Method for calculating migratory parent fish quantity by using fish early-stage resource quantity

Abstract

The invention provides a method for calculating migratory parent fish quantity by using fish early-stage resource quantity, which comprises the following steps: collecting early-stage resource quantity data, migratory parent fish quantity data and various parameters, and processing the collected data to obtain a comprehensive data set; constructing a basic model for representing the relationship among the number of roes, the number of larva fishes and the number of migratory parent fishes in combination with various parameters; designing a model architecture of the basic model by using a long-short-term memory network to obtain an initial prediction model, and quantifying the initial prediction model by using a Bayesian method to obtain a calculation prediction model; in combination with an MCMC method, performing parameter estimation and model training on the reckoning prediction model to realize optimization of the reckoning prediction model; and calculating the number of migratory parent fishes by using the optimized calculation and prediction model. According to the method, multiple data sources can be combined, the ecological environment complexity, the time sequence dynamics and the prediction uncertainty are fully considered, and accurate calculation of the migratory parent fish quantity is achieved.

Description

Technical Field

[0001] The invention relates to the technical field of estimating the number of migratory broodstock, and in particular to a method for estimating the number of migratory broodstock by using the early resource quantity of fish. Background Art

[0002] Migratory broodstock refer to fish that migrate upstream to spawn during the breeding season. They leave their original habitat and head to specific spawning grounds to find a suitable breeding environment to ensure the survival and reproduction of their offspring. Accurately estimating the number of migratory broodstock can help formulate sustainable fishing quotas and prevent overfishing. It can also help protect key habitats and breeding areas, and provide basic data for studying fish population dynamics, ecosystem health, and environmental changes.

[0003] Existing methods for estimating the number of migratory broodstock include but are not limited to: 1) Direct observation and mark-recapture: The number of fish is estimated by setting observation points on the migration path or using mark-recapture technology. This method relies on limited samples and cannot fully reflect the number of fish. 2) Acoustic monitoring: Use sonar technology to monitor the movement and number of fish. However, acoustic monitoring is limited by the coverage of the equipment and the interference of environmental noise. 3) Statistical models: Use historical fishing data and statistical models for estimation. However, statistical models usually rely on simplified assumptions and cannot capture complex ecological dynamics. As a result, the current estimation of the number of migratory broodstock is not accurate. Summary of the invention

[0004] In order to overcome the shortcomings of the prior art, the purpose of the present invention is to provide a method for estimating the amount of migratory broodstock using the early fish resource quantity, which can combine multiple data sources, fully consider the complexity of the ecological environment, the dynamics of time series and the uncertainty of prediction, and realize the accurate estimation of the amount of migratory broodstock.

[0005] To achieve the above object, the present invention provides the following scheme: a method for estimating the amount of migratory broodstock using the early fish resource amount, comprising the following steps:

[0006] Collect early resource quantity data, migratory broodstock quantity data and various parameters, and process the collected data to obtain a comprehensive data set;

[0007] Combining the above parameters, constructing a basic model for representing the relationship between the number of fish eggs, the number of fry and the number of migratory broodstock;

[0008] Using a long short-term memory network to design a model architecture of the basic model, obtaining an initial prediction model, and then using a Bayesian method to quantify the initial prediction model to obtain an inferred prediction model;

[0009] In combination with the MCMC method, parameter estimation and model training are performed on the inference prediction model to achieve optimization of the inference prediction model;

[0010] The optimized estimation and prediction model is used to estimate the number of migratory broodstock.

[0011] Optionally, the early resource quantity data include the number of fish eggs and the number of fry, and the various parameters include the proportion of male fish, the average relative fecundity of female fish in the corresponding river, the average weight of female fish in the corresponding river, the fertilization rate of the corresponding river, the hatching rate of fertilized eggs in the corresponding river and the survival rate of broodstock in the corresponding river.

[0012] Optionally, in combination with the above parameters, a basic model for representing the relationship between the number of fish eggs, the number of fry and the number of migratory broodstock is constructed, including:

[0013] The early resource quantity data and the various parameters are set as independent variables, the migratory parent fish quantity data is set as dependent variables, and the independent variables are standardized so that the mean value of the independent variables is 0 and the standard deviation is 1;

[0014] Performing feature selection on the comprehensive data set using a feature selection algorithm or an importance scoring method to identify the most influential variables;

[0015] A linear regression model is constructed by combining the independent variable, the dependent variable, the most influential variable and the interaction terms among the various parameters, and a regression analysis method is introduced into the linear regression model to handle the nonlinear effects of the various parameters;

[0016] The linear regression model is trained and evaluated by combining the k-fold cross validation method and the evaluation index, and the linear regression model is optimized according to the evaluation result to obtain a basic model.

[0017] Optionally, in combination with the various parameters, a basic model is constructed to represent the relationship between the number of fish eggs, the number of fry and the number of migratory parents, which also includes: in the basic model, the various parameters are used as state variables, and a state space sub-model is constructed in combination with the state variables, the dependent variables and the interaction terms between the various parameters, and the state space sub-model is used to dynamically simulate the impact of environmental changes on fish populations.

[0018] Optionally, a model architecture of the basic model is designed using a long short-term memory network to obtain an initial prediction model, including:

[0019] Integrate the early resource data and the various parameters into a multi-dimensional time series in chronological order to obtain input data, and standardize or normalize the input data to complete the design of the input layer;

[0020] Based on the multi-layer LSTM structure, a gating mechanism is introduced to process the time dependency of the input data, and the nonlinear dynamic characteristics of the multi-dimensional time series are learned to complete the design of the LSTM layer;

[0021] Map the output of the LSTM layer to the target space, generate prediction results, and complete the design of the fully connected layer;

[0022] Use the linear activation function to output the prediction results and complete the design of the output layer;

[0023] The input layer, the LSTM layer, the fully connected layer and the output layer are integrated to obtain the model architecture of the basic model.

[0024] Optionally, the initial prediction model is quantified using a Bayesian method to obtain an inferred prediction model, including:

[0025] Calculating the log-likelihood of the migratory broodstock quantity data using a likelihood function to obtain model parameter estimates, maximizing the model parameter estimates, and obtaining a parameter probability distribution of the initial prediction model;

[0026] The regularization term of the Bayesian method is used to constrain the prior distribution of the initial prediction model parameters, and then the initial prediction model is trained using the comprehensive data set, and the model parameters are adjusted to maximize the likelihood function, thereby completing the quantization of the initial prediction model and obtaining the inferred prediction model.

[0027] Optionally, in combination with the MCMC method, parameter estimation and model training are performed on the inference prediction model to optimize the inference prediction model, including:

[0028] Based on the Bayesian framework, a normal distribution with a mean of 0 is selected as the prior distribution for the model parameters;

[0029] Based on the prior distribution, the MCMC method is used to perform multiple iterations of sampling on the model parameters to generate a parameter sample set close to the posterior distribution of the parameters, thereby completing the parameter estimation of the inference prediction model;

[0030] In each iterative sampling process, the joint probability of the likelihood function and the prior distribution is calculated by updating the parameter sample, and the sampling process of the MCMC method is accelerated by using variational inference to complete the model training of the inference prediction model;

[0031] The posterior distribution of the inferred prediction model is analyzed to obtain a confidence space for estimating the parameter error range, and the inferred prediction model is verified and optimized by combining the confidence space, cross-validation method and evaluation index.

[0032] The present invention discloses the following technical effects by providing a method for estimating the amount of migratory broodstock using the early fish resource amount:

[0033] 1. The present invention takes into account the impact of the environment on the growth and survival of fish, collects data from different sources and types, and can more accurately use early resource data to estimate the amount of migratory broodstock, providing a scientific basis for fishery management and ecological protection.

[0034] 2. The present invention adds interaction terms between various parameters to the basic model, which can capture complex ecological relationships, dynamically simulate the impact of environmental changes on fish populations, and effectively establish a relationship model between the number of fish eggs and fry and the amount of migratory broodstock;

[0035] 3. The present invention combines LSTM with the Bayesian method to construct an inference and prediction model, effectively utilizing LSTM to capture the long-term dependencies in time series data and the Bayesian method to quantify prediction uncertainty, which can more accurately estimate the amount of migratory broodstock and provide prediction information with greater decision-making value.

[0036] 4. The present invention can quantify and dynamically adjust the uncertainty of prediction through parameter estimation after model construction, and then combine model training to achieve model evaluation optimization, which greatly improves the model performance and makes the inference results more reliable.

[0037] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0039] Figure 1 A schematic diagram of a method flow chart provided by an embodiment of the present invention;

[0040] Figure 2 A schematic diagram of a model building process provided by an embodiment of the present invention;

[0041] Figure 3 A schematic diagram of the model training optimization process provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0042] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0043] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0044] like Figure 1 As shown, the present invention provides a method for estimating the amount of migratory broodstock using the early fish resource amount, comprising the following steps:

[0045] 1. Collect early resource quantity data, migratory broodstock quantity data and various parameters, and process the collected data to obtain a comprehensive data set.

[0046] 1.1 Early resource data

[0047] Number of fish eggs and fry: During the parent fish migration breeding season, monitor and record the corresponding data according to the early fish resource monitoring process, record the sampling time, location, depth, water temperature, air temperature, transparency and number of fish eggs, and use the early resource data processing method to calculate the total number of resources in the early breeding season. Remove outliers and erroneous data to ensure the accuracy of the data, and store the data in the database for subsequent analysis.

[0048] 1.2 Data on the number of migratory broodstock

[0049] Data on the number of migratory broodstock are obtained through fishery records, acoustic monitoring and mark-recapture experiments. The time, location, quantity and individual biological characteristics (such as length and weight) of fishing are recorded, duplicate records and erroneous data are removed, and statistical analysis is used to verify the consistency and rationality of the data. Finally, the data is stored in a centralized database to facilitate correlation analysis with earlier resource data.

[0050] 1.3 Various parameters

[0051] Male fish proportion: In daily sampling, record the number of female and male fish caught and calculate the proportion of male fish. Remove outliers and missing values, and use interpolation methods to fill in missing data

[0052] Average relative fecundity of female fish in the corresponding river: Statistical data on reproductive biology of female fish in the corresponding river are collected, and the average relative fecundity of female fish in the river is calculated. Outliers and missing values ​​are processed to ensure data integrity.

[0053] Average weight of female fish in the corresponding river: Conduct routine biological data statistics on female fish in the estimated river, count the weight of female fish, and calculate the average weight. Record the time and location of the statistics, remove outliers, ensure the accuracy of the data, and then store the data in the ecological database to facilitate correlation analysis with other parameters.

[0054] Corresponding river fertilization rate: Underwater tracking and filming of breeding broodstock in the estimated river, observing the ovulation and fertilization of female fish, and calculating the fertilization rate. Record the time and location, remove outliers, and ensure the accuracy of the data.

[0055] Hatch rate of fertilized eggs in the corresponding river: track and observe the eggs that have been successfully fertilized, record their development, and calculate the final hatch rate based on the total number of fertilized eggs. Record the time and location, remove outliers, and ensure the accuracy of the data.

[0056] Survival rate of broodstock in corresponding rivers: record and observe the broodstock population in corresponding rivers and their survival rate during spawning. Record the time and location, etc., remove outliers, and ensure the accuracy of the data.

[0057] 2. If Figure 2 As shown, combining the above parameters, a basic model is constructed to represent the relationship between the number of fish eggs, the number of fry and the number of migratory broodstock. It includes:

[0058] Dynamic models suitable for time series data (such as autoregressive integrated moving average models or state-space models to capture the dynamic changes of fish populations) and machine learning models (such as random forests, support vector machines or neural networks to handle complex nonlinear relationships) can be selected as the framework of the basic model.

[0059] The early resource quantity data and the various parameters are set as independent variables, the migratory broodstock quantity data is set as dependent variables, and the independent variables are standardized so that the mean value of the independent variables is 0 and the standard deviation is 1, thereby eliminating the influence between different dimensions and improving the convergence speed and prediction accuracy of the model.

[0060] Using feature selection algorithms (such as recursive feature elimination, LASSO regression) or importance scoring methods (such as feature importance of random forests) to select features from the comprehensive data set, identify the variables with the greatest impact on the amount of migratory broodstock, and reduce model complexity;

[0061] A linear regression model is constructed by combining the independent variable, the dependent variable, the most influential variable and the interaction terms among the various parameters, and a regression analysis method (polynomial regression or piecewise regression) is introduced into the linear regression model to deal with the nonlinear effects of the various parameters.

[0062] The linear regression model is trained and evaluated by combining the k-fold cross validation method and the evaluation index, and the linear regression model is optimized according to the evaluation result to obtain a basic model.

[0063] In the basic model, the various parameters are used as state variables, and a state space sub-model is constructed by combining the state variables, the dependent variables and the interaction terms between the various parameters. The state space sub-model is used to dynamically simulate the impact of environmental changes on fish populations.

[0064] 3. If Figure 2 As shown, the model architecture of the basic model is designed using a long short-term memory network to obtain an initial prediction model, and then the initial prediction model is quantified using the Bayesian method to obtain an inferred prediction model. Including:

[0065] 3.1 Initial prediction model

[0066] Input layer: In chronological order, the early resource data and the various parameters are integrated into a multi-dimensional time series to obtain input data, and the input data is standardized or normalized to improve the convergence speed and prediction accuracy of the model.

[0067] LSTM layer: Based on the multi-layer LSTM structure (to improve the expressiveness of the model and enhance the ability to capture complex time series features), a gating mechanism (input gate, forget gate and output gate) is introduced to process the time dependency of the input data and learn the nonlinear dynamic features of the multidimensional time series.

[0068] Fully connected layer: maps the output of the LSTM layer to the target space, i.e., the predicted amount of migratory broodstock, to generate a prediction result.

[0069] Output layer: Use linear activation function to output prediction results.

[0070] The input layer, the LSTM layer, the fully connected layer and the output layer are integrated to obtain the model architecture of the basic model.

[0071] 3.2 Calculation of prediction model

[0072] 3.21 Bayesian LSTM

[0073] Parameter uncertainty: In traditional LSTM models, parameters such as weights and biases are fixed, whereas in Bayesian LSTM, these parameters are treated as probability distributions, meaning that the model not only outputs a predicted value but also provides confidence intervals for the prediction.

[0074] Bayesian Inference: Use Bayesian inference methods (such as variational inference or Markov Chain Monte Carlo methods) to estimate the posterior distribution of parameters, which can help quantify the uncertainty of the prediction and provide richer information to decision makers

[0075] 3.22 Loss Function

[0076] The loss function of Bayesian LSTM is based on the likelihood function, and the goal is to maximize the probability of the observed data. The log-likelihood of the migratory broodstock data is calculated using the likelihood function (a common method in statistics for estimating model parameters), and the model parameter estimates are obtained. The model parameter estimates are maximized to obtain the parameter probability distribution of the initial prediction model.

[0077] The regularization term of the Bayesian method (to prevent the model from overfitting) is used to constrain the prior distribution of the initial prediction model parameters.

[0078] 3.23 Initial prediction model training

[0079] The initial prediction model is trained using the comprehensive data set, model parameters are adjusted to maximize the likelihood function, and the initial prediction model is quantified to obtain an inferred prediction model.

[0080] The k-fold cross-validation method is used to evaluate the generalization ability of the model. Through different data segmentation, the performance of the model on unseen data is verified. During the verification process, the confidence interval of the prediction results is analyzed to evaluate the uncertainty of the model.

[0081] Bayesian methods allow expert knowledge to be introduced into the model to influence the prediction results through prior distribution. This combination helps to improve the reliability of the model when data is insufficient or uncertainty is high.

[0082] 4. If Figure 3 As shown, the MCMC method is combined to perform parameter estimation and model training on the inference prediction model to achieve optimization of the inference prediction model. It includes:

[0083] Based on the Bayesian framework, a normal distribution with a mean of 0 is selected as the prior distribution for the model parameters. The prior distribution is set for the model parameters to reflect the initial assumption that the parameters are unbiased. The mean and variance of the prior distribution can be adjusted based on domain knowledge.

[0084] Based on the prior distribution, the MCMC method is used to perform multiple iterative sampling on the model parameters to generate a parameter sample set close to the posterior distribution of the parameters, thereby completing the parameter estimation of the inference prediction model.

[0085] Among them, the MCMC method (Markov Chain Monte Carlo method) is a statistical sampling technique, which is mainly used to generate samples from complex probability distributions. The basic idea is to approximate the target probability distribution by constructing a Markov chain and generate samples from it. These samples can be used to estimate various statistics of the target distribution, such as expected value, variance, etc.

[0086] In each iterative sampling process, the joint probability of the likelihood function and the prior distribution is calculated by updating the parameter sample, and variational inference or other optimization strategies are used to accelerate the sampling process of the MCMC method, improve the training efficiency, and complete the model training of the inference prediction model.

[0087] The posterior distribution of the inferred prediction model is analyzed to obtain a confidence space for estimating the parameter error range. The inferred prediction model is verified and optimized by combining the confidence space, cross-validation method and evaluation indicators (mean square error, mean absolute error and determination coefficient, etc.).

[0088] 5. Use the optimized estimation and prediction model to estimate the number of migratory broodstock.

[0089] Therefore, the present invention provides a method for estimating the amount of migratory broodstock using the early fish resource quantity, which can combine multiple data sources, fully consider the complexity of the ecological environment, the dynamics of time series and the uncertainty of prediction, and realize the accurate estimation of the amount of migratory broodstock.

[0090] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0091] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A method for estimating the amount of migratory broodstock using the early fish resource amount, characterized in that: The following steps are involved: Collect early resource quantity data, migratory broodstock quantity data and various parameters, and process the collected data to obtain a comprehensive data set; Combining the above parameters, constructing a basic model for representing the relationship between the number of fish eggs, the number of fry and the number of migratory broodstock; Using a long short-term memory network to design a model architecture of the basic model, obtaining an initial prediction model, and then using a Bayesian method to quantify the initial prediction model to obtain an inferred prediction model; In combination with the MCMC method, parameter estimation and model training are performed on the inference prediction model to achieve optimization of the inference prediction model; The optimized estimation and prediction model is used to estimate the number of migratory broodstock.

2. The method for estimating the amount of migratory broodstock using the early fish resource amount according to claim 1, characterized in that: The early resource quantity data include the number of fish eggs and fry, and the various parameters include the proportion of male fish, the average relative fecundity of female fish in the corresponding river, the average weight of female fish in the corresponding river, the fertilization rate of the corresponding river, the hatching rate of fertilized eggs in the corresponding river and the survival rate of broodstock in the corresponding river.

3. The method for estimating the amount of migratory broodstock using the early fish resource amount according to claim 2, characterized in that: Combining the above parameters, a basic model is constructed to represent the relationship between the number of fish eggs, the number of fry and the number of migratory broodstock, including: The early resource quantity data and the various parameters are set as independent variables, the migratory parent fish quantity data is set as dependent variables, and the independent variables are standardized so that the mean value of the independent variables is 0 and the standard deviation is 1; Performing feature selection on the comprehensive data set using a feature selection algorithm or an importance scoring method to identify the most influential variables; A linear regression model is constructed by combining the independent variable, the dependent variable, the most influential variable and the interaction terms among the various parameters, and a regression analysis method is introduced into the linear regression model to handle the nonlinear effects of the various parameters; The linear regression model is trained and evaluated by combining the k-fold cross validation method and the evaluation index, and the linear regression model is optimized according to the evaluation result to obtain a basic model.

4. The method for estimating the amount of migratory broodstock using the early fish resource amount according to claim 3 is characterized in that: In combination with the various parameters, a basic model is constructed to represent the relationship between the number of fish eggs, the number of fry and the number of migratory parent fish, which also includes: in the basic model, the various parameters are used as state variables, and a state space sub-model is constructed in combination with the state variables, the dependent variables and the interaction terms between the various parameters, and the state space sub-model is used to dynamically simulate the impact of environmental changes on fish populations.

5. The method for estimating the amount of migratory broodstock using the early fish resource amount according to claim 4 is characterized in that: The model architecture of the basic model is designed using a long short-term memory network to obtain an initial prediction model, including: Integrate the early resource data and the various parameters into a multi-dimensional time series in chronological order to obtain input data, and standardize or normalize the input data to complete the design of the input layer; Based on the multi-layer LSTM structure, a gating mechanism is introduced to process the time dependency of the input data, and the nonlinear dynamic characteristics of the multi-dimensional time series are learned to complete the design of the LSTM layer; Map the output of the LSTM layer to the target space, generate prediction results, and complete the design of the fully connected layer; Use the linear activation function to output the prediction results and complete the design of the output layer; The input layer, the LSTM layer, the fully connected layer and the output layer are integrated to obtain the model architecture of the basic model.

6. A method for estimating the amount of migratory broodstock using the early fish resource amount according to claim 5, characterized in that: The initial prediction model is quantified using the Bayesian method to obtain an inferred prediction model, including: Calculating the log-likelihood of the migratory broodstock quantity data using a likelihood function to obtain model parameter estimates, maximizing the model parameter estimates, and obtaining a parameter probability distribution of the initial prediction model; The regularization term of the Bayesian method is used to constrain the prior distribution of the initial prediction model parameters, and then the initial prediction model is trained using the comprehensive data set, and the model parameters are adjusted to maximize the likelihood function, thereby completing the quantization of the initial prediction model and obtaining the inferred prediction model.

7. The method for estimating the amount of migratory broodstock using the early fish resource amount according to claim 6, characterized in that: In combination with the MCMC method, parameter estimation and model training are performed on the inference prediction model to achieve optimization of the inference prediction model, including: Based on the Bayesian framework, a normal distribution with a mean of 0 is selected as the prior distribution for the model parameters; Based on the prior distribution, the MCMC method is used to perform multiple iterations of sampling on the model parameters to generate a parameter sample set close to the posterior distribution of the parameters, thereby completing the parameter estimation of the inference prediction model; In each iterative sampling process, the joint probability of the likelihood function and the prior distribution is calculated by updating the parameter sample, and the sampling process of the MCMC method is accelerated by using variational inference to complete the model training of the inference prediction model; The posterior distribution of the inferred prediction model is analyzed to obtain a confidence space for estimating the parameter error range, and the inferred prediction model is verified and optimized by combining the confidence space, cross-validation method and evaluation index.

Images