A method for predicting the quantity of eel fry resources

Abstract

The application discloses a method for predicting eel fry resource quantity, comprising the following steps: obtaining initial factors, screening the initial factors to obtain influence factors, obtaining historical data according to the influence factors, and constructing a generalized linear model and a nonlinear model; optimizing and verifying the generalized linear model and the nonlinear model through the historical data, obtaining current data according to the influence factors, predicting the current data through the verified generalized linear model and the nonlinear model, and combining the prediction results by weighting to generate an eel fry resource quantity prediction result. Through the technical scheme, the eel fry resource quantity in an estuary can be effectively and universally predicted.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of seed resource prediction, and particularly relates to a method for predicting the resource quantity of eel fry. BACKGROUND

[0002] Eel, commonly known as eel and white eel, is a migratory fish with important economic value in Asia, and is known as "ginseng in water". It is deeply loved by consumers. However, in recent years, due to various human activities, the natural population of eel has decreased sharply, and the fry has decreased sharply. At the same time, due to the fact that artificial breeding has not been successful so far, the fry needed for breeding depends entirely on natural capture. In recent years, due to overfishing by fishermen in the main eel producing area, the number of fry has decreased sharply, the price has repeatedly reached a new high, the breeding cost has greatly increased, and the supply of fry is unstable, which has become the primary factor limiting the further development of eel breeding. Therefore, it is necessary to estimate the resource quantity of eel fry to provide relevant data support for the rational layout of eel breeding industry. SUMMARY

[0003] In order to solve the problem that the resource quantity of eel fry cannot be effectively estimated in the prior art, the present application provides a method for predicting the resource quantity of eel fry, which effectively and universally predicts the resource quantity of eel fry in estuary.

[0004] In order to achieve the above technical purpose, the present application provides the following technical scheme: a method for predicting the resource quantity of eel fry, comprising:

[0005] Obtaining initial factors and screening the initial factors to obtain influencing factors, obtaining historical data according to the influencing factors, and constructing a generalized linear model and a nonlinear model;

[0006] Optimizing and verifying the generalized linear model and the nonlinear model through the historical data;

[0007] Obtaining current data according to the influencing factors, predicting the current data through the verified generalized linear model and nonlinear model, and combining the prediction results by weighting to generate the prediction result of the resource quantity of eel fry.

[0008] Optionally, the initial factors include light, environmental temperature, tidal range, month distance and water temperature.

[0009] Optionally, the initial factors are screened by a grey correlation analysis method.

[0010] Optionally, the initial factor screening process comprises: generating an initial sequence and a resource amount sequence according to the initial factor and the eel fry resource amount; performing interpolation calculation on corresponding position elements of the initial sequence and the resource amount sequence, and generating two-level maximum difference and two-level minimum difference according to the difference calculation result; calculating the correlation coefficient of the initial sequence and the resource amount sequence according to the two-level maximum difference and the two-level minimum difference; calculating the correlation degree according to the correlation coefficient; and screening the initial factor according to the correlation degree.

[0011] Optionally, the influence factors include tidal range, month distance and water temperature.

[0012] Optionally, the generalized linear model is: CPUE = β0 + β1 tidal range + β2 month distance + β3 water temperature, wherein β0 is the intercept; β1, β2 and β3 represent the regression coefficients, and represent the influence degree of the tidal range, the month distance and the water temperature on the eel fry CPUE respectively.

[0013] Optionally, the nonlinear model adopts a deep learning model, and the deep learning model is a DNN model.

[0014] Optionally, the deep learning model comprises an input layer, two continuous hidden layers and an output layer, and the neurons of each layer are 3, 5, 5 and 1 in turn.

[0015] Optionally, the process of verifying the generalized linear model comprises: generating verification data by randomly selecting historical data, performing difference and variance calculation on the result of the generalized linear model and the eel fry resource amount corresponding to the verification data, and verifying the generalized linear model according to the difference and variance calculation result.

[0016] Optionally, the process of verifying the nonlinear model comprises: generating matching data by randomly selecting historical data, performing difference and variance calculation on the result of the generalized linear model and the result of the nonlinear model, and verifying the nonlinear model according to the difference and variance calculation.

[0017] The present application has the following technical effects:

[0018] This invention discloses a method for predicting eel fry resources. It uses grey relational analysis to screen corresponding influencing factors. After screening, based on the influencing factors—estuary tides, monthly intervals, and water temperature data—it analyzes the abundance data of eel fry from previous years. A linear and nonlinear correspondence is established between daily abundance data and tides and water temperature. By jointly calculating the linear and nonlinear results, the eel fry resource status is estimated. Compared with the traditional fixed-position gillnet fishing method, this method significantly reduces costs; combining fry data with environmental data makes the analysis results more accurate and reliable. This invention provides a universally applicable method for predicting estuarine eel fry resources. The estimation of eel fry resources is of great significance for the rational layout of eel aquaculture. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of the method flow provided in an embodiment of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] like Figure 1 This invention discloses a method for predicting eel fry resources. It uses grey relational analysis to screen corresponding influencing factors. After screening, based on the influencing factors—estuary tides, monthly intervals, and water temperature data—it analyzes the abundance data of eel fry from previous years. A linear and nonlinear correspondence is established between daily abundance data and tides and water temperature. By jointly calculating the linear and nonlinear results, the resource status of eel fry is estimated. Compared with the traditional fixed-net fishing method, this method significantly reduces costs; combining fry data with environmental data makes the analysis results more accurate and reliable. This invention is a universally applicable method for predicting estuarine eel fry resources.

[0023] Regarding the above technical solutions, the present invention is described as follows:

[0024] Using relevant statistical software, the parameters that have a correlation influence on the current eel fry resource amount are determined as the relevant parameters of the eel fry resource amount prediction through grey correlation analysis. The grey correlation analysis is to transform the eel fry resource amount and the relevant parameters into geometric shapes through a reasonable method and determine the correlation degree between the relevant influence parameters and the eel fry resource amount by comparison. The higher the correlation degree, the greater the influence on the eel fry resource amount. At the same time, when some data with low correlation degree are used as the relevant parameters of the eel fry resource amount, the accuracy of the eel fry resource amount prediction will be affected.

[0025] When the correlation degree is greater than 0.5, it is considered that there is a strong correlation between the relevant parameters and the eel fry resource amount prediction, and the relevant parameters are reserved as the influence parameters to participate in the prediction of the eel fry resource amount.

[0026] The specific content includes: selecting parameters that may affect the eel fry resource amount, including initial factors such as light, environmental temperature, tidal range, month distance, and water temperature; and statistics of eel fry resources in different months or days.

[0027] Correlation degree analysis between initial factors and eel fry resource: first, sort the above different initial factors and eel fry resource amount in time sequence, and fill in the missing values by filling down to generate the corresponding initial factor sequence and eel fry resource amount sequence;

[0028] Calculate the difference value of the corresponding position elements in each initial factor sequence and eel fry resource amount sequence; according to the calculated difference value, determine the maximum difference value and the minimum difference value of each initial factor sequence and eel fry resource amount sequence, and then determine the two-level maximum difference value and the two-level minimum difference value from the above maximum difference value and minimum difference value;

[0029] According to the two-level maximum difference value and the two-level minimum difference value, calculate the correlation coefficient of each initial factor sequence and eel fry resource amount sequence; wherein: the correlation coefficient y=(two-level minimum difference value+w·two-level maximum difference value) / (|r1-rs|+w·two-level maximum difference value), wherein r1 is the element of the eel fry resource amount sequence, rs is the corresponding element in the initial factor sequence, and w is the resolution coefficient, generally taking 0.5.

[0030] Calculate the correlation coefficient of each element in each initial factor sequence and eel fry resource amount sequence; generate the correlation degree of the initial factor sequence and the eel fry resource amount sequence according to the correlation coefficient; wherein the correlation degree is the average value of the correlation coefficient of each element in the initial factor sequence and the eel fry resource amount sequence, the higher the correlation degree, the stronger the correlation, and when the correlation degree is greater than 0.5, the initial factor sequence is reserved as the key influence factor.

[0031] By the above method, the tidal range, the month distance and the water temperature are screened out as the key influencing factors corresponding to the eel fry resource quantity.

[0032] The key influencing factors are used to predict the eel fry resource quantity, the eel fry resource quantity is distributed in the wild, and generally needs to be counted by artificial capture, and the movement and living habits of eels cannot be effectively recognized and counted by related image acquisition or other data acquisition methods.

[0033] A related prediction model is constructed by using other influencing factors. In the present application, a generalized linear model is used to map the relationship between the influencing factors and the eel fry resource quantity, and at the same time, the present application attempts to use a deep learning model into the prediction of the eel fry resource quantity, and uses the deep learning model as reference data to correct the linear prediction result.

[0034] For the generalized linear model, the above influencing factors are used, and the data of the eel fry capture quantity and the influencing factors in the past years are used as the basis to adjust the generalized linear model, wherein the initially constructed generalized linear model is eel fry CPUE = β0 + β1·tide range + β2·month distance + β3·water temperature, and the generalized linear model is fitted based on the historical data of the past years, to obtain the regression coefficients β1, β2, β3 and the intercept β0 corresponding to the influence degree of the tide range, the month distance and the water temperature on the eel fry CPUE.

[0035] At the same time, a deep learning model is constructed to reflect the non-linear relationship between the influencing factors and the eel fry, and the deep learning neural network is a mathematical model or a calculation model that simulates the structure and function of biological deep learning neural network, which can generally fit the non-linear relationship. However, due to the limitations of research,

[0036] The deep learning model can use a simple structure DNN model, use PyTorch software, select a suitable deep learning network from a deep learning network library, wherein the deep learning model DNN uses an input layer-hidden layer-hidden layer-output layer structure to construct an initialized network structure, the number of neurons in the input layer is the same as the number of influencing factors, the output layer outputs the numerical value of the eel fry resource quantity, the neuron is 1, the number of neurons in the hidden layer is 5, through the built neural network structure, the above network is trained through the data of the influencing factors and the eel fry resource quantity in the past years, and the network parameters are optimized by the loss function, the loss function is selected according to the actual requirements, the trained DNN model is input into the corresponding numerical value of the influencing factors, and the numerical value of the eel fry resource quantity is generated.

[0037] The deep learning model built in the application can fit certain nonlinear relationships, and the deep learning network model DNN is evaluated by the mean square error, when the mean square error is less than a certain threshold, it is considered to meet a certain accuracy, and the trained deep learning model can be used.

[0038] In the process of quantity statistics, relevant researchers usually use the fitted linear model to statistically analyze and predict the data, but there is no use of deep learning network model. For safety, the application uses linear model as the main model and deep learning network as the auxiliary model for prediction. First, the linear model is verified and adjusted. Some data are randomly extracted from the historical data, including the data corresponding to the influencing factors and the eel fry data, and are substituted into the linear model for calculation. The difference between the eel fry quantity output by the linear model and the actual eel fry quantity is calculated, and the mean and variance of the difference are calculated. When the mean and variance are less than a certain threshold, the linear model can be used. The relevant threshold is set according to the artificial experience of relevant personnel, otherwise, the data with the largest difference in the randomly selected data is removed and the linear model is fitted again with all the historical data to complete the adjustment of the linear model.

[0039] After the linear model is fitted, although the linear model is the commonly used model for relevant researchers, due to the limitations of the linear model, a nonlinear model is needed for certain correction. However, if there is a large deviation between the linear result and the nonlinear result, the prediction error will not be accurate. Therefore, in order to combine the linear model and the nonlinear model, some historical data are randomly selected, the difference between the results of the linear model and the deep learning network model is calculated, and the mean and variance of the distance of the deep learning network model are calculated. The mean and variance are judged. The mean means that there is a large gap between the deep learning network model and the linear model in terms of results, and the variance means that there may be a large deviation in some data. If the mean or variance is large and cannot be less than a certain threshold, the data with large error is removed, and the linear model and the nonlinear model are fitted again until a certain threshold requirement is met. The relevant threshold is obtained according to the artificial experience. When the linear model and the nonlinear model meet the above requirements, the data corresponding to the influencing factors are predicted by the linear model and the nonlinear model, and the weighted calculation of the results x of the linear model and the results f of the nonlinear model is performed, and the final result M=(1-α)x+αf, α is the weight value, which is set to 0.1, and M is the eel fry resource quantity prediction result, which is used to predict the eel fry resource quantity.

[0040] Example one

[0041] 1. Using grey correlation analysis to determine the relevant factors affecting the resource quantity of eel fry, this example selects tidal data, monthly distance and water temperature data as relevant variables.

[0042] 2. Using generalized linear model (GLM) to fit the relationship between relevant variables and eel fry:

[0043] (1) Based on the eel fry catch data of at least five consecutive years (n, n+1, n+2, n+3, n+4) in previous years, using generalized linear model (GLM), taking daily single-boat catch of eel fry CPUE as independent variable, taking daily tidal data, monthly distance data and water temperature data as dependent variable, establishing the relationship equation between eel fry abundance and tide:

[0044] Eel fry CPUE = β0 + β1·tidal range + β2·monthly distance + β3·water temperature ………… (1)

[0045] In the formula: β0 is the intercept; β1, β2, β3 represent the regression coefficients, which represent the influence of tidal range, monthly distance and water temperature on eel fry CPUE, respectively.

[0046] (2) Using statistical software to fit equation (1) and calculate the values of β0, β1, β2 and β3.

[0047] (3) When the tidal range, monthly distance and water temperature data of the (n+5) year are known, the data and the values of β0, β1, β2, β3 can be substituted into equation 1 in claim (1), and then the eel fry CPUE data of the (n+5) year linear model can be calculated.

[0048] Example:

[0049] Suppose the eel fry CPUE values, tidal range values, relative monthly distance values and water temperature data of the first to fifth years are as follows in Table 1:

[0050] Table 1

[0051]

[0052]

[0053]

[0054] Using generalized linear model (GLM), taking daily single-boat catch of eel fry CPUE as independent variable, taking daily tidal data and water temperature data as dependent variable, establishing the relationship equation between eel fry abundance and tide: eel fry CPUE = β0 + β1·tidal range + β2·monthly distance + β3·water temperature

[0055] Using statistical software to fit the equation, β0=-1073.836, β1=19.540, β2=19.307, β3=-2.664. It is concluded that: CPUE of Amur eel fry =-1073.836 + 19.540 tidal range + 19.307 month distance + 2.664 water temperature.

[0056] When the tidal range, month distance and water temperature data of any day in the 6th year are known, the CPUE data of the day of the linear model of Amur eel fry can be calculated.

[0057] 3、Using the above data to train the constructed deep learning model, and using the trained deep learning model for prediction, when the tidal range, month distance and water temperature data of any day in the 6th year are known, input the numerical values of the tidal range, month distance and water temperature data of any day in the 6th year into the trained deep learning model for prediction, and the CPUE data of the day of the nonlinear model of Amur eel fry can be obtained.

[0058] 4、Using the linear model and the nonlinear model for weighted calculation, finally the final Amur eel fry resource data of any day in the 6th year can be obtained.

[0059] 5、Before using the linear model, the above method is used for data verification and nonlinear model matching verification, and after the verification meets the relevant requirements, the linear model and the nonlinear model are used for prediction of the Amur eel fry resource data.

[0060] The above is only the preferred specific embodiment of the present application, but the protection scope of the present application is not limited to this, any person skilled in the art can easily think of changes or replacements within the technical range disclosed in the present application, which should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for predicting eel fry resources, characterized in that, include: Obtain initial factors and filter them to obtain influencing factors. Based on the influencing factors, obtain historical data and construct generalized linear and nonlinear models. Optimize and validate generalized linear and nonlinear models using historical data; Based on the influencing factors, current data is obtained, and the current data is predicted using a validated generalized linear model and a nonlinear model. The prediction results are then weighted and combined to generate a prediction result for the eel seedling resource quantity. The initial factor screening process includes: generating an initial sequence and a resource quantity sequence based on the initial factors and the eel seedling resource quantity; performing interpolation calculations on the corresponding elements of the initial sequence and the resource quantity sequence, and generating two levels of maximum and minimum differences based on the difference calculation results; calculating the correlation coefficient between the initial sequence and the resource quantity sequence based on the two levels of maximum and minimum differences; calculating the degree of association based on the correlation coefficient; and screening the initial factors based on the degree of association. The influencing factors include tidal range, lunar distance, and water temperature; The generalized linear model is: CPUE = β0 + β1 tidal range + β2 lunar distance + β3 water temperature, where β0 is the intercept; β1, β2, and β3 represent regression coefficients, which respectively indicate the degree of influence of tidal range, lunar distance, and water temperature on the CPUE of eel fry; The nonlinear model employs a deep learning model, specifically a DNN model. The deep learning model includes an input layer, two consecutive hidden layers, and an output layer, wherein the number of neurons in each layer is 3, 5, 5, and 1, respectively.

2. The method according to claim 1, characterized in that: The initial factors include light, ambient temperature, tidal range, lunar distance, and water temperature.

3. The method according to claim 1, characterized in that: The initial factors were screened using grey relational analysis.

4. The method according to claim 1, characterized in that: The process of validating the generalized linear model includes: generating validation data by randomly selecting historical data; calculating the difference and variance between the result of the generalized linear model and the eel seedling resource quantity corresponding to the validation data using the validation data; and validating the generalized linear model based on the difference and variance calculation results.

5. The method according to claim 1, characterized in that: The process of validating a nonlinear model includes: generating matching data by randomly selecting historical data; calculating the difference and variance between the results of the generalized linear model and the nonlinear model using the matching data; and validating the nonlinear model based on the difference and variance calculations.

Images