CN114398954A - Cloud server load prediction method based on hybrid optimization strategy extreme learning machine - Google Patents

Abstract

The invention discloses a cloud server load prediction method based on a hybrid optimization strategy extreme learning machine, which comprises the following steps: 1) establishing an extreme learning machine model; 2) constructing a whale optimizer, and 3) updating the whale optimizer by using a diversity mixing optimization strategy; 4) introducing the migration strategy into a whale optimizer to obtain a whale optimizer with a mixed strategy; 5) obtaining historical load data of a cloud server, and training a mixed strategy whale optimizer to obtain an optimal hyperparameter P; 6) inputting the optimal hyper-parameter into an extreme learning machine model, and establishing a cloud server load prediction model; 7) and acquiring current cloud server load data, inputting the current cloud server load data into a cloud server load prediction model, and predicting the future load of the cloud server. The invention provides a cloud server load prediction method based on a hybrid optimization strategy extreme learning machine, which can greatly improve the prediction precision of cloud server load.

Description

A cloud server load prediction method based on hybrid optimization strategy extreme learning machine

technical field

The invention relates to the field of cloud computing, in particular to a cloud server load prediction method based on a hybrid optimization strategy extreme learning machine.

Background technique

With the rapid development of communication technology, the number of various types of mobile terminal equipment has been increasing, and various software application services have also emerged, which directly leads to the explosive growth of the amount of data in the Internet, making traditional data processing into a bottleneck . Cloud computing has powerful computing resources, and can process massive data in a short time through distributed parallel computing. However, how to achieve efficient resource management has always been the most concerned issue for cloud service operators and suppliers. As one of the necessary measures to effectively monitor cloud network conditions, cloud load prediction has become a hot topic in the field. The more accurate the prediction results, the higher the resource utilization, the faster the response speed, and the better quality Service quality and higher user satisfaction rates. At the same time, accurate load prediction will also lay a solid foundation for cloud resource management and bring huge economic benefits to cloud service operators and suppliers.

A large number of cloud server load prediction methods have been proposed in previous studies, and some of them have been successfully applied in the industry. Generally speaking, these forecasting methods can be roughly divided into two categories, namely methods based on time series models and methods based on machine learning models. The method based on time series model is to convert non-stationary series into stationary series, but the time series data must be strictly continuous. At the same time, the data needs to be preprocessed before calculation, which makes the calculation process relatively complicated and is not suitable for dealing with data loss and other situations. Methods based on machine learning models include support vector machines, neural networks and Gaussian process regression, etc. In practice, although cloud load data has high complexity, machine learning models can effectively predict cloud load resources with nonlinear changes . In addition, the methods based on machine learning models have better generalization ability and mapping ability compared with methods based on time series forecasting models.

Extreme learning machine is a type of machine learning model based on feedforward neural network. Compared with other shallow machine learning models such as single-layer perceptron and support vector machine, extreme learning machine has more advantages in learning rate and generalization ability. However, the parameter setting of the extreme learning machine model has a great influence on the prediction effect of cloud load, and how to choose the optimal parameters is still a difficult problem.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide a cloud server load prediction method based on a hybrid optimization strategy extreme learning machine, comprising the following steps:

1) Establish an extreme learning machine model.

The extreme learning machine model includes an input layer, a hidden layer and an output layer. The input layer, hidden layer, and output layer are connected by neurons, and the number of neurons in each layer is u, r, and e. The weight coefficient between the hidden layer and the input layer is S, and the weight coefficient between the hidden layer and the output layer is P. θ is the set of neuron thresholds in the hidden layer,

Among them, the weight coefficient S, the weight coefficient P, and the set θ are respectively as follows:

In the formula, s _ru is the weight coefficient between the rth hidden layer neuron and the uth input layer neuron. _pre is the weight coefficient between the rth hidden layer neuron and the eth output layer neuron. θr is the threshold of the _rth hidden layer neuron.

The activation function of the extreme learning machine model is y( ), and the output is F, that is:

F=[f ₁ ,f ₂ ,...,f _M ] (4)

In the formula, the weight coefficient s _i =[s _i1 ,s _i2 ,...,s _iu ]. a _i =[a _1j ,a _2j ,...,a _uj ] ^T is the i-th column of the input matrix A. Transpose matrix F ^T =LP of output F.

Among them, the hidden layer output matrix L is as follows:

2) Build the whale optimizer,

The whale optimizer includes a foraging and encirclement phase, a random search phase, and a bubble-net attack phase.

During the feeding and hugging phase, the whale's position is updated as follows:

是最佳鲸鱼个体位置。K和Q是系数向量。系数向量K＝2k·n-k。参数k＝2-(2t)/MAX_iterβ。系数向量Q＝|J·pos(t)-pos(t)|。参数J＝2n。n是[0,1]中的随机向量。In the formula, pos(t+1) is the predicted position of the individual whale. t is the current iteration number.

is the best individual whale position. K and Q are coefficient vectors. The coefficient vector K=2k·nk. Parameter k=2-(2t)/MAX _iter β. The coefficient vector Q=|J·pos(t)-pos(t)|. Parameter J=2n. n is a random vector in [0,1].

During the random search phase, the whale's position is updated as follows:

pos(t+1)=pos _rand (t)-K·Q(8)

Q=|J·pos _rand (t)-pos(t)|(9)

where pos _rand (t) is the position of a random whale. When |K| ≤ 1, the optimal individual is selected as the solution to achieve the local optimization of the whale optimizer. When |K| > 1, individuals are randomly selected as solutions to achieve the global optimization of the whale optimizer.

During the bubble-net attack phase, the whale's position was updated as follows:

是鲸鱼个体与迄今为止获得的最佳个体之间的距离。o是定义螺旋形状的常数。l是范围在-1与1之间的数。In the formula,

is the distance between the whale individual and the best individual obtained so far. o is a constant that defines the shape of the spiral. l is a number in the range -1 to 1.

3) Update the whale optimizer with various mixed optimization strategies. v is a random number;

The steps to update the whale optimizer with a diverse mix of optimization strategies include:

3.1) Initialize the whale population using the Levy strategy. The mathematical model of the Levy policy is as follows:

where λ is the random step size.

where the parameter μ and the parameter η follow a normal distribution, namely:

The parameters δ _μ and δ _η are respectively as follows:

where Γ(·) is the standard gamma function.

3.2) Based on the Levy strategy, calculate the initial position of the individual, namely:

where L _b is the lower limit, U _b is the upper limit, and Levy(v) is a random vector with a step size that obeys the Levy distribution.

3.3) Introduce a nonlinear convergence factor β in the whale optimizer, namely:

where Max _iter is the maximum number of iterations.

4) Introduce the migration strategy into the whale optimizer to get the mixed strategy whale optimizer.

After introducing the migration strategy into the whale optimizer, the position of the whale swarm is updated as follows.

pos(t+1)=B*pos(t)+pos(t)*randn(0,σ ² ) (16)

In the formula, B is the convergence coefficient, and randn(0,σ ² ) conforms to the Gaussian distribution. B _max and B _min are the upper and lower limits of the convergence coefficient B.

5) Obtain the historical load data of the cloud server, and train the hybrid strategy whale optimizer to obtain the optimal hyperparameters.

The cloud server load data is normalized.

The cloud server load data includes a training set and a test set. Among them, the training set is used to train the mixed strategy whale optimizer, and the test set is used to test the mixed strategy whale optimizer.

The training set of the extreme learning machine model includes input matrix A and output matrix B, namely:

In the formula, M is the number of cloud server load data training sample sets.

During the training process of the mixed strategy whale optimizer, the weight coefficient s and the neuron threshold θ are constants, and the weight coefficient P is an unknown parameter. The weight coefficient P is solved by the least square method.

6) Input the optimal hyperparameters into the extreme learning machine model to establish a cloud server load prediction model.

7) Obtain the current cloud server load data, and input it into the cloud server load prediction model to predict the future load of the cloud server.

It is worth noting that the present invention considers that the cloud server load data has obvious nonlinear and non-stationary characteristics, and the method based on the time series model has relatively poor applicability, so the present invention proposes a hybrid optimization strategy extreme learning machine. Predict cloud server load. The extreme learning machine model with strong nonlinear mapping ability and generalization ability is used. Because the gradient descent method is not used to update the parameters in the cloud server load prediction, compared with the traditional neural network model, the Reduce model complexity and training time.

In addition, the final prediction accuracy is affected by the parameter settings of the extreme learning machine model. Therefore, the present invention proposes a hybrid optimization strategy based on the whale optimizer to determine the hyperparameters of the extreme learning machine model. Specifically, the present invention uses the Levy strategy to initialize the population, and the random walk mode of the Levy strategy expands the search range of the whale group and effectively avoids individual whales falling into local extreme values. Therefore, it helps the whale optimizer to have better convergence performance. At the same time in the foraging and round-up phases, since the whale optimizer converges at the same speed in the early and late stages, this greatly reduces the overall convergence speed of the optimization. Therefore, a nonlinear convergence factor is introduced in the whale optimizer to solve this problem.

The technical effect of the present invention is beyond doubt. The invention provides a cloud server load prediction method based on a hybrid optimization strategy extreme learning machine, which can greatly improve the prediction accuracy of the cloud server load.

The invention comprehensively considers the nonlinear and non-stationary characteristics of cloud server load data, proposes a hybrid strategy whale optimizer with better optimization performance, and combines the proposed hybrid strategy whale optimizer with an extreme learning machine model. , providing effective and accurate predictions, greatly enhancing the efficiency of cloud server computing resources.

Description of drawings

Figure 1 is a framework diagram of a cloud server load prediction algorithm;

Fig. 2 is the working flow chart of the cloud server load prediction algorithm;

Detailed ways

The present invention will be further described below in conjunction with the examples, but it should not be understood that the scope of the above-mentioned subject matter of the present invention is limited to the following examples. Without departing from the above-mentioned technical idea of the present invention, various substitutions and changes can be made according to common technical knowledge and conventional means in the field, which shall be included in the protection scope of the present invention.

Example 1:

Referring to Figure 1 to Figure 2, a cloud server load prediction method based on hybrid optimization strategy extreme learning machine, including the following steps:

1) Establish an extreme learning machine model.

The extreme learning machine model includes an input layer, a hidden layer and an output layer. The input layer, hidden layer, and output layer are connected by neurons, and the number of neurons in each layer is u, r, and e. The weight coefficient between the hidden layer and the input layer is S, and the weight coefficient between the hidden layer and the output layer is P. θ is the set of neuron thresholds in the hidden layer,

Among them, the weight coefficient S, the weight coefficient P, and the set θ are respectively as follows:

In the formula, s _ru is the weight coefficient between the rth hidden layer neuron and the uth input layer neuron. _pre is the weight coefficient between the rth hidden layer neuron and the eth output layer neuron. θr is the threshold of the _rth hidden layer neuron.

The activation function of the extreme learning machine model is y( ), and the output is F, that is:

F=[f ₁ ,f ₂ ,...,f _M ] (4)

In the formula, the weight coefficient s _i =[s _i1 ,s _i2 ,...,s _iu ]. a _i =[a _1j ,a _2j ,...,a _uj ] ^T is the i-th column of the input matrix A; Transpose matrix F ^T =LP of output F.

Among them, the hidden layer output matrix L is as follows:

2) Build the whale optimizer,

The whale optimizer includes a foraging and encirclement phase, a random search phase, and a bubble-net attack phase.

During the feeding and hugging phase, the whale's position is updated as follows:

是最佳鲸鱼个体位置。K和Q是系数向量。系数向量K＝2k·n-k。参数k＝2-(2t)/MAX_iterβ。系数向量Q＝|J·pos(t)-pos(t)|。参数J＝2n。n是[0,1]中的随机向量。In the formula, pos(t+1) is the predicted position of the individual whale. t is the current iteration number.

is the best individual whale position. K and Q are coefficient vectors. The coefficient vector K=2k·nk. Parameter k=2-(2t)/MAX _iter β. The coefficient vector Q=|J·pos(t)-pos(t)|. Parameter J=2n. n is a random vector in [0,1].

During the random search phase, the whale's position is updated as follows:

pos(t+1)=pos _rand (t)-K·Q(8)

Q=|J·pos _rand (t)-pos(t)| (9)

where pos _rand (t) is the position of a random whale. When |K| ≤ 1, the optimal individual is selected as the solution to achieve the local optimization of the whale optimizer. When |K| > 1, individuals are randomly selected as solutions to achieve the global optimization of the whale optimizer.

During the bubble-net attack phase, the whale's position was updated as follows:

是鲸鱼个体与迄今为止获得的最佳个体之间的距离。o是定义螺旋形状的常数。l是范围在-1与1之间的数。In the formula,

is the distance between the whale individual and the best individual obtained so far. o is a constant that defines the shape of the spiral. l is a number in the range -1 to 1.

3) Update the whale optimizer with various mixed optimization strategies. v is a random number;

The steps to update the whale optimizer with a diverse mix of optimization strategies include:

3.1) Initialize the whale population using the Levy strategy. The mathematical model of the Levy strategy is as follows:

where λ is the random step size.

where the parameter μ and the parameter η follow a normal distribution, namely:

The parameters δ _μ and δ _η are respectively as follows:

where Γ(·) is the standard gamma function.

3.2) Based on the Levy strategy, calculate the initial position of the individual, namely:

where L _b is the lower limit, U _b is the upper limit, and Levy(v) is a random vector with a step size that obeys the Levy distribution.

3.3) Introduce a nonlinear convergence factor β in the whale optimizer, namely:

where Max _iter is the maximum number of iterations.

4) Introduce the migration strategy into the whale optimizer to get the mixed strategy whale optimizer.

After introducing the migration strategy into the whale optimizer, the position of the whale swarm is updated as follows.

pos(t+1)=B ^* pos(t)+pos(t) ^* randn(0,σ ² ) (16)

In the formula, B is the convergence coefficient, and randn(0,σ ² ) conforms to the Gaussian distribution. B _max and B _min are the upper and lower limits of the convergence coefficient B.

5) Obtain the historical load data of the cloud server, and train the hybrid strategy whale optimizer to obtain the optimal hyperparameters.

The cloud server load data is normalized.

The cloud server load data includes a training set and a test set. Among them, the training set is used to train the mixed strategy whale optimizer, and the test set is used to test the mixed strategy whale optimizer.

The training set of the extreme learning machine model includes input matrix A and output matrix B, namely:

In the formula, M is the number of cloud server load data training sample sets.

During the training process of the mixed strategy whale optimizer, the weight coefficient s and the neuron threshold θ are constants, and the weight coefficient P is an unknown parameter. The weight coefficient P is solved by the least square method.

6) Input the optimal hyperparameters into the extreme learning machine model to establish a cloud server load prediction model.

7) Obtain the current cloud server load data, and input it into the cloud server load prediction model to predict the future load of the cloud server.

Example 2:

A cloud server load prediction method based on a hybrid optimization strategy extreme learning machine, comprising the following steps:

1) Divide the cloud server load data, determine the training set and test set of the cloud server load prediction model, and uniformly normalize the data set;

2) Build an extreme learning machine model.

2.1) The whole model consists of three-layer network structure, namely input layer, hidden layer and output layer, which are connected by neurons. For the extreme learning machine model, the number of neurons in each layer is u, r, and e. The weight coefficient between the hidden layer and the input layer is S, and the weight coefficient between the hidden layer and the output layer is P. θ is the set of neuron thresholds in the hidden layer, where,

2.2) The extreme learning machine model has M cloud server load data training sample sets. A is the input matrix and B is the output matrix, where,

The activation function of the extreme learning machine model is y( ), the output is F, and the expressions of the two are

F=[f ₁ , f ₂ ,...,f _M ], (6)

where s _i =[s _i1 ,s _i2 ,...,s _iu ], a _i =[a _1j ,a _2j ,...,a _uj ] ^T . At the same time F ^T =LP, where,

where L is the hidden layer output matrix and F ^T is the transpose matrix of F. Specifically, when y is infinitely divisible, s and θ will be constant during training. At the same time, the only unknown parameter is P, which can be solved by the least squares method, ie min _P ||LP-F ^T ||.

3) Build a whale optimizer to simulate the hunting strategy of humpback whales.

3.1) Foraging and enveloping stage, in this process, the whale group shares the location information of the prey, and finally surrounds the prey. Assuming that the current best individual of the whale swarm is the target position, the other individuals update their positions and try to approach the best individual position. The whale's position update formula is as follows:

是最佳鲸鱼个体位置，K和Q是系数向量。K和Q的可分别由K＝2k·n-k，J＝2n，k＝2-(2t)/MAX_iterβ,Q＝|J·pos(t)-pos(t)|计算得出，其中n是[0,1]中的随机向量，k在迭代过程中从2减小到0。where pos(t) is the current whale individual position, t is the current iteration number,

is the optimal whale individual position, and K and Q are coefficient vectors. K and Q can be calculated by K=2k·nk, J=2n, k=2-(2t)/MAX _iter β, Q=|J·pos(t)-pos(t)|, where n is a random vector in [0,1], k decreases from 2 to 0 during iteration.

3.2) Random search stage, in which the current single whale position is randomly selected as the best solution. In contrast to the previous process, the whale positions were updated based on randomly selected whales rather than the best individual whales found so far. Adjusting the K value will keep other individuals away from the selected whale, and the update formula is as follows:

pos(t+1)=pos _rand (t)-K·Q, (10)

Q=|J·pos _rand (t)-pos(t)|, (11)

where pos _rand (t) is the position of a random whale. When |K| ≤ 1, the optimal individual is selected as the solution to achieve the local optimization of the whale optimizer. When |K|>1, individuals are randomly selected as solutions to achieve the global optimization of the whale optimizer.

3.3) Foam net attack stage

During this phase, whales have two behaviors: retraction, circle, and hover. The orbital motion of the whale is achieved in step 2.1 by reducing k. The spiral function is used to simulate the behavior of a whale spiral. The choice of these two behaviors is determined by a random number v to achieve simultaneous shrinking encirclement along a spiral path. The positions of individual whales are updated as follows:

是鲸鱼个体与迄今为止获得的最佳个体之间的距离，o是定义螺旋形状的常数，l是范围在-1与1之间的数。in

is the distance between the whale individual and the best individual obtained so far, o is a constant defining the shape of the spiral, and l is a number ranging between -1 and 1.

4) Update the whale optimizer with various hybrid optimization strategies to further enhance the local optimization and global optimization capabilities.

4.1) Use the Levy strategy to initialize the whale population. The population initialization has an important impact on the optimization process of the optimizer. At the same time, a high-quality initial state can speed up the convergence of the optimizer. The Levy strategy is used to initialize the population. Thanks to its random walk mode, the search range of the whale group can be expanded, and the individual whales can be effectively prevented from falling into local extreme values. The mathematical model of the Levy strategy is as follows:

where λ is the random step size, and μ and η follow a normal distribution i.e.

, the values of δ _μ and δ _η can be given by

Calculated, where Γ(·) is the standard gamma function. So the initial position based on the Levy strategy can be calculated by the following formula:

where L _b is the lower bound, U _b is the upper bound, and Levy(v) is a random vector with a step size obeying the Levy distribution.

4.2) Non-linear convergence factor, in the foraging and round-up stages, a non-linear convergence factor is introduced in the whale optimizer to improve the convergence speed. The nonlinear convergence factor is:

Where Max _iter is the maximum number of iterations.

5) Introduce the migration strategy into the whale optimizer to increase the diversity of whale groups and enhance the ability of whales to jump out of local minima. The formula for the latest position of the whale population after migration is as follows:

pos(t+1)=B ^* pos(t)+pos(t) ^* randn(0,σ ² ), (18)

where B (B _max =0.9, B _min =0.4) is the convergence coefficient, and randn(0,σ ² ) conforms to a Gaussian distribution.

6) Initialize the parameters of the hybrid strategy whale optimizer, and input the cloud server load training set into the optimizer for model training, and then input the optimal hyperparameters to the extreme learning machine model. Finally, the cloud server load data test set is input into the extreme learning machine model to predict the cloud server load.

Claims (9)

1. A cloud server load prediction method based on a hybrid optimization strategy extreme learning machine is characterized by comprising the following steps:

1) establishing the extreme learning machine model;

2) and constructing a whale optimizer.

3) Updating the whale optimizer by using a diversity mixing optimization strategy;

4) introducing the migration strategy into a whale optimizer to obtain a whale optimizer with a mixed strategy;

5) obtaining historical load data of a cloud server, and training a mixed strategy whale optimizer to obtain an optimal hyperparameter P;

6) inputting the optimal hyper-parameter into an extreme learning machine model, and establishing a cloud server load prediction model;

7) and acquiring current cloud server load data, inputting the current cloud server load data into a cloud server load prediction model, and predicting the future load of the cloud server.

2. The cloud server load prediction method based on the hybrid optimization strategy extreme learning machine according to claim 1, wherein the extreme learning machine model comprises an input layer, a hidden layer and an output layer; the input layer, the hidden layer and the output layer are connected by neurons, and the number of the neurons in each layer is u, r and e; the weight coefficient between the hidden layer and the input layer is S, and the weight coefficient between the hidden layer and the output layer is P; theta is the set of neuron thresholds for the hidden layer,

the weighting coefficient S, the weighting coefficient P, and the set θ are respectively as follows:

in the formula, s_ruIs a weight coefficient between the r hidden layer neuron and the u input layer neuron; p is a radical of_reIs a weight coefficient between the r hidden layer neuron and the e output layer neuron; theta_rIs the r-th hidden layer neuron threshold.

3. The cloud server load prediction method based on the hybrid optimization strategy extreme learning machine according to claim 1, characterized in that: the cloud server load data is subjected to normalization processing.

4. The method for predicting the load of the cloud server based on the hybrid optimization strategy extreme learning machine according to claim 3, wherein the cloud server load data comprises a training set and a test set; the training set is used for training the mixed strategy whale optimizer, and the testing set is used for testing the mixed strategy whale optimizer;

the training set of the extreme learning machine model includes an input matrix a and an output matrix B, i.e.:

in the formula, M is the number of the cloud server load data training sample sets.

5. The cloud server load prediction method based on the hybrid optimization strategy extreme learning machine as claimed in claim 4, wherein in the training process of the hybrid strategy whale optimizer, a weight coefficient s and a neuron threshold theta are constants, and a weight coefficient P is an unknown parameter; the weight coefficient P is solved by a least square method.

6. The cloud server load prediction method based on the hybrid optimization strategy extreme learning machine according to claim 1, wherein the activation function of the extreme learning machine model is y (-) and the output is F, that is:

F＝[f₁,f₂,...,f_M] (6)

in the formula, the weight coefficient s_i＝[s_i1,s_i2,…,s_iu]；a_i＝[a_1j,a_2j,...,a_uj]^TIs the ith column of the input matrix A; transposed matrix F of output F^T＝LP；f_MIs an element of the output matrix F;

wherein, the hidden layer output matrix L is as follows:

7. the cloud server load prediction method based on the hybrid optimization strategy extreme learning machine as claimed in claim 1, wherein the whale optimizer comprises a foraging embracing phase, a random search phase and a foam net attack phase;

during the foraging embracing phase, the whale position is updated as follows:

wherein pos (t +1) is the predicted location of the individual whale; t is the current iteration number;

is the best individual position of whale; k and Q areA number vector; the coefficient vector K is 2K · n-K; parameter k is 2- (2t)/MAX_iterBeta, beta is a nonlinear convergence factor; coefficient vector Q ═ J · pos (t) -pos (t) |; the parameter J is 2 n; n is [0,1 ]]A random vector of (1);

in the random search phase, the whale's location is updated as follows:

pos(t+1)＝pos_rand(t)-K·Q (10)

Q＝|J·pos_rand(t)-pos(t)| (11)

in the formula, pos_rand(t) is the location of random whales; when the absolute value K is less than or equal to 1, selecting an optimal individual as a solution so as to realize local optimization of the whale optimizer; when K > 1, individuals are randomly selected as solutions to achieve global optimization of the whale optimizer.

During the foam net attack phase, the whale's location is updated as follows:

in the formula (I), the compound is shown in the specification,

is the distance between the individual whale and the best individual obtained so far; o is a constant defining the shape of the helix; l is a number ranging between-1 and 1; v is a random number.

8. The cloud server load prediction method based on the hybrid optimization strategy extreme learning machine as claimed in claim 1, wherein the step of updating the whale optimizer using the diverse hybrid optimization strategy comprises:

1) initializing a whale population by using a Levy strategy; the mathematical model of the Levy strategy is as follows:

wherein λ is a random step size; v is a random number;

wherein the parameter μ and the parameter η follow a normal distribution, namely:

parameter delta_μAnd parameter delta_ηRespectively as follows:

wherein Γ (·) is a standard gamma function;

2) based on the Levy policy, the individual initial position is calculated, namely:

in the formula, L_bIs the lower limit, U_bIs the upper bound, Levy (v) is the random vector of steps that obey the Levy distribution;

3) introducing a nonlinear convergence factor beta in a whale optimizer, namely:

in the formula, Max_iterIs the maximum number of iterations.

9. The cloud server load prediction method based on the hybrid optimization strategy limit learning machine as claimed in claim 1, characterized in that after the migration strategy is introduced into the whale optimizer, the position of the whale colony is updated as follows;

pos(t+1)＝B^*pos(t)+pos(t)*randn(0,σ²) (18)

where B is the convergence coefficient, randn (0, σ)²) The Gaussian distribution is met; b is_max、B_minAre the upper and lower limits of the convergence coefficient B.

Images