more information in explaining the algal bloom phenomena, despite that the complete mechanisms causing algal bloom and the relationships among vari ables are still unclear 4. Conclusions To avoid algal blooms in drinking water lakes or reservoirs, it is very crucial to have accurate the phytoplankton abun dance prediction and forecast. This study proposes the psosvm based prediction and forecast models for monthly phyto plankton abundance time series in MSr pSo was applied to obtain the optimal parameters of svr. It can be seen that the forecast model have better performance with the r of up to 0.876 than prediction model with the r of 0.767. This means the algal bloom problem is a complicated nonlinear dynamic system that is affected not only by the water variables in cur rent month but also by those in a couple of previous months. Furthermore, SvR in the study showed superior forecast power and root mean square errors with the latest 3 months data in the forecast model, and this indicates that the historical water parameters and phytoplankton abundance have impact on the phytoplankton dynamics of the reservoir. From numerical results, it can be concluded that this work proposes an effective approach for water system modeling and management, and it can also be noted that this approach can be further improved by combining other numerical approaches and optimiza on algorithms such as genetic algorithms [3234 Acknowledgments We thank macao Water Co Ltd. for providing historical data of water quality parameters and phytoplankton abundances The financial support from the fundo para o desenvolvimento das Ciencias e da Tecnologia(fDct)(grant FDCT/069/2014 5916 L. Lou et aL. / Applied Mathematical Modelling 39(2015)59075916 A2)and Research Committee at University of Macau are gratefully acknowledged. This project was also supported by University of Macau MYRG201400074F References [1Z. Selman, S Greenhalgh, R Diaz, Eutrophication and Hypoxia in Coastal Areas: A Global Assessment of the State of Knowledge, World Resources Inst Washington D.C., 2008 [2]S W. Fei, M. Wang, Y.B. Miao, J. Tu, C L Liu, Particle swarm optimizationbased support vector machine for forecasting dissolved gases content in power transformer oil, Energy Convers Manage. 50(2009)16041609 [3]MR. Amiryousefi, M. Mohebbi, F. Khodaiyan, S Asadi, An empowered adaptive neurofuzzy inference system using selforganizing map clustering to predict mass transfer kinetics in deepfat frying of ostrich meat plates, Comput. Electron. Agric. 76(2011)8995 [4]IG. Hatvania, J. Kovacsa, L.S. Kovacsb, P. Jakuschc, J. Korponaid, Analysis of longterm water quality changes in the KisBalaton water protection system with time series, cluster analysis and Wilks' lambda distribution, Ecol. Eng. 37(2011)629635 [5]CS Li, J W. Hu, A new ARIMAbased neurofuzzy approach and swarm intelligence for time series forecasting, Eng. Appl. Artif Intell. 25 (2012)295 308 [6]J Pallant, I Chorus, J. Bartram, Toxic cyanobacteria in water, SPSS Survival Manual, McGraw Hill, 2007 [7 G B Huang O.Y. Zhu, C K Siew, Extreme learning machine: a new learning scheme of feed forward neural networks. IEEE Int Conf Neural Networks Conf.Proc.2(2004)985990 [8G B Huang, Q.Y. Zhu, C K Siew, Extreme learning machine: theory and applications, Neurocomputing 70(2006)489501 9]G B Huang, L Chen, C K Siew, Universal approximation using incremental constructive feedforward networks with random hidden nodes, IEEE Trans Neural Networks 17(2006)879892 [10] H.G. Han, J.F. Qiao, Prediction of activated sludge bulking based on a selforganizing RBF neural network, J Process Control 22(2012)11031112 [11]L. Belanche, J.J. Valdes, J Comas, I.R. Roda, M. Poch, Prediction of the bulking phenomenon in wastewater treatment plants, Artif Intell. Eng. 14(2000) [12]R HechtNielsen, Kolmogorov's mapping neural network existence theorem, in: Proceedings of 1st IEEE International Joint Conference of Neural Networks, New York, 1987, pp. 1113 [13]N Cristianini, ShaweTaylor, An introduction to support vector machines and other kernelbased learning methods, first ed, Cambridge University Press, Cambridge, 2000 [14]VN Vapnik, The Nature of Statistical Learning Theory, second ed, Springer, New York, 2000 [15] R.C. Eberhart, ]. Kennedy, A new optimizer using particle swarm theory, Proceedings of the Sixth International Symposium on Micro Machine and Human Science(1995)3943 [16SW. Lin, K.C. Ying, S.C. Chen, Z.J. Lee, Particle swarm optimization for parameter determination and feature selection of support vector machines, Expert Syst. Appl. 35(2008)18171824 [17J. Kennedy, R.C. Eberhart, Swarm Intelligence, Morgan Kaufmann Press, San Francisco, 2001 [18S.S Keerthi, Efficient tuning of SVM hyperparameters using radius/margin bound and iterative algorithms, IEEE Trans. Neural Network 13(2000) 12251229 [19 K Duan, S Keerthi, A Poo, Evaluation of simple performance measures for tuning SVM hyper parameters, Neurocomputing 51(2003)4159 [20] P T. Lin, Support Vector Regression: Systematic Design and Performance Analysis(Doctoral dissertation), Department of Electronic Engineerin National Taiwan University of Science and Technology, PhD Thesis, 2001 [21] S.W. Fei, M. Wang, Y B. Miao, ]. Tu, C L. Liu, Particle swarm optimizationbased support vector machine for forecasting dissolved gases content in power transformer oil, Energy Convers Manage. 50(2009)16041609 [22 W.C. Hong, Chaotic particle swarm optimization algorithm in a support vector regression electric load forecasting model, Energy Convers Manage. 50 (2009)105117. [23]X Tang, L Zhuang, C Jiang, Prediction of silicon content in hot metal using support vector regression based on chaos particle swarm optimization, Expert Syst.Appl.36(2009)(2009)1185311857 [24 H B Zhao, S Yin, Geomechanical parameters identification by particle swarm optimization and support vector machine, Appl. Math. Modell. 33(2009) 39974012. [251Q. Wu, R. Law, Cauchy mutation based on objective variable of Gaussian particle swarm optimization for parameters selection of SvM, Expert Syst Appl.38(2011)64056411 [26]Q Wu, S.Y. Wu, ]. Liu, Hybrid model based on SVM with Gaussian loss function and adaptive Gaussian PSO, Eng. Appl. Artif Intell. 23(2010)487494 [27Q. Wu, R. Law, Complex system fault diagnosis based on a fuzzy robust wavelet support vector classifier and an adaptive Gaussian particle swarmoptimization, Inf. Sci. 180(2010)45144528 [28 APHA, Standard Methods for the Examination of Water and Wastewater, A W. W. A a. W.E. F, 2002 [29] G B Huang, L Chen, Convex incremental extreme learning machine, Neurocomputing 70(2007)30563062 [30]N. Cristianini, ShaweTaylor, An introduction to support vector machines and other kernelbased learning methods, first ed, Cambridge University Press, Cambridge, 2002. [31]Z. Xie, I. Lou, W.K. Ung, K M. Mok, Freshwater algal bloom prediction by support vector machine in Macau storage reservoirs, Math. Prob. Eng. 2012 (2012)397473 [32 C Cattani, S Chen, G. Aldashev, Information and modeling in complexity, Math. Prob. Eng 2012(2012)868413 [33]S Chen, Y Zheng, C Cattani, W. Wang, Modeling of biological intelligence for SCM system optimization, Comput. Math. Methods Med. 2012(2012) 769702, [34]P Lu, S Chen, Y. Zheng, Artificial intelligence in civil engineering, Math. Prob. Eng 2013(2013)145974
 41KB
PSO_SVM.zip
20190706程序可用，自带训练和测试数据集，每个模块是单独编，可读性强，利用率高。
 44.32MB
PSO优化SVM参数matlab实例
20170924matlab环境下使用PSO算法对SVM多分类器的参数进行优化的案例，代码有详细的注释，另有一篇博客对算法的大致过程有介绍.
 9KB
（psoSVM）的MATLAB程序
20181101这是pso优化的支持向量机程序，主要用于各种数据的分类
训练1400个样本数据，用pso优化SVM大概需要多长时间？还是说，这个时间和训练数据的数量有关？（我用的matlab)_course
20190920最近在用pso优化svm中的参数,用matlab运行的时间比较长，大概好几分钟吧，想问下这个优化一般需要多长时间？ 还是说我的程序不好，所以运行时间长。
pso_svm.mmatlab程序。下载_course
20200713pso_svm算法，里面有注释，每个模块是单独编，可读性强，利用率高。里面的数据可能没有，你需要可以自己去数据库下载。 相关下载链接：//download.csdn.net/download/lccn
 2KB
粒子群（PSO）优化SVM做预测
20180707粒子群优化SVM的两个参数，利用数据做预测，数据代码都在压缩包里，可以直接运行。
 6KB
PSO—SVM
20121101SVM PSO 通过粒群算法择优选择支持向量机的敏感参数
 3KB
PSOSVM 粒子算法优化支持向量机MATLAB代码
20180321用粒子群算法优化支持向量机的matlab程序，用于对股价、经济的预测作用，优化后预测精确
 813KB
matlab pso算法包含使用说明
20181212关于pso算法的matlab实现，可以方便初学者学习，内包含多套学习资料，可以有效进行学习等
 9KB
各种改进的PSO算法代码matlab.zip
20130424一些关于粒子群优化算法的matlab程序实现，里面有各种的PSO改进算法
 44.32MB
PSO优化SVM参数
20180511PSO优化SVM参数 PSO优化SVM参数 PSO优化SVM参数 PSO优化SVM参数
 13KB
PSOLSSVM代码
20130419PSOLSSVM代码
 56KB
pso+svm预测鲍鱼的年龄.zip
20190712通过pso（粒子群算法）优化支持向量机回归模型中的惩罚系数和gamma系数，用于预测鲍鱼的年龄。数据集也在文件中，模型还不够完善，下次改进再发出来，不过能够运行，准确度还是有所提高，只是计算时间稍长，
 712KB
SVM实现负荷预测，其中包含基本SVM，PSO优化SVM，改进SVM三种方法
20180917简单易读的SVM负简单易读的SVM负荷预测实验，并包含PSO、改进PSO等多种方法简单易读的SVM负荷预测实验，并包含PSO、改进PSO等多种方法简单易读的SVM负荷预测实验，并包含PSO、改进PSO
PSO+SVM基于matlab下载_course
20191002PSC最优寻参的SVM支持向量机的 nu 参数(取值越小，异常点就越少) 相关下载链接：//download.csdn.net/download/qq_41186145/10445269?utm_so

博客
python的hanoi并输出移动步数
python的hanoi并输出移动步数

博客
蚂蚁金服研发岗三面全过程，附阿里年薪技能图
蚂蚁金服研发岗三面全过程，附阿里年薪技能图

学院
数据分析基础班 05期直播回放
数据分析基础班 05期直播回放

学院
方法的定义和使用 视频
方法的定义和使用 视频

博客
20201127
20201127

学院
Laravel Excel 快速入门课程
Laravel Excel 快速入门课程

学院
Xlua原理详解（C#与Lua交互，Unity热更新）
Xlua原理详解（C#与Lua交互，Unity热更新）

博客
lineheight：1
lineheight：1

博客
failed to open stream: php_network_getaddresses: getaddrinfo failed: Name or service not known
failed to open stream: php_network_getaddresses: getaddrinfo failed: Name or service not known

博客
20201127
20201127

学院
Java Web快速入门
Java Web快速入门

博客
20201127
20201127

下载
Dynamics CRM工具插件
Dynamics CRM工具插件

下载
单片机与DSP中的TI新推音频DSP TAS3108
单片机与DSP中的TI新推音频DSP TAS3108

博客
佳能Canon LBP3910 打印机驱动
佳能Canon LBP3910 打印机驱动

下载
通信与网络中的Atmel 推出独立LIN2.0转发器
通信与网络中的Atmel 推出独立LIN2.0转发器

学院
thinkphpCms系统
thinkphpCms系统

下载
webpack入门必知必会
webpack入门必知必会

学院
“动画讲解+手写代码”走进数据结构+算法
“动画讲解+手写代码”走进数据结构+算法

学院
接口测试入门到实战精通【软件测试】
接口测试入门到实战精通【软件测试】

下载
美术学院毕业生英文简历模板
美术学院毕业生英文简历模板

学院
社群本质拆解25讲——为什么90%的人认为的「社群」概念都是错的？
社群本质拆解25讲——为什么90%的人认为的「社群」概念都是错的？

学院
小程序开发快速入门
小程序开发快速入门

下载
RedisDesktopManagerfor2020.6.0.0Windows
RedisDesktopManagerfor2020.6.0.0Windows

学院
【数据分析实战训练营】数据可视化
【数据分析实战训练营】数据可视化

学院
Net网站开发必学知识点（基于c#和razor）、razor教程
Net网站开发必学知识点（基于c#和razor）、razor教程

学院
VMware Converter P2V SQL Cluster
VMware Converter P2V SQL Cluster

下载
电源技术中的飞思卡尔三款射频LDMOS产品符合WiMAX基站所需性能
电源技术中的飞思卡尔三款射频LDMOS产品符合WiMAX基站所需性能

学院
大白话系列算法
大白话系列算法

学院
Windows Server 2016 NLB集群部署实战
Windows Server 2016 NLB集群部署实战