CN110222826A - One kind being based on improved EEMD-IndRNN ship method for predicting - Google Patents

Abstract

The invention discloses an improved EEMD-IndRNN-based ship flow prediction method, which uses an ensemble empirical mode decomposition algorithm to decompose nonlinear and non-stationary ship flow data into a series of high and low frequency eigenmode function sequences with stability and a monotone The remainder sequence of the original sequence not only retains the information of the original sequence to the greatest extent, but also makes full use of the internal laws of the sequence to improve the prediction accuracy; then use the Pearson correlation coefficient to calculate the correlation between each component and the original ship flow data, and according to the correlation size will be The components are recombined into three new components: high, medium and low; finally, the above components are processed separately by using an independent cyclic neural network, and a deep learning neural network is constructed by superimposing multiple hidden layers, combined with a large amount of ship traffic data, in data training The time hidden feature information of the ship flow is fully extracted to complete the prediction. The invention not only refines the processing of data components, but also improves the prediction accuracy and has better adaptability.

Description

An Improved EEMD-IndRNN Ship Flow Forecasting Method

technical field

The invention relates to the technical field of time series prediction, in particular to a method for predicting ship flow based on an improved EEMD-IndRNN (ensemble empirical mode division-independent recurrent neural network).

Background technique

With the development of my country's maritime economy and trade, the number of ships is gradually increasing. In order to solve the problems of ship traffic accidents caused by the increase in the density of water routes, channel planning and improve the efficiency of ship traffic, scientific and accurate prediction of ship flow is required. Nowadays, due to the rapid development of science and technology and higher forecast accuracy requirements, a single forecast method for ship flow can no longer meet people's needs, usually an effective combination of multiple forecast algorithms.

The research on ship traffic flow forecasting is mainly divided into: one is to seek a unique internal relationship according to the influencing factors of ship flow data, and form a function map with multiple inputs and outputs, so as to achieve the purpose of forecasting. The second is to use the existing forecasting model in the forecasting of ship traffic flow through application innovation. At present, domestic and foreign research on ship traffic flow forecasting is very rich, mainly including neural network, wavelet transform, support vector machine, long short-term memory network and combined forecasting.

In view of a series of problems in the existing technology, such as low prediction accuracy, low adaptability, and cyclic neural gradient explosion, a ship flow prediction method based on improved ensemble empirical mode decomposition and independent cyclic neural network can be developed, which can solve the problem of For the adaptability of time series forecasting on different time scales, it can further improve the forecasting accuracy of ship flow.

Contents of the invention

The object of the present invention is to provide a kind of ship flow forecasting method based on improved EEMD-IndRNN (ensemble empirical mode and independent cyclic neural network), which belongs to the framework of deep neural network method, and can solve the problem of time series prediction for different time scales Adaptability issues, while improving prediction accuracy.

In order to achieve the above object, the present invention is achieved through the following technical solutions:

A ship flow prediction method based on improved ensemble empirical mode decomposition and independent cyclic neural network, comprising the following steps: S1, preprocessing the original ship flow data; S2, performing stationarity verification on the original ship flow data; S3, the non-stationary original ship flow data that obtains by described step S2 verification carries out set empirical mode decomposition, obtains several eigenmode functions and a remainder; S4, calculates the correlation coefficient of above-mentioned each component and original ship flow data , superimposed and combined into new components according to the degree of correlation. S5. Use the independent cyclic neural network model to predict the new components. Since the new combined component represents the long-term, medium-term and short-term trend of the ship flow, it is necessary to use an independent cyclic neural network prediction model with different parameters. Finally, the component prediction results Superposition is performed to obtain the ship flow prediction result. .

Preferably, in the step S2, further comprising:

The ADF test method is used to check the stationarity of the time series of the original ship flow data. When the time series of the original ship flow data is stable, there is no unit root; otherwise, there is a unit root.

Preferably, further comprising in the step S3:

S31. Add white noise to the original ship flow data X(t) to obtain ship flow data X _n1 (t)=X(t)+n(t) after adding white noise, where n(t) is white noise ;

S32. Find all the maximum and minimum values of the original ship flow data X _n (t), respectively use cubic spline interpolation to fit the upper envelope q1 and the lower envelope q2, and take their average value m(t)=(q1+q2)/2, to obtain a new sequence h1=X _n1 (t)-m(t); when the new sequence h1 has a positive minimum value or a negative maximum value, then repeat the Describe step S32, until finding first intrinsic mode function IMF1, further obtain new data Xn1(t) _-IMF1 ;

S33, according to the new data X _n1 (t)-IMF1 of described step S32, and using new data X _n1 (t)-IMF1 as X _n1 (t) in the step S32 of the next cycle, step S32 is executed in a loop until Decompose the original data X _n (t) into m intrinsic modulus functions and a monotonic remainder r(t), then:

X _n1 (t) = IMF ₁ +IMF ₂ +...+IMF _m +r(t)

Step S34, adding K times of different white noises to the original data X(t), and repeating the above steps S31-S33, correspondingly obtaining m eigenmode functions and a monotonic remainder;

Step S35, performing overall average calculation on each decomposed component, as follows:

In the formula, IMF' _m is the average value of the sum of the m eigenmode functions obtained after adding white noise for K times and decomposing it, i means adding white noise for the ith time; r'(t) means K times The average value of the sum of the monotonous remainder r(t) obtained after adding white noise and decomposing;

Step S36, the decomposition result of the original ship flow data X(t) is:

X(t)=IMF' ₁ +IMF' ₂ +...IMF' _m + r'(t)

Preferably, the step S4 further includes

Step S41, calculating the Pearson correlation coefficients of appeal components IMF' ₁ , IMF' ₂ , ..., IMF' _m , r'(t) and X(t)

Among them, X is each component, Y is the original data, E is the mathematical expectation, and cov is the covariance. When the correlation coefficient is 0, it indicates that there is no relationship between the two variables. When one variable increases (decreases) as the other variable increases (decreases), it is called a positive correlation between the two variables, Pearson correlation The coefficient takes a value between 0 and 1.

Preferably, in the step S42, the plurality of setting intervals include weak correlation intervals, medium correlation intervals and strong correlation intervals, and correspondingly obtain three new components M1, M2 and M3, respectively representing short, short, medium and long-term trends;

Among them, the component M1 is equal to the sum of the average values of all decomposed eigenmode functions located in the weak correlation interval, the component M2 is equal to the sum of the average values of all decomposed eigenmode functions located in the medium correlation interval, and the component M3 It is equal to the sum of the average values of all decomposed eigenmode functions in the strongly correlated interval.

Preferably, in the step S42, the plurality of set intervals include intervals (0, 0.3], (0.3, 0.6], (0.6, 1.0].

Preferably, further comprising in the step S5:

Step S51, using the new components M1, M2, and M3 recombined in the step S4 as the input of the independent cyclic neural network;

Step S52, the hidden layer state expression of the independent recurrent neural network is changed to:

h _t ＝σ(WX _t +u⊙h _t-1 +b)

In the formula, t is the time, X _t is the input at time t, that is, the above-mentioned recombined new components M1, M2, M3; W is the weight between hidden layers, σ is the activation function of neurons; u is the input layer and the weight between the hidden layer; b is the bias value; ⊙ represents the product of matrix elements; h _t-1 represents the hidden layer output at time t-1 (that is, the previous time), that is, each hidden layer neuron at time t Only accept the output at this moment and its own state at time t-1 as input;

Step S53, when it is necessary to construct a multi-layer recurrent neural network, the output of the new hidden layer is:

h′ _t = σ(W′h _t +u′⊙h′ _t-1 +b′)

In the formula, t is the time, h _t is the output of the previous hidden layer at time t; h′ _t-1 represents the output of the new hidden layer at time t-1; h′ _t represents the output of the new hidden layer; W′ is the weight between the new hidden layers, σ is the activation function of neurons; u' is the weight between the previous hidden layer and the current hidden layer; b' is the bias value of the current layer; ⊙ represents the matrix element product;

Step S54, the output of the independent recurrent neural network is:

Y( _t )=Vh't+c

In the formula, V is the weight coefficient between the last hidden layer and the output layer, h′ _t is the output of the last hidden layer, and c is the threshold;

Step S55, output each predicted value component through each independent cyclic neural network, and superimpose the obtained predicted value components of each independent cyclic neural network to obtain a predicted result:

Y＝Y ₁ +Y ₂ +...+Y _m +Y _r

Among them, Y _m and Y _r are the predicted values of different ship flow components through the independent recurrent neural network.

Compared with prior art, the beneficial effect of the present invention is:

(1) In actual life, the ship traffic flow will be affected by factors such as seasons, climate, human activities and form irregular fluctuations, which has the characteristics of non-stationary and nonlinear, which brings great difficulties to the prediction. The invention makes full use of the time information of the time series by utilizing the advantages of the independent cyclic neural network, and uses the ensemble empirical mode algorithm to decompose the nonlinear and non-stationary ship flow data into a series of stationary eigenmode function sequences and a The monotonous residual sequence not only retains the information of the original sequence to the greatest extent, but also makes full use of the internal laws of the sequence to improve the prediction accuracy;

(2) Obtain a series of stable high and low frequency components in the present invention, high frequency component represents the short-term change of ship flow, low frequency component represents the long-term change trend of ship flow, and calculates the Pearson correlation of high and low frequency component and original ship flow data again Coefficient, according to the size of the correlation, the components are combined into new components to obtain new combined components;

(3) In the present invention, since the new combination component represents the long-term, short-term change trend of the ship flow, the independent cyclic neural network prediction model of different parameters is used to process the above-mentioned components separately, and the depth is constructed by superimposing multiple hidden layers Learn the neural network, combine a large amount of ship flow data, fully extract the time hidden feature information of the ship flow in the data training, and complete the prediction;

(4) Due to the problem of gradient explosion and gradient disappearance when the cyclic neural network processes the time series problem, the present invention uses an independent cyclic neural network with suitable optimization parameters to process the decomposed ship traffic flow sequence, which can achieve good prediction results; and traditional Compared with the judgment of factors affecting the flow of ships based on subjective factors, the present invention has better adaptability.

Description of drawings

Fig. 1 is the flow chart of the ship flow prediction method based on the improved ensemble empirical mode decomposition and independent cyclic neural network of the present invention;

Fig. 2 is a schematic diagram of the method for ensemble empirical mode decomposition of the present invention;

Fig. 3 is the set empirical mode decomposition diagram shown in Fig. 2 of the present invention;

Fig. 4 is a schematic diagram of the development of the independent cyclic neural network of the present invention.

Detailed ways

The features, objects and advantages of the present invention will become more apparent by reading the detailed description of a non-limiting embodiment made with reference to FIGS. 1-4 . Referring to Figures 1-4 which illustrate embodiments of the present invention, the present invention will be described in more detail below. However, this invention may be embodied in many different forms and should not be construed as limited to the embodiments set forth herein.

The present invention is mainly used to predict the number of ships passing in a certain port or water area, as shown in the combination of Figures 1 to 4, the ship flow prediction method based on ensemble empirical mode decomposition and independent cyclic neural network of the present invention includes the following steps:

Step S1, ship flow data preprocessing:

Among them, because the ship flow data will be affected by subjective and objective factors, some data are abnormal. Although the number is very small, it has a great impact on the overall prediction model. Therefore, it is necessary to preprocess the ship flow data X(t) (raw data), including deleting zero data and abnormally large data, etc., and the ship flow data {x ₁ , x ₂ ,..., x _n } represent time t The number of ships entering and leaving the port.

Step S2, verifying the stationarity of the ship flow data:

Among them, the ADF (Augmented Dickey-Fuller, added DF unit root) test method is used to check the stationarity of the ship flow data X(t) (original data). If the time series of the data is stationary, there is no unit root, otherwise, there will be a unit root;

In the step S2, the ADF null hypothesis is: if there is no unit root in the time series, then the time series data is stable; if there is a unit root in the time series, that is, non-stationary, for a stable time series data, it is necessary to The null hypothesis is rejected; if the obtained statistic is significantly less than the critical statistical value of 3 confidence levels (1%, 5%, 10%), it means that the null hypothesis is rejected;

Step S3, if after the verification of step S2, it is found that the ship flow data is non-stationary, in order to subsequently improve the prediction accuracy, the non-stationary ship flow data X(t) is subjected to ensemble empirical mode decomposition, and a A series of smooth high and low frequency components.

Among them, the high-frequency component represents the short-term change of ship flow, and the low-frequency component represents the long-term change trend of ship flow. As shown in FIG. 2 , it is a schematic diagram of the principle of performing ensemble empirical mode decomposition on data signals.

The following process is further included in the step S3:

Step S31, adding white noise to the ship flow data X(t), to obtain the ship flow data _Xn1 (t)=X(t)+n(t) after adding white noise, where n(t) is white noise, so that Data of different scales are automatically mapped to an appropriate reference scale to overcome the mode aliasing defect of EMD (Empirical Mode Decomposition, Empirical Mode Decomposition), so as to obtain better decomposition results;

Step S32, find out all the maximum and minimum values of the ship flow data X _n (t), respectively use cubic spline interpolation to fit the upper envelope q1 and the lower envelope q2, and take the average value m( t)=(q1+q2)/2, a new sequence h1=X _n1 (t)-m(t) is obtained. Wherein, if there is a positive minimum value or a negative maximum value in the new sequence h1, repeat this step S32 until the first intrinsic modulus function IMF1 is found, so that new data X _n1 (t)-IMF1 can be obtained;

Step S33, according to the new data X _n1 (t)-IMF1 of step S32, and using new data X _n1 (t)-IMF1 as X _n1 (t) in the step S32 of next cycle, step S32 is executed in a loop until the The original data X _n (t) is decomposed into m intrinsic modulus functions and a monotonous remainder r(t), and finally:

X _n1 (t) = IMF ₁ +IMF ₂ +...+IMF _m +r(t) (1)

Step S34, based on the above principle, the present invention adds the original data X(t) to K times of different white noises, and repeats the above steps S31-Step S33, and correspondingly obtains m eigenvalues decomposed after adding white noise each time Modulo function and a monotonic data.

Step S35, in order to eliminate the added noise, perform overall average calculation on each decomposed component, as follows:

In the formula, IMF _im refers to the m-th eigenmode function obtained by adding white noise for the ith time and decomposing it; IMF′ _m is the sum of the m-th eigenmode function obtained by adding white noise for K times and decomposing i refers to adding white noise for the ith time; r _i (t) refers to the monotonous data r(t) obtained by adding white noise for the ith time and decomposing it; r'(t) refers to adding K times White noise and the average value of the sum of the monotonous data r(t) obtained after decomposition.

Therefore, the decomposition result of the original ship flow data X(t) is:

X(t) = IMF' ₁ + IMF' ₂ + ... IMF' _m + r'(t) (4)

Step S4, calculating the correlation coefficients between the above components and the original ship flow data, superimposing and combining them into new components according to the degree of correlation. In step S3, if a neural network model is established for each component of ship flow data, the amount of calculation is very large, and there is a certain amount of noise in the high-frequency component, and the components can be superimposed and recombined according to the degree of correlation without changing On the basis of prediction accuracy, the amount of calculation is greatly reduced. Therefore, in this step, the Pearson correlation coefficient is added to classify according to the correlation, and the weak correlation is medium correlation.

The following process is further included in the step S4:

Step S41, calculate the Pearson correlation coefficient of the above-mentioned components IMF' ₁ , IMF' ₂ , ..., IMF' _m , r'(t) and the original ship flow data X(t), so that each The contribution of the component to the original data is as follows:

In the formula (5), the parameter X is each component, Y is the original data, E is the mathematical expectation, and cov means the covariance. Among them, when the correlation coefficient is 0, it shows that the two variables X and Y in formula (5) have no relationship; when one variable increases (decreases) with the increase (decrease) of the other variable, it is called two There is a positive correlation between the two variables, and the value of the Pearson correlation coefficient above is between 0 and 1.

Step S42, according to the correlation result calculated in the above formula (5), according to the correlation size of the Pearson correlation coefficient, according to the interval (0,0.3], (0.3,0.6], (0.6,1.0]) recombined into Three new components M1, M2 and M3 are weakly correlated, moderately correlated and strongly correlated. Each new component is the result of adding one or more IMFs. The three components M1, M2 and M3 respectively represent the Short-term, medium-term and long-term trends. For example, if the correlation is in the weak correlation interval, there are IMF' ₁ and IMF' ₂ , then M1=IMF' ₁ +IMF' ₂ , that is, M1 is equal to all the decompositions in the weak correlation interval The sum of the average values of the eigenmode function components; similarly, M2 is also equal to the sum of the average values of all decomposed eigenmode function components located in the medium correlation interval, and M3 is also equal to all the eigenmode function components located in the strong correlation The sum of the mean values of the modulo function components.

Step S5, among the three components M1, M2 and M3 obtained in the above step S4, the high-frequency component M1 represents the short-term change of the ship flow, and the low-frequency component M3 represents the long-term change trend of the ship flow. Therefore, it is necessary to predict the components M1, M2, and M3 using independent cyclic neural network models optimized with different parameters. The flow chart is shown in Figure 1.

Figure 4 shows the model structure of an independent recurrent neural network (IndRNN, Independent Recurrent Neural Networks). The time series t represents the ship flow data of different dates. Each time requires a ship data with a length of time_step (assumed to be in the range of 1-10) as input. Output data with the same length as time_step (assumed range is 2-11), and obtain a model with optimized weights through a large amount of data training, so as to achieve the purpose of prediction.

In the step S5 of the present invention, independent recurrent neural network (IndRNN) prediction further comprises the following process:

Step S51, use the new combined components M1, M2 and M3 obtained in step S4 as the input of the independent cyclic neural network, different components need to use the IndRNN model with different parameters, and the input in the following steps is uniformly expressed as X′ _t instead of the step The combined components M1, M2, M3 in S4, ie each combined component (M1, M2, M3) is _X't in the following method.

Step S52, the hidden layer state expression of the independent recurrent neural network is:

h _t ＝σ(WX′ _t +u⊙h _t-1 +b) (6)

In the formula, t is the time; X′ _t is the input at time t (that is, the above-mentioned components M1, M2, M3 respectively); W is the weight between hidden layers; σ is the activation function of neurons; u is the input layer and the weight between the hidden layer; ⊙ represents the matrix element product; h _t-1 represents the hidden layer output at time t-1 (that is, the previous time), that is, each hidden layer neuron at time t only accepts the output at this moment and The state of itself at time t-1 is taken as input.

In ship traffic forecasting, the weights W and u can extract characteristic information of the data through independent recurrent neural network model training. In the traditional RNN, each neuron at time t receives the state of all neurons at time t-1 as input. That is, the hidden layer state expression of the traditional RNN is:

h _t =σ(WX′ _t +Uh _t-1 +b) (7)

From Formula 6 and Formula 7, it can be seen that in the connection of hidden layer weights, the independent cyclic neural network is simplified, which can effectively solve the problem of gradient disappearance and gradient explosion, so that the neural network can be superimposed with 90 layers to build a deep learning network , to extract more characteristic information from ship flow data, so as to improve the prediction accuracy of ship flow more effectively.

Step S53, when it is necessary to construct a multi-layer recurrent neural network, the output of the new hidden layer is:

h′ _t = σ(W′h _t +u′⊙h′ _t-1 +b′) (8)

In the formula, t is the time; h′ _t-1 represents the output of the new hidden layer at time t-1; h′ _t represents the output of the new hidden layer; h _t is the output of the previous hidden layer at time t; W′ is the weight between the new hidden layers; σ is the activation function of neurons; u' is the weight between the previous hidden layer and the current hidden layer; b' is the bias value of the current layer; ⊙ represents the matrix element product.

Step S54, the output of the independent recurrent neural network is:

Y(t)=Vh′ _t +c (9)

In the formula, V is the weight coefficient between the hidden layer and the output layer, h′ _t is the output of the last hidden layer, and c is the threshold.

Step S55, obtaining the predicted value components through the independent cyclic neural networks with different parameters, and superimposing the obtained predicted value components of the independent cyclic neural networks to obtain the predicted results:

Y＝Y ₁ +Y ₂ +...+Y _m +Y _r (10)

Among them, Y ₁ , Y ₂ , . . . Y _m and Y _r are the predicted values of different ship flow components through the independent recurrent neural network.

To sum up, in the present invention, the ship flow data is preprocessed first, and then the stationarity verification is performed. Decompose the set empirical mode into a series of high and low frequency components, and transform it into a stationary data sequence. Among them, the high-frequency component represents the short-term change of ship flow, and the low-frequency component represents the long-term change trend of ship flow. This kind of processing of ship flow data can greatly improve the prediction accuracy, and can overcome the problem that the ship traffic flow is affected by many complex factors such as man-made and natural, and the non-stationary and nonlinear data bring great difficulties to the prediction. At the same time, there are gradient explosion and gradient disappearance problems when the cyclic neural network deals with time series problems. Using an independent cyclic neural network with appropriate parameters to process the decomposed ship traffic flow sequence can avoid such problems and build a multi-layer neural network. Network, to achieve a very good prediction effect.

Although the content of the present invention has been described in detail through the above preferred embodiments, it should be understood that the above description should not be considered as limiting the present invention. Various modifications and alterations to the present invention will become apparent to those skilled in the art upon reading the above disclosure. Therefore, the protection scope of the present invention should be defined by the appended claims.

Claims (7)

1. a kind of ship method for predicting based on improved set empirical mode decomposition and independent loops neural network, special Sign is comprising the steps of:

S1, original ship data on flows is pre-processed, deletes the value for being zero in original ship data on flows, then delete one Maximum value and minimum value；

S2, stationarity verifying is carried out to the original ship data on flows；

S3, set empirical mode decomposition is carried out by the original ship data on flows of the step S2 non-stationary verified, Obtain several intrinsic mode functions and a remainder；

The Pearson came phase relation of S4, several intrinsic mode functions in calculating step S3 and a remainder and original ship flow data Number, and be multiple new components according to degree of relevancy size stack combinations；

S5, the multiple new component is predicted using the independent loops neural network model of different parameters respectively, is obtained The predicted value component of each independent neural network is simultaneously superimposed, and ship volume forecasting result is obtained.

2. the ship flow as described in claim 1 based on improved set empirical mode decomposition and independent loops neural network Prediction technique, which is characterized in that

In the step S2, further include:

Stationarity verification is carried out using time series of the ADF method of inspection to the original ship data on flows, when described original The time series of ship data on flows is steady, then unit root is not present, conversely, then there is unit root.

3. the ship as claimed in claim 1 or 2 based on improved set empirical mode decomposition and independent loops neural network Method for predicting, which is characterized in that

It is further included in the step S3:

S31, white noise is added in the original ship data on flows X (t), the ship data on flows after obtaining addition white noise X_n1(t)=X (t)+n (t), n (t) are white noise；

S32, the original ship data on flows X is found out_n(t) all maximum and minimum use cubic spline interpolation respectively Coenvelope line q1 and lower envelope line q2 are fitted, its average value m (t)=(q1+q2)/2 is taken, obtains new sequences h 1=X_n1(t)-m (t)；When the new sequences h 1 is there are positive minimum or negative maximum, then repeating said steps S32, until finding first Intrinsic mode functions IMF1 further obtains new data X_n1(t)-IMF1；

S33, according to the new data X of the step S32_n1(t)-IMF1, and by new data X_n1(t)-IMF1 is as subsequent cycle X in step S32_n1(t), circulation executes step S32, until by initial data X_n(t) m intrinsic mode functions and one are decomposed into Dull remainder r (t), then:

X_n1(t)=IMF₁+IMF₂+...+IMF_m+r(t)

Step S34, K different white noise is added in the initial data X (t), and repeats above step S31- step S33, Accordingly obtain every time be added white noise after decomposed after m intrinsic mode functions and a dull remainder；

Step S35, ensemble average calculating is carried out to each component of decomposition, as follows:

In formula, IMF_imRefer to m-th of intrinsic mode functions that i-th is added white noise and is decomposed；IMF′_mBe K times plus Enter white noise and the average value of the sum of m-th of intrinsic mode functions obtaining after being decomposed, i refers to that white noise is added in i-th；r_i (t) refer to the dull data r (t) that i-th is added white noise and is decomposed；R ' (t) refers to K addition white noise simultaneously The average value of the sum of the dull data r (t) obtained after being decomposed；

Step S36, the decomposition result of the described original ship flow data X (t) are as follows:

X (t)=IMF '₁+IMF′₂+...IMF′_m+r′(t)。

4. the ship volume forecasting side as claimed in claim 3 based on set empirical mode decomposition and independent loops neural network Method, which is characterized in that

It is further included in the step S4:

Step S41, the component IMF ' after above-mentioned decomposition is calculated₁, IMF '₂..., IMF '_m, r ' (t) and original ship data on flows X (t) Pearson correlation coefficient, as follows:

In formula (5), X indicates that each component, Y indicate original ship data on flows, and E is mathematic expectaion, and cov indicates covariance；Wherein, When related coefficient is 0, show that X and two variables of Y are not related；Increase or reduce with another variable when a variable and It increases or reduces, then to be positively correlated between two variables, Pearson correlation coefficient value is between 0 to 1；

Step S42, according to the calculated correlation results of step S41, it is multiple for reconfiguring respectively according to multiple set intervals New component, each component are the result that one or more intrinsic mode functions are added.

5. the ship volume forecasting side as claimed in claim 4 based on set empirical mode decomposition and independent loops neural network Method, which is characterized in that

In the step S42, the multiple set interval includes weak related interval, medium related interval and strong correlation section, right It obtains three new component M1, M2 and M3 with answering, respectively indicates the variation tendency of the short, medium and long phase of ship flow；

Wherein, component M1 is equal to intrinsic mode functions average value sum after all decomposition positioned in weak related interval, component M2 is equal to intrinsic mode functions average value sum after all decomposition positioned in medium related interval, and component M3 is equal to all positions Intrinsic mode functions average value sum after decomposition in strong correlation section.

6. the ship volume forecasting side as claimed in claim 5 based on set empirical mode decomposition and independent loops neural network Method, which is characterized in that

In the step S42, the multiple set interval include section (0,0.3], (0.3,0.6], (0.6,1.0].

7. as the ship flow described in claim 5 or 6 based on set empirical mode decomposition and independent loops neural network is pre- Survey method, which is characterized in that

It is further included in the step S5:

Step S51, using three reconfigured in the step S4 new component M1, M2 and M3 as independent loops neural network Input X '_t；

Step S52, the hidden layer state expression formula of the described independent loops neural network are as follows:

In formula, t is the moment；X′_tIt is the input of t moment, i.e., respectively above-mentioned component M1, M2, M3；W is the power between hidden layer Weight；σ is the activation primitive of neuron；U is the weight between input layer and hidden layer；B is bias；Representing matrix element Product；h_t-1The hidden layer output for indicating t-1 moment (i.e. previous moment), i.e., only receive this in each hidden layer neuron of t moment The state at the output at quarter and t-1 moment itself is as input；

Step S53, when constructing multilayer circulation neural network, new hidden layer output are as follows:

In formula, t is the moment；h_tIt is the preceding layer hidden layer output of t moment；h′_t-1Indicate that new hidden layer is defeated at the t-1 moment Out；h′_tIndicate new hidden layer output；W ' is the weight between new hidden layer；σ is the activation primitive of neuron；Before u ' is Weight between one hidden layer and current hidden layer；B ' is the bias of current layer；Representing matrix element product；

Step S54, the output of the described independent loops neural network are as follows:

Y (t)=Vh '_t+c

In formula, weight coefficient of the V between the last one hidden layer and output layer, h '_tFor the output of the last one hidden layer, c is threshold Value；

Step S55, each predicted value component, and each independent loops nerve net that will be obtained are exported by each independent loops neural network Each predicted value component of network is overlapped to obtain prediction result:

Y=Y₁+Y₂+...+Y_m+Y_r

Wherein, Y₁、Y₂、……Y_mAnd Y_rTo pass through the predicted value of the different ship flow component of independent loops neural network.