A method and system for managing fish catches at fishing grounds based on big data analytics

Abstract

This invention proposes a method and system for managing fish catches at fishing grounds based on big data analysis. The method includes the following steps: the user selects a fishing ground and target fish species, obtains a fishing tag, and enters initial user fishing information; historical fishing information of users at the fishing ground is obtained and preprocessed to obtain a fishing feature dataset for the corresponding fishing ground; a recommendation model is built based on a machine learning network, and the fishing feature datasets of each corresponding fishing ground are input into the recommendation model for iterative training. The recommendation model outputs recommended fish species and recommended fishing spots for the current time at the corresponding fishing ground; the user selects a fishing spot from the recommended fishing spot information to fish, weighs the catch according to the fish species, obtains the unit price of the fish species to calculate the total price of the catch, and calls the payment interface to remotely settle the user's account; this method significantly improves the efficiency of fishing ground management and the user's fishing experience by collecting user fishing information, analyzing historical data, building a recommendation model, and realizing an automated settlement process.

Description

Technical Field

[0001] This invention relates to the field of big data technology, and in particular to a method and system for managing fish catches in fishing grounds based on big data analysis. Background Technology

[0002] Fishing ground return management refers to the scientific regulation and sustainable utilization of fish resources in fishing areas. It aims to balance fishery ecology, economic benefits and user experience. With the development of recreational fishing, fishing grounds often face problems such as reduced fish numbers and species imbalance, which leads to a decline in the fishing experience and an increase in ecological risks. Fishing ground return management ensures the ecological stability of fishing grounds by regularly releasing fish fry, controlling the amount of catch, and monitoring water quality and fish growth.

[0003] A method and system for weighing live fish, disclosed in CN114812767A, includes the following steps: acquiring the jumping frequency of the live fish using a weighing device; determining the weighing time based on the jumping frequency; weighing the live fish using the weighing time to obtain the standard weight of the fish; using the weight of the live fish landing on the weighing device as the initial weight of the fish and storing the initial weight; collecting multiple weight data of the live fish on the weighing device as temporary weights of the fish; comparing the temporary weights with the initial weights; discarding temporary weights that are less than the initial weights; averaging the remaining temporary weights to obtain the effective weight of the fish; and determining whether the effective weight is equal to the standard weight.

[0004] In traditional fishing ground management, due to the lack of effective means of collecting and analyzing user fishing information, fishing grounds struggle to provide users with accurate recommendations for fish species and fishing spots based on real-time environmental parameters and fish activity predictions, resulting in low fishing success rates and poor user experience. At the same time, traditional settlement processes are cumbersome and require manual operation, which is not only inefficient but also prone to errors and disputes, thereby reducing the efficiency of fishing ground management and the user fishing experience. Summary of the Invention

[0005] In view of this, the present invention proposes a method and system for managing fish return at fishing grounds based on big data analysis. By collecting user fishing information, analyzing historical data, building recommendation models, and realizing automated settlement processes, it can not only recommend the best fish species and fishing spots to users based on real-time environmental parameters and fish condition predictions, but also automatically settle accounts remotely, simplifying the transaction process and significantly improving the efficiency of fishing ground management and the user fishing experience.

[0006] In a first aspect, the present invention provides a method for managing fish return at fishing grounds based on big data analysis, comprising the following steps:

[0007] S1, the user selects the fishing spot and target fish species, obtains the fishing tag, and enters the initial user fishing information;

[0008] S2, obtain historical fishing information of users at the fishing ground, and perform data preprocessing to obtain the fishing feature dataset of the corresponding fishing ground;

[0009] S3. A recommendation model is built based on a machine learning network. The fishing feature datasets of each corresponding fishing spot are input into the recommendation model for iterative training. The recommendation model outputs the recommended fish species and recommended fishing spots for the corresponding fishing spot at the current time.

[0010] S4: The user selects a fishing spot from the recommended fishing spot information to fish. The catch is weighed according to the fish species, the unit price of the fish species is obtained to calculate the total price of the fish, and the payment interface is called to remotely settle the account of the user.

[0011] Based on the above technical solutions, preferably, in step S1, the user selects the fishing spot and target fish species, obtains a fishing tag, and enters initial user fishing information. This includes the user obtaining a fishing spot and fishing tag, and entering initial user fishing information via a mobile terminal. The user fishing information includes personal information, fishing spot information, target fish species, fishing time, environmental information, and fishing spot information. The personal information includes name, age, contact information, fishing experience, account, and preferred fish species. The fishing spot information includes fishing spot name, geographical location, water quality type, and development time. The fishing time includes season, time period, and holiday marking status. The environmental information includes water temperature, air pressure, temperature and humidity, weather conditions, pH value, and dissolved oxygen level. The fishing spot information includes fishing price, fishing spot number, weight of each fish species caught, and fishing spot status.

[0012] Based on the above technical solutions, preferably, step S2 involves obtaining historical user fishing information at the fishing grounds and performing data preprocessing to obtain the corresponding standard training dataset for fishing ground users, including the following sub-steps:

[0013] S21, retrieve the historical fishing information of all users at the corresponding fishing spot from the database based on the fishing spot name;

[0014] S22, perform data cleaning, transformation and merging of the historical fishing information of all users in the corresponding fishing spot to obtain standard user fishing data;

[0015] S23, extract and calibrate features from standard user fishing data, including fishing age, preferred fish species, season, time period, holiday marking status, water temperature, air pressure, temperature and humidity, weather conditions, pH value, dissolved oxygen, fishing price, fishing spot number, fish species type, and the weight of each fish species caught, to obtain standard feature data.

[0016] S24. Add the standard feature data to the dataset to obtain the fishing feature dataset for the corresponding fishing spot.

[0017] Based on the above technical solution, preferably, in step S23, the fish species type, fish catch weight, and fishing spot utility of the standard user fishing data are calibrated. The fishing spot utility calibration includes the following sub-steps:

[0018] The preset time window length is used to obtain the weight of each fish species caught at each fishing spot each day within the past time window, forming a fish catch sequence for each fish species.

[0019] Sort the total catch within the time window in ascending order to obtain the ascending sequence W of the catch for each fish species. s,f ={w1,w2,...,w T};

[0020] The percentile is preset, and the percentile position index of the catch sequence is calculated based on the catch sequence of each fish species and the preset percentile. The expression is:

[0021] S p = (p / 100) × (T+1)

[0022] In the formula, S p This is the percentile position index, where p is the preset percentile and T is the number of days in the time window;

[0023] Determine S p Is it an integer? If S p If the integer is S, then the Sth digit of the ascending sequence of the catch for the corresponding fish species is taken. p Each value is used to obtain the calibrated value of the fish catch at the fishing spot.

[0024] P p =W s,f [S p -1]

[0025] In the formula, P p To determine the target value for fish catch at the fishing spot, W s,f [] represents the ascending sequence of fish catches of species f at fishing spot s;

[0026] If S p If the value is not an integer, then the fish catch calibration value of the fishing spot is calculated using linear interpolation, and the expression is:

[0027]

[0028] In the formula, To round down, To round up;

[0029] Based on the preset percentiles, the caliber of fish catch at each fishing spot, and the ascending sequence of fish catches for the corresponding fish species, the utility value corresponding to the fishing spot is calculated, expressed as:

[0030] U s,f =(W s,f -P(100-p) (W s,f )) / (P p (W s,f )-P (100-p) (W s,f ))

[0031] In the formula, U s,f Let P be the utility value of fish species f at fishing spot s. (100-p) Let P be the calibrated value of the fish catch at the fishing position corresponding to the 100-p percentile. p =P (100-p) , then U s,t =0;

[0032] Based on the utility value, a pre-defined classification level is recommended. If the utility value is in the range of [0, 0.4), it is marked as an inefficient fishing spot. If the utility value is in the range of [0.4, 0.7), it is marked as a normal fishing spot. If the utility value is in the range of [0.7, 1], it is marked as an excellent fishing spot. The fishing spot number and fish species type are then associated.

[0033] Based on the above technical solutions, preferably, step S3, which involves constructing a recommendation model based on a machine learning network, inputting the fishing feature datasets of each corresponding fishing spot into the recommendation model for iterative training, and the recommendation model outputting the recommended fish species and recommended fishing spots for the corresponding fishing spot at the current time, includes the following sub-steps:

[0034] S31, a recommendation model is constructed based on a machine learning network. The recommendation model includes an input layer, a shared layer, a fish species branch module, and a fishing spot branch module. The output of the input layer is connected to the input of the shared layer. The fishing feature datasets of each corresponding fishing spot are input into the recommendation model for iterative training. The output of the shared layer is connected to the input of the fish species branch module and the fishing spot branch module, respectively, for extracting features from the fishing feature datasets of each corresponding fishing spot. The fish species branch module outputs the catch rate of each fish species, and the fishing spot branch module outputs the utility value of each fish species at each fishing spot.

[0035] S32, obtain the top three fish species in terms of catch rate as recommended fish species, obtain the fish species utility value corresponding to the fishing spot based on the recommended fish species, and select the fishing spot with the fish species utility value in the excellent fishing spot range as the recommended fishing spot.

[0036] S33 calculates the score for each recommended fishing spot by comprehensively considering the historical total catch, number of fishing attempts, and fishing price. s The expression is:

[0037] Score s =α×T s / max(T)+β×(1-C s / max(C))+γ×((1-(Ps -min(P)) /

[0038] (max(P)-min(P)))

[0039] In the formula, T s C represents the historical total catch at fishing spot s. s P represents the number of fishing attempts at fishing spot s. s Let s be the current fishing price, T be the historical total catch of the fishing spot, C be the maximum number of times the fishing spot can be used, min(P) be the lowest fishing price of the fishing spot, max(P) be the highest fishing price of the fishing spot, α be the weighting coefficient of the catch, β be the weighting coefficient of the number of fishing attempts, and γ be the weighting coefficient of the fishing price.

[0040] S34. The evaluation values ​​of each recommended fishing spot are associated with the fishing spot number and the recommended fish species in descending order to obtain the recommended fishing list information for users to view.

[0041] Based on the above technical solution, preferably, after step S34, the method further includes: determining whether the recommended fish species at the current time of the fishing spot is consistent with the target fish species selected by the user; if consistent, obtaining the available fishing spots in the recommended fishing list information according to the fishing spot status information, and selecting the fishing spot with the highest evaluation value as the recommended fishing spot and allocating it to the user; if inconsistent, determining whether the user has changed the target fish species to the recommended fish species; if changed, obtaining the fishing spot with the highest evaluation value as the recommended fishing spot and allocating it to the user; otherwise, obtaining the utility value corresponding to the target fish species of each fishing spot from the time window to the current time, and selecting the fishing spots with fish species utility values ​​in the excellent fishing spot range as candidate fishing spots; obtaining the available fishing spots in the candidate fishing spots according to the fishing spot status information and calculating the evaluation value of the available fishing spots in each candidate fishing spot, and selecting the fishing spot with the highest evaluation value as the recommended fishing spot and allocating it to the user.

[0042] Based on the above technical solutions, preferably, the shared layer includes a first feature extraction unit, a second feature extraction unit, and a third feature extraction unit, which are connected sequentially. The third feature extraction unit outputs shared feature data.

[0043] The expression for the first feature extraction unit is:

[0044] h1 = σ(W1X + b1)

[0045] In the formula, X represents the standard feature data corresponding to each fishing spot, σ is the Sigmoid function, and W1 is the weight matrix with dimension R. 256×64 b1 is the bias vector;

[0046] The expression for the second feature extraction unit is:

[0047] h2 = LayerNorm(W2h1 + b2)

[0048] In the formula, h1 is the output feature of the first feature extraction unit, and W2 is the weight matrix with dimension R. 128×256 b2 is the bias vector, and LayerNorm() is the layer normalization function;

[0049] The expression for the third feature extraction unit is:

[0050] h3=Dropout(σ(W3h2+b3), c=0.3)

[0051] In the formula, h2 is the output feature of the second feature extraction unit, and W3 is the weight matrix with dimension R. 64×128 b3 is the bias vector, Dropout() is the regularization function, and c is the neuron dropout probability.

[0052] Based on the above technical solution, preferably, the fish species branching module includes a first fully connected layer, a second fully connected layer, and a first output layer, wherein the input of the first fully connected layer is connected to the output of the third feature extraction unit, reducing the 64-dimensional features to 32-dimensional features, as expressed in the following expression:

[0053] z1=W4h3+b4

[0054] In the formula, W4 is the weight matrix with dimension R. 32×64 b4 is the bias vector with dimension R. 32 ;

[0055] The output of the first fully connected layer is connected to the input of the second fully connected layer, mapping the 32-dimensional features to the m-dimensional representation of the number of fish species. The expression is as follows:

[0056] z2 = RELU(z1)W5 + b5

[0057] In the formula, RELU() is the ReLU function, W5 is the weight matrix with dimension R. m×32 b5 is the bias vector with dimension R. m ;

[0058] The output of the second fully connected layer is connected to the first output layer. The first output layer is used to output the probability of each fish species, and its expression is:

[0059]

[0060] In the formula, p i Let be the predicted probability of the i-th fish species, and let be the sum of the predicted probabilities of all fish species, which is 1.

[0061] Based on the above technical solutions, preferably, the fishing position branch module includes a third fully connected layer, a fourth fully connected layer, and a second output layer, wherein,

[0062] The input of the third fully connected layer is connected to the output of the third feature extraction unit, reducing the 64-dimensional features to 16-dimensionality, as shown in the expression:

[0063] u1 = W6h3 + b6

[0064] In the formula, W6 is the weight matrix with dimension R. 16×64 b6 is the bias vector with dimension R. 16 ;

[0065] The output of the third fully connected layer is connected to the input of the fourth fully connected layer, mapping the 16-dimensional features to a 1-dimensional scalar, as shown in the expression:

[0066] y=σ(u1)W7+b7

[0067] In the formula, σ is the Sigmoid function, W7 is the weight matrix with dimension R. 1×16 b7 is the bias vector with dimension R;

[0068] The output of the fourth fully connected layer is connected to the input of the second output layer, and the second output layer outputs the fishing spot utility value, y∈[0,1].

[0069] Secondly, the present invention also provides a fishing ground return management system based on big data analysis, implemented using a fishing ground return management method based on big data analysis, the system comprising:

[0070] The information entry module is used by users to select the fishing spot and target fish species, obtain the fishing tag, and enter the initial user fishing information.

[0071] The data processing module is used to acquire historical fishing information of users at the fishing grounds and perform data preprocessing to obtain the fishing feature dataset of the corresponding fishing grounds.

[0072] The recommendation module is used to build a recommendation model based on machine learning networks. The fishing feature datasets of each corresponding fishing spot are input into the recommendation model for iterative training. The recommendation model outputs the recommended fish species and recommended fishing spots for the corresponding fishing spot at the current time.

[0073] The settlement module is used by users to select fishing spots from the recommended fishing spots, weigh the catch according to the fish species, obtain the unit price of the fish species to calculate the total price of the catch, and call the payment interface to remotely settle the account of the user.

[0074] The fishing ground return management method and system based on big data analysis of the present invention have the following advantages over the prior art:

[0075] (1) By collecting user fishing information, analyzing historical data, building recommendation models, and implementing automated settlement processes, it can not only recommend the best fish species and fishing spots to users based on real-time environmental parameters and fish condition predictions, thus increasing the success rate of fishing; but also the automated weighing, pricing, and remote settlement simplifies the transaction process, ensures the convenience of transactions, and enhances user stickiness, thereby significantly improving the efficiency of fishing ground management and the user fishing experience.

[0076] (2) By calculating the utility value independently for each fish species, the fishing effect of each fishing spot for different fish species can be accurately evaluated, avoiding the ambiguity of the overall evaluation and providing users with more accurate fishing spot recommendations. The utility value is calculated based on percentiles, effectively filtering out data interference from extreme weather or occasional big catches, thus improving the accuracy of recommendations. (3) By building a recommendation model through machine learning networks, it can accurately capture complex factors such as the fishing environment and fish species distribution, thus improving the accuracy of recommended fish species and recommended fishing spots. By comprehensively evaluating fishing spots, the fishing spot with the highest cost performance and most suitable for the current fishing conditions can be selected, providing users with more accurate and personalized fishing recommendations, thus improving the success rate and satisfaction of fishing. Attached Figure Description

[0077] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art 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.

[0078] Figure 1 The flowchart shows the fishing ground return management method based on big data analysis according to the present invention.

[0079] Figure 2 This is a structural diagram of the recommended model for the big data analysis-based fish return management method of the present invention. Detailed Implementation

[0080] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0081] Firstly, such as Figure 1 As shown, this invention provides a method for managing fish catches in fishing grounds based on big data analysis, including the following steps:

[0082] S1, the user selects the fishing spot and target fish species, obtains the fishing tag, and enters the initial user fishing information.

[0083] In this embodiment, step S1 includes the user obtaining a fishing spot tag and entering initial user fishing information via a mobile device. The user fishing information includes personal information, fishing spot information, target fish species, fishing time, environmental information, and fishing spot information. The personal information includes name, age, contact information, fishing experience, account, and preferred fish species. The fishing spot information includes fishing spot name, geographical location, water quality type, and development time. The fishing time includes season, time period, and holiday marking status. The environmental information includes water temperature, air pressure, temperature and humidity, weather conditions, pH value, and dissolved oxygen level. The fishing spot information includes fishing price, fishing spot number, weight of each fish species caught, and fishing spot status.

[0084] It should be noted that users select their target fishing spot and species through the fishing spot management platform or mobile application. After selecting the fishing spot, users need to go to the fishing spot management office or obtain a fishing tag through the online reservation system. The fishing tag serves as an access pass for fishing activities, recording not only the user's basic information and fishing permissions, but also being linked to the resource allocation and management of the fishing spot. After obtaining the fishing tag, users need to enter their initial fishing information through the mobile application, which can be done through the fishing spot's APP, mini-program, or a third-party fishing service platform. This information comprehensively covers all aspects of fishing activities, providing strong support for fishing spot management, user services, and data analysis.

[0085] Personal information includes name, age, contact information, fishing experience, account information, and preferred fish species, which helps the fishing ground understand the user's background and provide personalized services. Fishing ground information includes the fishing ground name, geographical location, water quality type (freshwater, brackish water, mixed water, etc.), and development time. Accurate entry of fishing ground information helps the fishing ground allocate resources, maintain the environment, and guide users, ensuring the smooth progress of fishing activities. Users need to specify their target fish species for this fishing trip, which helps the fishing ground arrange fishing spots, prepare bait, and provide fishing advice. Fishing time includes season (e.g., spring, summer, autumn, and winter) and time of day (e.g., morning, afternoon). The system records fishing time (including evenings and holidays) and user activity status (whether it is a weekend or holiday). This helps the fishing ground analyze user behavior patterns, optimize fishing spot allocation and activity arrangements. Detailed environmental information records include water temperature, air pressure, temperature and humidity, weather conditions (sunny, rainy, and cloudy), pH value, and dissolved oxygen levels. This environmental information helps the system predict fish activity and remind users to adjust their fishing strategies. Fishing spot information includes fishing price (i.e., spot fee), spot number, weight of each fish species caught (historical data), and spot status (available, booked, and under maintenance). Real-time updates of fishing spot information help users quickly understand the fishing ground situation and make reasonable choices.

[0086] In addition, an electronic fishing tag is obtained, and the ID corresponding to the electronic fishing tag is unique. When a user needs to fish, the obtained electronic fishing tag is associated with the user's information. After the fishing is settled, the system automatically unbinds the electronic fishing tag from the user's information.

[0087] Understandably, fishing grounds can use the collected user fishing information to allocate resources, maintain the environment, and arrange activities, thereby improving the operational efficiency of the fishing grounds and providing strong data support for fishing ground data analysis.

[0088] S2: Obtain historical fishing information of users at the fishing spot and perform data preprocessing to obtain the fishing feature dataset of the corresponding fishing spot.

[0089] Step S2 includes the following sub-steps:

[0090] S21, retrieve the historical fishing information of all users at the corresponding fishing spot from the database based on the fishing spot name;

[0091] S22, perform data cleaning, transformation and merging of the historical fishing information of all users in the corresponding fishing spot to obtain standard user fishing data;

[0092] S23, extract and calibrate features from standard user fishing data, including fishing age, preferred fish species, season, time period, holiday marking status, water temperature, air pressure, temperature and humidity, weather conditions, pH value, dissolved oxygen, fishing price, fishing spot number, fish species type, and the weight of each fish species caught, to obtain standard feature data.

[0093] S24. Add the standard feature data to the dataset to obtain the fishing feature dataset for the corresponding fishing spot.

[0094] It should be noted that by using the fishing spot name as the query keyword, the database storing user fishing information is accessed to retrieve and extract the historical fishing records of all users at that fishing spot. This provides comprehensive basic data for subsequent analysis. The historical fishing information of all users at the fishing spot is cleaned to remove duplicate records, ensuring that each data point is unique. Missing values ​​are filled using the median, and erroneous data is corrected. The cleaned data is then standardized, and related data from different data sources or tables are merged. Data from multiple fishing activities is integrated to form a panoramic view of the user's fishing history, improving data quality, ensuring data consistency and usability, and laying a solid foundation for subsequent feature extraction.

[0095] In step S23 of this embodiment, the fish species type, fish catch weight, and fishing spot utility of the standard user fishing data are calibrated. The fishing spot utility calibration includes the following sub-steps:

[0096] The preset time window length is used to obtain the weight of each fish species caught at each fishing spot each day within the past time window, forming a fish catch sequence for each fish species.

[0097] Sort the total catch within the time window in ascending order to obtain the ascending sequence W of the catch for each fish species. s,f ={w1,w2,...,w T};

[0098] The percentile is preset, and the percentile position index of the catch sequence is calculated based on the catch sequence of each fish species and the preset percentile. The expression is:

[0099] S p = (p / 100) × (T+1)

[0100] In the formula, S p This is the percentile position index, where p is the preset percentile and T is the number of days in the time window;

[0101] Determine S p Is it an integer? If S p If the integer is S, then the Sth digit of the ascending sequence of the catch for the corresponding fish species is taken. p Each value is used to obtain the calibrated value of the fish catch at the fishing spot.

[0102] P p =W s,f [S p -1]

[0103] In the formula, P p To determine the target value for fish catch at the fishing spot, W s,f [] represents the ascending sequence of fish catches of species f at fishing spot s;

[0104] If S p If the value is not an integer, then the fish catch calibration value of the fishing spot is calculated using linear interpolation, and the expression is:

[0105]

[0106] In the formula, To round down, To round up;

[0107] Based on the preset percentiles, the caliber of fish catch at each fishing spot, and the ascending sequence of fish catches for the corresponding fish species, the utility value corresponding to the fishing spot is calculated, expressed as:

[0108] U s,f =(W s,f -P (100-p) (W s,f )) / (P p (W s,f )-P (100-p) (W s,f))

[0109] In the formula, U s,f Let P be the utility value of fish species f at fishing spot s. (100-p) Let P be the calibrated value of the fish catch at the fishing position corresponding to the 100-p percentile. p =P (100-p) , then U s,t =0;

[0110] Based on the utility value, a pre-defined classification level is recommended. If the utility value is in the range of [0, 0.4), it is marked as an inefficient fishing spot. If the utility value is in the range of [0.4, 0.7), it is marked as a normal fishing spot. If the utility value is in the range of [0.7, 1], it is marked as an excellent fishing spot. The fishing spot number and fish species type are then associated.

[0111] It should be noted that the time window length is the past 7 days. The weight of each fish species caught at each fishing spot each day within this time window is obtained to form a catch sequence for each fish species, reflecting the fishing effect of the fishing spot at different time periods. The total catch within the time window is sorted in ascending order to obtain an ascending catch sequence for each fish species, preparing for subsequent calculations of percentiles and fishing spot utility values. The preset percentile is 75. Based on the catch sequence for each fish species and the preset percentile, the percentile position index of the catch sequence is calculated to determine the position of the corresponding percentile in the ascending catch sequence, thus obtaining the catch calibration value, reflecting the fishing effect of the fishing spot at that percentile. The utility value corresponding to the fishing spot is used to quantify the fishing effect of the fishing spot for different fish species, forming a utility value that can be used for comparison. The fishing spot utility value is then converted into an intuitive recommendation level and associated with the fishing spot number and fish species type for easy user understanding and selection.

[0112] It should be noted that by calculating utility values ​​independently for each fish species, the effectiveness of each fishing spot for different fish species can be accurately evaluated, avoiding the ambiguity of the overall evaluation and providing users with more accurate fishing spot recommendations. Utility values ​​calculated based on percentiles effectively filter out data interference from extreme weather or occasional big catches, thus improving the accuracy of recommendations.

[0113] In addition, the sliding time window design can quickly capture the migration patterns of fish schools. When the temperature rises suddenly and causes fish schools to move to deeper water areas, the utility value of deep water fishing spots is automatically updated within 24 hours. This allows for real-time response to environmental changes and adjustments to the length of the time window, thereby improving the accuracy of model recommendations.

[0114] like Figure 2 As shown in Figure S3, a recommendation model is built based on a machine learning network. The fishing feature datasets of each corresponding fishing spot are input into the recommendation model for iterative training. The recommendation model outputs the recommended fish species and recommended fishing spots for the corresponding fishing spot at the current time.

[0115] Step S3 includes the following sub-steps:

[0116] S31, a recommendation model is constructed based on a machine learning network. The recommendation model includes an input layer, a shared layer, a fish species branch module, and a fishing spot branch module. The output of the input layer is connected to the input of the shared layer. The fishing feature datasets of each corresponding fishing spot are input into the recommendation model for iterative training. The output of the shared layer is connected to the input of the fish species branch module and the fishing spot branch module, respectively, for extracting features from the fishing feature datasets of each corresponding fishing spot. The fish species branch module outputs the catch rate of each fish species, and the fishing spot branch module outputs the utility value of each fish species at each fishing spot.

[0117] S32, obtain the top three fish species in terms of catch rate as recommended fish species, obtain the fish species utility value corresponding to the fishing spot based on the recommended fish species, and select the fishing spot with the fish species utility value in the excellent fishing spot range as the recommended fishing spot.

[0118] S33 calculates the score for each recommended fishing spot by comprehensively considering the historical total catch, number of fishing attempts, and fishing price. s The expression is:

[0119] Score s =α×T s / max(T)+β×(1-C s / max(C))+γ×((1-(P s -min(P)) /

[0120] (max(P)-min(P)))

[0121] In the formula, T s C represents the historical total catch at fishing spot s. s P represents the number of fishing attempts at fishing spot s. s Let s be the current fishing price, T be the historical total catch of the fishing spot, C be the maximum number of times the fishing spot can be used, min(P) be the lowest fishing price of the fishing spot, max(P) be the highest fishing price of the fishing spot, α be the weighting coefficient of the catch, β be the weighting coefficient of the number of fishing attempts, and γ be the weighting coefficient of the fishing price.

[0122] S34. The evaluation values ​​of each recommended fishing spot are associated with the fishing spot number and the recommended fish species in descending order to obtain the recommended fishing list information for users to view.

[0123] It should be noted that the input layer receives fishing feature datasets for each corresponding fishing spot, including information such as fishing age, preferred fish species, season, time period, holiday marking status, water temperature, air pressure, temperature and humidity, weather conditions, pH value, dissolved oxygen level, fishing price, fishing spot number, fish species type, and the weight of each fish species caught. The shared layer performs preliminary feature extraction and shared representation on the input feature dataset, providing basic features for subsequent branch modules. The fish species branch module further extracts fish species-related features based on the output of the shared layer and outputs the catch rate of each fish species. The fishing spot branch module, also based on the output of the shared layer, extracts features related to the fishing spot and outputs the utility value of each fish species at the fishing spot. The fishing feature datasets of each corresponding fishing ground are input into the recommendation model for iterative training. The model parameters are optimized through backpropagation to improve the model's prediction accuracy. Taking into account the historical total catch, number of fishing attempts, and fishing price of the recommended fishing spot, the evaluation value of each fishing spot is calculated using a weighted summation. The weight coefficient α of the catch represents the relative importance of the historical total catch in the evaluation value. When α is large, the evaluation value focuses more on reflecting the fishing... Historical catch data for a fishing spot is suitable for users who prioritize catch quantity. The weighting coefficient β for the number of fishing attempts indicates the relative importance of the number of fishing attempts in the evaluation value. When β is large, the evaluation value focuses more on reflecting the popularity or fishing activity of the fishing spot, which is suitable for users who like a lively atmosphere and pursue a good fishing experience. The weighting coefficient γ for the fishing price indicates the relative importance of the fishing price in the overall evaluation of the fishing spot's quality. The larger the γ value, the greater the impact of the fishing price on the evaluation value. α+β+γ=1. By comprehensively evaluating the fishing spots, the fishing spot with the highest cost performance and the most suitable for the current fishing conditions can be selected.

[0124] Understandably, building recommendation models through machine learning networks can accurately capture complex factors such as the fishing environment and fish species distribution, improving the accuracy of recommended fish species and fishing spots. Furthermore, by comprehensively evaluating fishing spots, the most cost-effective and suitable spots for the current fishing conditions can be selected, providing users with more accurate and personalized fishing recommendations, thereby increasing fishing success rates and satisfaction.

[0125] Step S34 is followed by: determining whether the recommended fish species at the current fishing spot are consistent with the target fish species selected by the user. If they are consistent, then the available fishing spots in the recommended fishing list are obtained based on the fishing spot status information, and the fishing spot with the highest evaluation value among the available fishing spots is selected as the recommended fishing spot and assigned to the user. If they are inconsistent, then the user is determined whether the target fish species has been changed to the recommended fish species. If the user has changed, then the fishing spot with the highest evaluation value among the available fishing spots is selected as the recommended fishing spot and assigned to the user. Otherwise, the utility value of the target fish species at each fishing spot from the time window to the current time is obtained, and the fishing spots with the fish species utility value in the excellent fishing spot range are selected as candidate fishing spots. Based on the fishing spot status information, the available fishing spots among the candidate fishing spots are obtained, and the evaluation value of the available fishing spots among the candidate fishing spots is calculated. The fishing spot with the highest evaluation value is selected as the recommended fishing spot and assigned to the user.

[0126] The system first determines whether the recommended fish species at the current fishing spot match the user's selected target fish species, ensuring that the recommended fishing spot matches the user's fishing preferences and improving user satisfaction. If the recommended fish species matches the user's target fish species, the system prioritizes the available fishing spot with the highest evaluation value to provide the best fishing experience. If the recommended fish species does not match the user's target fish species, the system determines whether the user is willing to change their target fish species to the recommended fish species. If the user is willing to adjust their fishing target, the system still provides the optimal fishing spot recommendation. If the user insists on fishing for the target fish species, the system calculates the utility value and evaluation value of the target fish species to find the most suitable fishing spot for the user, improving satisfaction.

[0127] In this embodiment, the shared layer includes a first feature extraction unit, a second feature extraction unit, and a third feature extraction unit, which are connected sequentially. The third feature extraction unit outputs shared feature data.

[0128] The expression for the first feature extraction unit is:

[0129] h1 = σ(W1X + b1)

[0130] In the formula, X represents the standard feature data corresponding to each fishing spot, σ is the Sigmoid function, and W1 is the weight matrix with dimension R. 256×64 b1 is the bias vector;

[0131] The expression for the second feature extraction unit is:

[0132] h2 = LayerNorm(W2h1 + b2)

[0133] In the formula, h1 is the output feature of the first feature extraction unit, and W2 is the weight matrix with dimension R. 128×256 b2 is the bias vector, and LayerNorm() is the layer normalization function;

[0134] The expression for the third feature extraction unit is:

[0135] h3=Dropout(σ(W3h2+b3), c=0.3)

[0136] In the formula, h2 is the output feature of the second feature extraction unit, and W3 is the weight matrix with dimension R. 64×128 b3 is the bias vector, Dropout() is the regularization function, and c is the neuron dropout probability.

[0137] It should be noted that by sequentially connecting the first feature extraction unit, the second feature extraction unit, and the third feature extraction unit, the system can gradually extract more abstract and advanced feature representations from the original standard feature data. This hierarchical feature extraction method is beneficial for capturing complex patterns and relationships in the data. Furthermore, the shared feature data output by the third feature extraction unit is used by both the fish species branch module and the fishing position branch module, realizing feature sharing and reuse. This helps reduce redundant calculations, improve model efficiency, and enables the two branch modules to make predictions based on the same feature foundation.

[0138] In addition, the layer normalization function is used in the second feature extraction unit, which helps to accelerate the convergence speed of the model and improve the training efficiency of the model. At the same time, layer normalization can also enhance the robustness of the model, making the model less sensitive to small changes in the input data. Furthermore, the regularization function is used in the third feature extraction unit, which helps to prevent the model from overfitting, improves the model's generalization ability, and enables the model to maintain good prediction performance when facing new and unseen data.

[0139] In this embodiment, the fish species branching module includes a first fully connected layer, a second fully connected layer, and a first output layer. The input of the first fully connected layer is connected to the output of the third feature extraction unit, reducing the 64-dimensional features to 32 dimensions. The expression is as follows:

[0140] z1=W4h3+b4

[0141] In the formula, W4 is the weight matrix with dimension R. 32×64 b4 is the bias vector with dimension R. 32 ;

[0142] The output of the first fully connected layer is connected to the input of the second fully connected layer, mapping the 32-dimensional features to the m-dimensional representation of the number of fish species. The expression is as follows:

[0143] z2 = RELU(z1)W5 + b5

[0144] In the formula, RELU() is the ReLU function, W5 is the weight matrix with dimension R. m×32 b5 is the bias vector with dimension R. m ;

[0145] The output of the second fully connected layer is connected to the first output layer. The first output layer is used to output the probability of each fish species, and its expression is:

[0146]

[0147] In the formula, p i Let be the predicted probability of the i-th fish species, and let be the sum of the predicted probabilities of all fish species, which is 1.

[0148] It should be noted that the fish species branching module transforms shared feature data into predicted probabilities for each fish species through a layer-by-layer mapping of the first fully connected layer, the second fully connected layer, and the first output layer. This enables the model to learn the complex relationships between different fish species and features, thereby achieving accurate fish species prediction. The first output layer outputs the predicted probabilities for each fish species, and the sum of the predicted probabilities for all fish species is 1. This helps users understand the likelihood of different fish species appearing and provides more comprehensive information for fishing decisions.

[0149] In this embodiment, the fishing spot branch module includes a third fully connected layer, a fourth fully connected layer, and a second output layer, wherein...

[0150] The input of the third fully connected layer is connected to the output of the third feature extraction unit, reducing the 64-dimensional features to 16-dimensionality, as shown in the expression:

[0151] u1 = W6h3 + b6

[0152] In the formula, W6 is the weight matrix with dimension R. 16×64 b6 is the bias vector with dimension R. 16 ;

[0153] The output of the third fully connected layer is connected to the input of the fourth fully connected layer, mapping the 16-dimensional features to a 1-dimensional scalar, as shown in the expression:

[0154] y=σ(u1)W7+b7

[0155] In the formula, σ is the Sigmoid function, W7 is the weight matrix with dimension R. 1×16 b7 is the bias vector with dimension R;

[0156] The output of the fourth fully connected layer is connected to the input of the second output layer, and the second output layer outputs the fishing spot utility value, y∈[0,1].

[0157] It should be noted that the fishing spot branch module transforms shared feature data into fishing spot utility values ​​through layer-by-layer mapping of the third fully connected layer, the fourth fully connected layer, and the second output layer. This reflects the attractiveness and potential value of the fishing spot for fishing, helping users to choose the optimal fishing spot. The fishing spot utility values ​​output by the second output layer are normalized to the [0,1] interval, making the utility values ​​between different fishing spots comparable and allowing users to more intuitively understand the differences in advantages and disadvantages of each fishing spot.

[0158] S4: The user selects a fishing spot from the recommended fishing spot information to fish. The catch is weighed according to the fish species, the unit price of the fish species is obtained to calculate the total price of the fish, and the payment interface is called to remotely settle the account of the user.

[0159] It should be noted that users select a fishing spot based on the system's recommended information. The system records the selected spot. After fishing, users take their catch to a designated weighing area, insert the electronic fishing tag into the scale, and select the fish species. The weighing device records the weight and transmits the data to the system. The system then uses this information to determine the corresponding fish species price, which can be dynamically adjusted based on market conditions and seasonal changes. The system calculates the total return price based on the weight and price, retrieves the payment interface, and transmits the total return price to the payment platform. The payment platform sends the total return price to the fishing area manager for confirmation based on the user's account information. After confirmation, the payment platform deducts the payment from the user's account and sends the settlement result back to the system. The system records the settlement result for future queries and statistics and stores the user's fishing information in the database.

[0160] This embodiment significantly improves the efficiency of fishing ground management and the user fishing experience by comprehensively collecting user fishing information, accurately analyzing historical data, constructing an intelligent recommendation model, and implementing an automated settlement process. The system not only provides fishing grounds with a scientific basis for resource allocation but also recommends optimal fish species and fishing spots to users based on real-time environmental parameters and fish behavior predictions, increasing the success rate of fishing. Simultaneously, automated weighing, pricing, and remote settlement functions simplify the transaction process, ensuring fairness and convenience, and enhancing user stickiness. Furthermore, this method continuously optimizes the recommendation algorithm through data accumulation and analysis, promoting the continuous improvement of fishing ground service quality and achieving a win-win situation for both fishing grounds and users.

[0161] Secondly, the present invention also provides a fishing ground fish return management system based on big data analysis, implemented using a fishing ground fish return management method based on big data analysis, the system comprising:

[0162] The information entry module is used by users to select the fishing spot and target fish species, obtain the fishing tag, and enter the initial user fishing information.

[0163] The data processing module is used to acquire historical fishing information of users at the fishing grounds and perform data preprocessing to obtain the fishing feature dataset of the corresponding fishing grounds.

[0164] The recommendation module is used to build a recommendation model based on machine learning networks. The fishing feature datasets of each corresponding fishing spot are input into the recommendation model for iterative training. The recommendation model outputs the recommended fish species and recommended fishing spots for the corresponding fishing spot at the current time.

[0165] The settlement module is used by users to select fishing spots from the recommended fishing spots, weigh the catch according to the fish species, obtain the unit price of the fish species to calculate the total price of the catch, and call the payment interface to remotely settle the account of the user.

[0166] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0167] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the system and modules described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0168] In the embodiments provided by this invention, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0169] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0170] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0171] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, ROM, RAM, magnetic disks, or optical disks.

[0172] Furthermore, it should be noted that in the system and method of the present invention, it is obvious that the components or steps can be decomposed and / or recombined. These decompositions and / or recombinations should be considered equivalent solutions of the present invention. Moreover, the steps performing the above series of processes can naturally be executed in the order described, but are not necessarily required to be executed in chronological order; some steps can be executed in parallel or independently of each other. Those skilled in the art will understand that all or any step or component of the method and apparatus of the present invention can be implemented in any computing device (including processors, storage media, etc.) or network of computing devices, in hardware, firmware, software, or a combination thereof. This is something that those skilled in the art can achieve by using their basic programming skills after reading the description of the present invention.

[0173] Therefore, the object of the present invention can also be achieved by running a program or a set of programs on any computing system. The computing system can be a known general-purpose system. Therefore, the object of the present invention can also be achieved simply by providing a program product containing program code for implementing the method or apparatus. That is, such a program product also constitutes the present invention, and the storage medium storing such a program product also constitutes the present invention. Obviously, the storage medium can be any known storage medium or any storage medium developed in the future. It should also be noted that in the apparatus and method of the present invention, it is obvious that the components or steps can be decomposed and / or recombined. These decompositions and / or recombinations should be considered equivalent to the present invention. Furthermore, the steps for performing the above series of processes can naturally be performed in the order described, but are not necessarily required to be performed in chronological order. Some steps can be performed in parallel or independently of each other.

[0174] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A fishing ground fish returning management method based on big data analysis, characterized by, The method comprises the following steps: S1, a user selects a fishing ground and a target fish species for fishing, obtains a fishing card, and inputs initial user fishing information; S2, historical user fishing information of the fishing ground is obtained, and data preprocessing is performed to obtain fishing feature data sets corresponding to the fishing ground; It includes: fish species type, fish catch weight and fishing site utility calibration of user historical fishing data, wherein the fishing site utility calibration comprises the following sub-steps: A preset time window length is set, the fishing weight of each fish species of each fishing site in the past time window is obtained, and the fish catch sequence of each fish species is formed; Sort the total amount of fish in the time window in ascending order to get the ascending sequence of fish catch of each fish species W s,f = {w1, w2,..., w T} A preset percentile is set, and the percentile position index of the fish catch sequence is calculated according to the fish catch sequence of each fish species and the preset percentile, and the expression is: S p = (p / 100) x (T+1) In the formula, S p is a percentile position index, p is a preset percentile, and T is the number of days of a time window. determination S p whether S is an integer, if S p is an integer, the S p th value in the ascending sequence of fish catch of the corresponding fish species is taken to obtain the fishing site fish catch index value; P p = W s,f [ S p -1] where P p is the catchability at position s, W s,f [] is the ascending order sequence of the catch of fish f at position s. If S p is not an integer, the linear interpolation method is used to calculate the fishing position fish catch calibration value, and the expression is: wherein is rounded down, is rounded up; According to the preset percentile, the fishing catch calibration value of the fishing site and the ascending sequence of the fish catch of the corresponding fish species, the utility value corresponding to the fishing site is calculated, and the expression is: U s,f = (W s,f -P (100-p) (W s,f )) / (P p (W s,f )-P (100-p) (W s,f )) In the formula, U s,f is the utility value of the fishing site s for the fish species f, P (100-p) is the 100-pth percentile of the fish catch index of the fishing site s, and if P p = P (100-p) , then U s,t is 0. According to the utility value, a preset recommended classification level is set, if the utility value is located in the interval [0, 0.4), it is marked as a low-efficiency fishing site, if the utility value is located in the interval [0.4, 0.7), it is marked as a general fishing site, if the utility value is located in the interval [0.7, 1], it is marked as an excellent fishing site, and the fishing site number and fish species type are associated; S3, a recommendation model is constructed based on a machine learning network, and the fishing feature data sets of each corresponding fishing ground are input into the recommendation model for iterative training, and the recommendation model outputs the recommended fish species and the recommended fishing site information of the corresponding fishing ground at the current time; S4, the user selects the fishing site in the recommended fishing site information for fishing, weighs the fish catch according to the fish species, obtains the total price of the fish catch, and calls the payment interface to remotely settle the user's account.

2. The big data analysis-based fishing ground fish return management method of claim 1, characterized by: In step S1, the user selects a fishing ground and a target fish species for fishing, obtains a fishing card, and inputs initial user fishing information, including obtaining a fishing card of a fishing ground by a user, and inputting initial user fishing information through a mobile terminal. The user fishing information includes personal information, fishing ground information, target fish species, fishing time, environment information and fishing site information. The personal information includes name, age, contact information, fishing age, account and preferred fish species. The fishing ground information includes fishing ground name, geographical location, water quality type and development time. The fishing time includes season, time period and holiday marking state. The environment information includes water temperature, air pressure, temperature and humidity, weather condition, pH value and dissolved oxygen content. The fishing site information includes fishing price, fishing site number, fishing weight of each fish species and fishing site state.

3. The fishing ground fish management method based on big data analysis according to claim 2, characterized by: In step S2, the historical user fishing information of the fishing ground is obtained, and data preprocessing is performed to obtain the corresponding fishing ground user standard training data set, comprising the following sub-steps: S21, obtaining the historical fishing information of all users of the corresponding fishing ground from the database according to the fishing ground name; S22, performing data cleaning, conversion and merging processing on the historical fishing information of all users of the corresponding fishing ground to obtain standard user fishing data; S23, performing feature extraction and calibration on the fishing age, preferred fish species, season, time period, holiday marking state, water temperature, air pressure, temperature and humidity, weather condition, pH value, dissolved oxygen content, fishing price, fishing site number, fish species type and fishing weight information of each fish species in the standard user fishing data to obtain standard feature data; S24, add the standard feature data to the data set to obtain the fishing feature data set of the corresponding fishing ground.

4. The fishing ground fish management method based on big data analysis according to claim 3, characterized by: In step S3, the recommendation model is constructed based on the machine learning network, and the fishing feature data set of each corresponding fishing ground is input into the recommendation model for iterative training. The recommendation model outputs the recommended fish species and the recommended fishing position information of the corresponding fishing ground at the current time, including the following sub-steps: S31, a recommendation model is constructed based on a machine learning network, the recommendation model includes an input layer, a shared layer, a fish species branch module and a fishing position branch module, the output end of the input layer is connected with the input end of the shared layer, the fishing feature data set of each corresponding fishing ground is input into the recommendation model for iterative training, the output end of the shared layer is connected with the input end of the fish species branch module and the fishing position branch module respectively, and the fishing feature data set of each corresponding fishing ground is extracted, the fish species branch module outputs the fish species release rate, and the fishing position branch module outputs the fishing position fish species utility value; S32, the fish species with the top three release rates is obtained as the recommended fish species, the fishing position fish species utility value corresponding to the recommended fish species is obtained, and the fishing position with the fish species utility value in the excellent fishing position interval is selected as the recommended fishing position; S33, the total fish catch, fishing times and fishing price of the recommended fishing position are obtained to comprehensively calculate the evaluation value Score of each fishing position s , the expression is: Score s = a x T s / max(T) + b x (1 - C s / max(C)) + g x ((1 - (P s - min(P)) / (max(P)-min(P)) In the formula, T s is the total fish catch of the fishing site s, C s is the number of fishing times of the fishing site s, P s is the current fishing price of the fishing site s, T is the total fish catch of the fishing site, C is the maximum number of fishing times of the fishing site, min(P) is the minimum fishing price of the fishing site, max(P) is the maximum fishing price of the fishing site, a is the weight coefficient of the fish catch, β is the weight coefficient of the number of fishing times, and γ is the weight coefficient of the fishing price. S34, the evaluation value of each recommended fishing position is associated with the fishing position number and the recommended fish species in descending order to obtain the recommended fishing list information for the user to view.

5. The big data analysis based fishing ground fish return management method according to claim 4, characterized by: After step S34, it is further judged whether the recommended fish species of the fishing ground at the current time is consistent with the fishing target fish species selected by the user. If yes, the idle fishing position in the recommended fishing list information is obtained according to the fishing position state information, and the fishing position with the highest evaluation value in the idle fishing position is selected as the recommended fishing position and distributed to the user. If not, it is judged whether the fishing target fish species is replaced by the recommended fish species. If yes, the fishing position with the highest evaluation value in the idle fishing position is obtained as the recommended fishing position and distributed to the user. Otherwise, the utility value corresponding to the fishing target fish species of each fishing position in the time window to the current time is obtained, and the fishing position with the fish species utility value in the excellent fishing position interval is selected as the candidate fishing position. The idle fishing position in the candidate fishing position is obtained according to the fishing position state information, and the evaluation value of the idle fishing position in each candidate fishing position is calculated. The fishing position with the highest evaluation value is selected as the recommended fishing position and distributed to the user.

6. The big data analysis based fishing ground fish returning management method of claim 5, characterized by: The shared layer includes a first feature extraction unit, a second feature extraction unit and a third feature extraction unit, the first feature extraction unit, the second feature extraction unit and the third feature extraction unit are sequentially connected, and the third feature extraction unit outputs shared feature data, wherein, The expression of the first feature extraction unit is: h1=σ(W1X+b1) In the formula, X is the standard feature data corresponding to each fishing ground, σ is a Sigmoid function, W1 is a weight matrix, the dimension of which is R 256×64 , and b1 is a bias vector. The expression of the second feature extraction unit is: h2=LayerNorm(W2h1+b2) In the formula, h1 is the output feature of the first feature extraction unit, W2 is a weight matrix with a dimension of R 128×256 , b2 is a bias vector, and LayerNorm() is a layer normalization function. The expression of the third feature extraction unit is: h3=Dropout(σ(W3h2+b3), c=0.3) In the formula, h2 is the output feature of the second feature extraction unit, W3 is a weight matrix with a dimension of R 64×128 , b3 is a bias vector, Dropout() is a regularization function, and c is a neuron dropout probability.

7. The big data analysis based fishing ground fish returning management method according to claim 6, characterized by: The fish species branch module includes a first full connection layer, a second full connection layer and a first output layer, wherein the input of the first full connection layer is connected with the output of the third feature extraction unit, the 64-dimensional feature is reduced to 32-dimensional, and the expression is: z1=W4h3+b4 where W4 is a weight matrix of dimension R 32×64 , and b4 is a bias vector of dimension R 32 ; The output of the first full connection layer is connected with the input of the second full connection layer, 32-dimensional features are mapped to m-dimensional species quantity, and the expression is: z2 = RELU(z1) W5 + b5 where RELU() is a ReLU function, W5 is a weight matrix with dimension R m×32 , and b5 is a bias vector with dimension R m . The output of the second full connection layer is connected with the first output layer, and the first output layer is used for outputting the probability of each species, and the expression is: where p i is the predicted probability of the i-th fish species, and the sum of the predicted probabilities of all fish species is 1.

8. The big data analysis based fishing ground fish return management method according to claim 7, characterized by: The fishing position branch module comprises a third full connection layer, a fourth full connection layer and a second output layer, wherein, The input of the third full connection layer is connected with the output of the third feature extraction unit, 64-dimensional features are reduced to 16-dimensional features, and the expression is: u1 = W6h3 + b6 where W6 is a weight matrix of dimension R 16×64 , and b6 is a bias vector of dimension R 16 ; The output of the third full connection layer is connected with the input of the fourth full connection layer, 16-dimensional features are mapped to 1-dimensional scalar, and the expression is: y = sigma(u1) W7 + b7 where σ is a sigmoid function, W7is a weight matrix of dimension R 1×16 , and b7is a bias vector of dimension R. The output of the fourth full connection layer is connected with the input of the second output layer, and the second output layer outputs the fishing position utility value, y [0, 1].

9. A fishing ground fish returning management system based on big data analysis, which is implemented by using the fishing ground fish returning management method based on big data analysis according to any one of claims 1-8. The system comprises: An information input module is used for a user to select a fishing ground and a target fish species, obtain a fishing card and input initial user fishing information; A data processing module is used for obtaining historical fishing information of a user in a fishing ground, and performing data preprocessing to obtain fishing feature data sets corresponding to the fishing ground; A recommendation module is used for constructing a recommendation model based on a machine learning network, inputting the fishing feature data sets corresponding to each fishing ground into the recommendation model for iterative training, and outputting recommended fish species and recommended fishing position information of the corresponding fishing ground at the current time; A settlement module is used for a user to select a fishing position in the recommended fishing position information for fishing, weigh fish catch according to fish species, obtain fish total price calculated by fish species unit price, and remotely settle accounts of a user account by calling a payment interface.

Images