Intelligent fishing method, device and system based on machine learning

Abstract

The present application discloses an intelligent fishing method, device and system based on machine learning, the method comprising: collecting environmental data and user behavior data, uploading environmental data and user behavior, analyzing environmental data and user behavior data in a distributed system, and outputting the probability of a user catching fish and the category of the user's fishing behavior based on the analysis results. Through this application, a variety of environmental factors can be analyzed, covering climate and environmental data, location area data and information data obtained on the Internet, especially a specific regional location can be identified, and then the unique geographical climate environment of this regional location can be analyzed, and on this basis, real-time interaction with the user is formed, the user's fishing process is monitored in real time, and various situations in the fishing process are returned to the user in the form of a report, so that the user can predict his or her fishing results and even fishing operations during the fishing process, which greatly improves the user's fishing experience.

Description

Technical Field

[0001] The present application relates to the field of application technology of outdoor sports, and specifically to an intelligent fishing method, device and system based on machine learning. Background Art

[0002] Fishing is an activity that many people enjoy, and anglers often go to different places to fish. When going to unfamiliar places to fish, they are not familiar with the local weather and fishing environment. Sometimes they return empty-handed due to unfamiliarity with the environment. In addition, for some users who have just started fishing, they are not sure whether there are problems in the fishing process. Often after trying many times, they have no gains and cannot get a good fishing experience. Currently, there is no device on the market that can simultaneously detect the surrounding environment and the user's own behavior during fishing to form a good experience with the user.

[0003] Common fishing devices usually use simple floats and fishing rods, and rely entirely on the angler's own experience, or use sensors to prompt the occurrence of fish bites. They are unable to judge environmental factors, cannot predict the probability of catching fish at that time and place, and cannot enable users to judge their own fishing behavior during the user's fishing behavior, thereby causing a bad fishing experience for novice anglers. Summary of the invention

[0004] The present application provides an intelligent fishing method, device and system based on machine learning.

[0005] The embodiment of the present application provides an intelligent fishing method based on machine learning, including the steps of collecting environmental data and user behavior data, uploading data, analyzing environmental data, analyzing user behavior data after analyzing environmental data when the result meets a threshold, and returning an analysis report.

[0006] An embodiment of the present application provides an intelligent fishing device based on machine learning, including: an environmental data acquisition module, used to collect environmental data, including climate data, location area, and information searched on the Internet with fishing location as a keyword; a user behavior data acquisition module, used to collect user behavior data including the user's rod lifting speed, rod lifting acceleration, rod lifting frequency, etc.; a data transmission module, used to transmit the collected environmental data and user behavior data to a cloud analysis platform; an environmental data analysis module, used to analyze and judge the uploaded environmental data; a user behavior data analysis module, used to analyze the uploaded user behavior data; a data return module, used to transmit the analyzed results back to the user terminal.

[0007] An embodiment of the present application provides a system, which includes: a distributed server cluster, a user terminal, a sensor, a processor, a memory, an input device and an output device, an intelligent float and an intelligent fishing rod, and an intelligent fishing device based on machine learning; wherein the distributed server cluster includes multiple nodes; the intelligent float and the intelligent fishing rod can both collect data information such as speed and acceleration; the processor includes one or more processors; the memory is used to store one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the intelligent fishing method based on machine learning in the embodiment of the present application.

[0008] The embodiments of the present application provide a method, device and system for intelligent fishing based on machine learning, which predicts the probability of fishing success in the next cycle by collecting a variety of environmental data and user behavior data, and then classifies the user's fishing behavior based on the prediction results. In the present application, a variety of environmental factors can be analyzed, covering climate and environmental data, location area data and information data obtained on the Internet, especially a specific regional location can be identified, and then the unique geographical climate environment of this regional location can be analyzed, and on this basis, real-time interaction with the user is formed, the user's fishing process is monitored in real time, and various situations in the fishing process are returned to the user in the form of a report, so that the user can predict his fishing results and even fishing operations during the fishing process, which greatly improves the user's fishing experience.

[0009] With regard to the above embodiments and other aspects of the present application and their implementation, further description is provided in the accompanying drawings, detailed description and claims. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 This is a flow chart of an intelligent fishing method based on machine learning in an embodiment of the present application; DETAILED DESCRIPTION

[0011] In order to make the purpose, technical solution and advantages of the present application more clear, the embodiments of the present application will be described in detail below with reference to the accompanying drawings. It should be noted that the embodiments and features in the embodiments of the present application can be combined with each other arbitrarily without conflict.

[0012] Embodiment 1

[0013] Figure 1 The flowchart of the intelligent fishing method based on machine learning in the embodiment of the present application is as follows. The embodiment can be applied to the case of analyzing environmental factors and user behaviors. The method can be executed by the intelligent fishing device based on machine learning in the embodiment of the present application. The device can be implemented by software and / or hardware and can generally be integrated in an intelligent system. The method of the embodiment of the present application includes:

[0014] Step 101: Collect environmental data and user behavior data.

[0015] Specifically, with time T1 as a period, environmental data is collected based on sensors on the float and the fishing rod; the environmental data includes natural data, location area, and information searched on the Internet with fishing location as a keyword, and the environmental data is used as the first data set; with time T2 as a period, the user behavior data is collected as the second data set, and the user behavior data includes the user's rod lifting speed, rod lifting acceleration, rod lifting frequency, etc.; the natural data includes water flow speed, water flow acceleration, wind speed, etc.

[0016] Further optionally, T1 and T2 can be set to the same value or different values. The value of T1 can be set to change according to the change of climate. When the climate changes rapidly, the value of T1 is reduced, and when the climate is stable, the value of T1 is increased.

[0017] Step 102: Upload data.

[0018] Specifically, the data of the first data set is returned to a cloud analysis platform, the cloud analysis platform is composed of a distributed system, there are multiple nodes in the distributed system, the multiple nodes correspond to different location areas, and each node processes the data uploaded by its corresponding location area;

[0019] Preferably, the computing resources of each node in the distributed system are dynamically allocated according to the amount of data uploaded in each location area. The distributed system obtains the fishing data of multiple users in the corresponding location area in real time. If there are more users uploading data in a certain location area, more computing resources are allocated to the corresponding nodes. If there are fewer users uploading data in a certain location area, fewer computing resources are allocated to the corresponding nodes, so as to flexibly configure the computing resources of the distributed system.

[0020] Step 103: Analyze environmental data.

[0021] Specifically, the cloud analysis platform obtains the location area in the first data set, sends the data in the first data set to the node corresponding to the location area, and inputs the data into the first training model in the node, and the output of the first training model is the probability value of catching fish in the next T1 period;

[0022] Further optionally, the first training model is a CNN model, and the first historical data set is obtained as a training set with T1 as a cycle, and the training set data is input into the CNN model to obtain the recognition result, and the model recognition result is compared with the correct result. The parameters in the model structure are corrected, and the training steps are repeated multiple times to complete the training and obtain the trained model. The CNN model is composed of three convolutional layers, three pooling layers and two fully connected layers; wherein the convolutional layer extracts features on the original input through the weight sharing and local connection characteristics of the convolutional neural network, and the pooling layer reduces the data dimension by merging the output of a layer of neuron clusters into a single neuron in the next layer; the fully connected layer connects the neurons of each layer to all neurons in the previous layer to integrate the global information of the entire input; the convolutional layer and the pooling layer are alternately connected, and a convolutional layer is connected to a pooling layer, and the number of units in the two fully connected layers are 128 and 32 respectively. The function of the convolutional layer is mainly feature extraction. Each convolutional layer performs a convolution operation on the feature map of the input layer through multiple convolution kernels to output a new feature map. Use It represents the j-th feature map output by the i-th layer, as shown in the following formula.

[0023]

[0024] Specifically, in the above formula, f represents a nonlinear activation function, is the weight matrix of the jth convolution kernel in the i-th layer, Represents the convolution operation, that is, the weight matrix Slide the feature map m in a certain direction and perform the bitwise sum operation on the current area. Represents the bias term.

[0025] Specifically, if the output probability value is higher than a preset threshold, the probability value is returned to the bound user terminal; and the collected data of the second data set is input into the second training model in the node;

[0026] Further optionally, if the probability value is lower than a preset threshold, a text or voice prompt is output to the terminal, such as "Sorry, the probability of you catching fish after today is low, do you want to continue?"

[0027] Step 104: Analyze user behavior data.

[0028] Specifically, in the training data set of the second training model, the data is classified according to preset user fishing behaviors to obtain different classification categories.

[0029] Further optionally, the classification categories include normal fishing behavior, abnormal fishing behavior - lifting the rod too early, abnormal fishing behavior - lifting the rod too late, abnormal fishing behavior - lifting the rod too lightly, etc.

[0030] Specifically, the user's fishing behavior is identified for the data input into the second data set. The second training model is based on the LSTM model and includes three LSTM branches. The historical second data set obtained with T2 as the period is used as the training set for training. The time series features of different time spans in the training set are input into different LSTM branches. The output results of the three LSTM branches are spliced ​​and fused, and then input into the fully connected layer for feature combination. Finally, the output of the fully connected layer is input into the sigmoid layer for final classification output to obtain the classification of user behavior. Among them, the basic unit of LSTM includes output gate, input gate, forget gate and cell state, among which the cell state can save long-term historical information to ensure the effective flow of information during model training, and the gate mechanism can limit the amount of information. LSTM uses forget gate and input gate to control the content of state unit C t , forget gate f t Determine the cell state C at the previous moment t-1 How many cell states C can be retained to the current moment? t , the forget gate can retain information from a long time ago. Input gate i t Determines how many inputs x there are at the current moment t Retain to unit state C t , which can not only ensure that valid content enters the memory, but also avoid the irrelevant content from being saved again. The content of the unit state C t From the last unit state C t-1 and the current input cell state and input gate i t And the forget gate f t The input gate is used to control the cell state C t How many outputs are there to the LSTM model? The current output value h t , which is used to control the impact of long-term memory on the current output, and finally obtain the output o(t). The final output value of LSTM h t By the current state unit C t With the current output gate value o t Joint decision. Forget gate f t , input gate i t , unit status C t , output gate o t The calculation formula of the hidden layer h(t) is as follows:

[0031] f t =σ(W f ·[h t-1 , x t ]+b f )

[0032] i t=σ(W i ·[h t-1 ,x t ]+b i )

[0033]

[0034] o(t)=σ(W o ·[h t-1 , x t ]+b o )

[0035] h t =o t .tanhC t

[0036] Among them, W f , W i , W o Indicates the input x at the current time t The weight of f 、b i 、b o is the bias term.

[0037] Step 105: Return the result.

[0038] Preferably, the fishing behavior category output by the second training model is sent to the user terminal interface in the form of a report, which is in the form of any one of text, voice, animation, or a combination of the above methods, and prompts the user to correct the fishing behavior, such as "the reason why your fishing success rate was not high just now was that you lifted the rod too quickly, you can be a little more patient";

[0039] Step 106: Repeat the above steps to continuously monitor and report the user's phishing behavior.

[0040] In this embodiment, by analyzing environmental data, the probability of successful fishing within a period can be predicted and provided to users so that they can decide whether to adjust their fishing itinerary accordingly; by analyzing user behavior, user behavior can be judged and corrected, and real-time interaction can be carried out with users during the fishing process, thereby enhancing the fun of fishing and obtaining a better fishing experience.

[0041] Embodiment 2

[0042] The intelligent fishing device based on machine learning provided in the embodiment of the present application can execute the operation method provided in any embodiment of the present application, and has the corresponding modules and beneficial effects of the execution method. The device can be implemented by software and / or hardware, specifically including: an environmental data acquisition module, a user behavior data acquisition module, a data transmission module, an environmental data analysis module, a user behavior data analysis module, and a data return module.

[0043] Among them, the environmental data acquisition module is used to collect environmental data, including climate data, location area, and information searched on the Internet using fishing location as a keyword.

[0044] The user behavior data acquisition module is used to obtain user behavior data, specifically the user's rod lifting speed, rod lifting acceleration, rod lifting frequency, etc.

[0045] The data transmission module is used to return the environmental data and user behavior data to the cloud analysis platform.

[0046] The environmental data analysis module is used to analyze the uploaded environmental data.

[0047] The user behavior data analysis module is used to analyze the uploaded user behavior data.

[0048] The data return module is used to return the analysis results of the environment data analysis module and the user behavior data analysis module in the form of a report.

[0049] In one embodiment, the environmental data acquisition module includes:

[0050] The climate and environment data acquisition unit is used to collect climate and environment data, including water flow velocity, water flow acceleration, wind speed, etc.

[0051] A location area data acquisition unit, used to obtain the location area where the user is located;

[0052] Other information acquisition units are used to acquire information searched on the Internet using the fishing location as a keyword.

[0053] In one embodiment, the user behavior data acquisition module includes:

[0054] The fishing rod data collection unit specifically collects the user's rod lifting speed and rod lifting acceleration.

[0055] The float data collection unit specifically collects the frequency of lifting the rod.

[0056] In one embodiment, the environmental data analysis module includes:

[0057] The probability analysis unit is used to analyze the probability of catching fish in the next T1 period.

[0058] A probability judgment unit, used for deciding, according to a judgment result, whether to return a result to a user terminal or to input the collected data of the second data set into a second training model in the node;

[0059] In one embodiment, the user behavior data analysis module includes:

[0060] The user behavior identification unit is used to identify the user's fishing behavior.

[0061] The user behavior category output unit outputs the phishing behavior category.

[0062] Embodiment 3

[0063] An embodiment of the present application provides an intelligent fishing system based on machine learning, which includes a sensor, an input device, a memory, a processor, an intelligent fishing device based on machine learning, an output device, and a user terminal; the above components can be connected via a bus or other means.

[0064] In one embodiment, the sensor is used as a sensing medium to obtain parameters collected by a fishing rod and a float, and the acquisition method is not limited herein.

[0065] In one implementation, the input device is responsible for input processing of data acquired by the sensor.

[0066] In one embodiment, the memory, as a computer-readable storage medium, can be used to store software programs, computer executable programs and modules, such as the modules corresponding to the intelligent fishing device based on machine learning in the embodiment of the present application (environmental data acquisition module, user behavior data acquisition module, data transmission module, environmental data analysis module, user behavior data analysis module, data return module). The processor executes various functional applications and data processing of the device by running the software programs, instructions and modules stored in the memory, that is, realizing the above-mentioned intelligent float anti-interference method.

[0067] In one embodiment, the memory may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system and an application required for at least one function; the data storage area may store data created according to the use of the terminal, etc. In addition, the memory may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one cache memory device, a flash memory device, or other non-volatile solid-state memory devices.

[0068] In one embodiment, the number of processors may be one or more;

[0069] In one embodiment, the output device is responsible for outputting data;

[0070] In one implementation, the user terminal is responsible for receiving the returned data;

[0071] Exemplarily, the device in the embodiment of the present application may include a processing device (such as a central processing unit, etc.), a read-only memory, an input / output (I / O) interface, etc., wherein the device in the embodiment of the present application also includes the following modules: an environmental data acquisition module, a user behavior data acquisition module, a data transmission module, an environmental data analysis module, a user behavior data analysis module, and a data return module. The device in the embodiment of the present application can implement an intelligent fishing method based on machine learning, and the specific steps can be as follows:

[0072] (1) Collecting environmental data and user behavior data:

[0073] With time T1 as a period, environmental data is collected based on sensors on the float and the fishing rod; the environmental data includes natural data, location area, and information searched on the Internet with fishing location as a keyword, and the environmental data is used as the first data set; with time T2 as a period, the user behavior data is collected as the second data set, and the user behavior data includes the user's rod lifting speed, rod lifting acceleration, rod lifting frequency, etc.; the natural data includes water flow speed, water flow acceleration, wind speed, etc.

[0074] T1 and T2 can be set to the same value or different values. The value of T1 can be set to change according to the change of climate. When the climate changes rapidly, the value of T1 can be reduced, and when the climate is stable, the value of T1 can be increased.

[0075] (2) Upload data:

[0076] Returning the data of the first data set to a cloud analysis platform, wherein the cloud analysis platform is composed of a distributed system, wherein there are multiple nodes in the distributed system, wherein the multiple nodes correspond to different location areas, and each node processes the data uploaded by the corresponding location area;

[0077] The computing resources of each node in the distributed system are dynamically allocated according to the amount of data uploaded in each location area. The distributed system obtains the fishing data of multiple users in the corresponding location area in real time. If there are more users uploading data in a certain location area, more computing resources are allocated to the corresponding nodes. If there are fewer users uploading data in a certain location area, fewer computing resources are allocated to the corresponding nodes, so as to flexibly configure the computing resources of the distributed system.

[0078] (3) Analyze environmental data:

[0079] The cloud analysis platform obtains the location area in the first data set, sends the data in the first data set to the node corresponding to the location area, and inputs the data into the first training model in the node, and the output of the first training model is the probability value of catching fish in the next T1 period;

[0080] The first training model is a CNN model. The first historical data set is obtained as a training set with T1 as a cycle. The training set data is input into the CNN model to obtain the recognition result. The model recognition result is compared with the correct result. The parameters in the model structure are corrected. The training steps are repeated multiple times to complete the training and obtain the trained model. The CNN model consists of three convolutional layers, three pooling layers and two fully connected layers. Among them, the convolutional layer extracts features on the original input through the weight sharing and local connection characteristics of the convolutional neural network. The pooling layer reduces the data dimension by merging the output of a layer of neuron clusters into a single neuron in the next layer. The fully connected layer connects the neurons of each layer to all neurons in the previous layer to integrate the global information of the entire input. The convolutional layer and the pooling layer are connected alternately. A pooling layer is connected after a convolutional layer. The number of units in the two fully connected layers are 128 and 32 respectively. The function of the convolutional layer is mainly feature extraction. Each convolutional layer performs a convolution operation on the feature map of the input layer through multiple convolution kernels to output a new feature map. Use It represents the j-th feature map output by the i-th layer, as shown in the following formula.

[0081]

[0082] In the above formula, f represents the nonlinear activation function, is the weight matrix of the jth convolution kernel in the i-th layer, Represents the convolution operation, that is, the weight matrix Slide the feature map m in a certain direction and perform the bitwise sum operation on the current area. Represents the bias term.

[0083] If the output probability value is higher than the preset threshold, the probability value is returned to the bound user terminal; and the collected data of the second data set is input into the second training model in the node;

[0084] If the probability value is lower than a preset threshold, a text or voice prompt is output to the terminal, such as "Sorry, the probability of you catching fish after today is low, do you want to continue?".

[0085] (4) Analyze user behavior data:

[0086] In the training data set of the second training model, the data is classified according to the preset user fishing behavior to obtain different classification categories.

[0087] The classification categories include normal fishing behavior, abnormal fishing behavior - lifting the rod too early, abnormal fishing behavior - lifting the rod too late, abnormal fishing behavior - lifting the rod too lightly, etc.

[0088] According to the data of the second data set input, the user's fishing behavior is identified. The second training model is based on the LSTM model and includes three LSTM branches. The historical second data set obtained with T1 as the period is used as the training set for training. The time series features of different time spans in the training set are input into different LSTM branches. The output results of the three LSTM branches are spliced ​​and fused, and then input into the fully connected layer for feature combination. Finally, the output of the fully connected layer is input into the sigmoid layer for final classification output to obtain the classification of user behavior. Among them, the basic unit of LSTM includes output gate, input gate, forget gate and cell state. The cell state can save long-term historical information to ensure the effective flow of information during model training. The gate mechanism can limit the amount of information. LSTM uses forget gate and input gate to control the content of state unit C t , forget gate f t Determine the cell state C at the previous moment t-1 How many cell states C can be retained to the current moment? t , the forget gate can retain information from a long time ago. Input gate i t Determines how many inputs x there are at the current moment t Retain to unit state C t , which can not only ensure that valid content enters the memory, but also avoid the irrelevant content from being saved again. The content of the unit state C t From the last unit state C t-1 and the current input cell state and input gate i t And the forget gate f t The input gate is used to control the cell state C t How many outputs are there to the LSTM model? The current output value h t , which is used to control the impact of long-term memory on the current output, and finally obtain the output o(t). The final output value of LSTM h t By the current state unit C t With the current output gate value o t Joint decision. Forget gate f t , input gate i t , unit status C t , output gate o t The calculation formula of the hidden layer h(t) is as follows:

[0089] f t =σ(W f ·[h t-1 , x t ]+b f )

[0090] i t =σ(Wi ·[h t-1 ,x t ]+b i )

[0091]

[0092] o(t)=σ(W o ·[h t-1 , x t ]+b o )

[0093] h t =o t tanhC t

[0094] Among them, W f , W i , W o Indicates the input x at the current time t The weight of f 、b i 、b o is the bias term.

[0095] (5) Return result:

[0096] The fishing behavior category output by the second training model is sent to the user terminal interface in the form of a report, wherein the report is in the form of any one of text, voice, animation, or a combination of the above methods, and the user is prompted to correct the fishing behavior, such as "the reason why your fishing success rate was not high just now was that you lifted the rod too quickly, you can be a little more patient";

[0097] (6) Repeat the above steps to continuously monitor and report back the user's phishing behavior.

[0098] Through the above description of the implementation method, the technicians in the relevant field can clearly understand that the present application can be implemented with the help of software and necessary general hardware, and of course it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as a processor, a read-only memory (ROM), a random access memory (RAM), a flash memory (FLASH), etc. to execute the methods described in the embodiments of the present application.

[0099] It is worth noting that in the embodiment of the above-mentioned intelligent float anti-interference operation device, the various units and modules included are only divided according to functional logic, but are not limited to the above-mentioned division, as long as the corresponding functions can be achieved; in addition, the specific name of each functional unit is only for the convenience of distinguishing each other, and is not used to limit the scope of protection of this application.

[0100] The above description is merely an exemplary embodiment of the present application and is not intended to limit the protection scope of the present application.

[0101] In general, various embodiments of the present application can be implemented in hardware or dedicated circuits, software, logic or any combination thereof. For example, some aspects can be implemented in hardware, while other aspects can be implemented in firmware or software that can be executed by a controller, microprocessor or other computing device, although the present application is not limited thereto.

[0102] Embodiments of the present application may be implemented by executing computer program instructions by a data processor of a device, for example in a processor entity, or by hardware, or by a combination of software and hardware. The computer program instructions may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages.

[0103] The block diagram of any logic flow in the accompanying drawings of the present application can represent program steps, or can represent interconnected logic circuits, modules and functions, or can represent a combination of program steps and logic circuits, modules and functions. The computer program can be stored in a memory. The memory can have any type suitable for the local technical environment and can be implemented using any suitable data storage technology, such as but not limited to read-only memory (ROM), random access memory (RAM), optical memory device and system, etc. Computer-readable media can include non-transient storage media. The data processor can be any type suitable for the local technical environment, such as but not limited to a general-purpose computer, a special-purpose computer, a microprocessor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a programmable logic device (FGPA) and a processor based on a multi-core processor architecture.

[0104] By way of exemplary and non-limiting examples, a detailed description of exemplary embodiments of the present application has been provided above. However, various modifications and adjustments to the above embodiments will be apparent to those skilled in the art when considered in conjunction with the accompanying drawings and claims, but will not depart from the scope of the present invention. Therefore, the proper scope of the present invention will be determined according to the claims.

Claims

1. An intelligent fishing method based on machine learning, comprising the following steps: Step S1: Collecting environmental data and user behavior data, with time T1 as a period, collecting environmental data based on sensors on the float and the fishing rod, the environmental data including natural data, location area and information searched on the Internet with fishing location as a keyword, and taking the environmental data as the first data set; Taking time T2 as a period, collecting the user behavior data as a second data set, wherein the user behavior data includes the user's rod lifting speed, rod lifting acceleration, and rod lifting frequency; The natural data include water velocity, water acceleration, and wind speed; Step S2: Upload data and return the data of the first data set to a cloud analysis platform, wherein the cloud analysis platform is composed of a distributed system, wherein there are multiple nodes in the distributed system, and the multiple nodes correspond to different location areas, and each node processes the data uploaded by its corresponding location area; Step S3: Analyze environmental data. The cloud analysis platform obtains the location area in the first data set, sends the data in the first data set to the node corresponding to the location area, and inputs the data into the first training model in the node. The output of the first training model is the probability value of catching fish in the next T1 period. If the output probability value is higher than a preset threshold, the probability value is returned to the bound user terminal, and the collected data of the second data set is input into the second training model in the node. Step S4: in the training data set of the second training model, the data is classified according to the preset user fishing behavior to obtain different classification categories; for the data input into the second training model, the user fishing behavior is identified and classified; Step S5: sending the fishing behavior category output by the second training model to the user terminal interface in the form of a report, and prompting the user to correct the fishing behavior; Step S6: Repeat the process of S1-S5 to continuously monitor and report back the user's phishing behavior.

2. According to the method described in claim 1, in step S2, for each node in the distributed system, the computing resources of each node in the distributed system are dynamically allocated according to the amount of data uploaded in each location area.

3. According to the method of claim 1, in step S3, the first training model is a CNN model, a historical first data set is obtained as a training set for training with a period of T1, the training set data is input into the CNN model to obtain a recognition result, the model recognition result is compared with the correct result, the parameters in the model structure are corrected, the training steps are repeated multiple times, the training is completed, and a trained model is obtained; the CNN model is composed of three convolutional layers, three pooling layers and two fully connected layers; wherein, The convolution layer extracts features from the original input through the weight sharing and local connection characteristics of the convolutional neural network. The pooling layer reduces the data dimension by merging the output of a layer of neuron clusters into a single neuron in the next layer. The fully connected layer connects the neurons of each layer to all neurons in the previous layer to integrate the global information of the entire input. The convolution layer and the pooling layer are connected alternately, and a convolution layer is connected to a pooling layer. The number of units in the two fully connected layers is 128 and 32 respectively. The function of the convolution layer is mainly feature extraction. Each convolution layer performs a convolution operation on the feature map of the input layer through multiple convolution kernels and outputs a new feature map. Represents the jth feature map output by the i-th layer, as shown in the following formula: In the above formula, f represents the nonlinear activation function, is the weight matrix of the jth convolution kernel in the i-th layer, Represents the convolution operation, that is, the weight matrix Slide the feature map m in a certain direction and perform the bitwise sum operation on the current area. Represents the bias term.

4. The method according to claim 1, in step S3, if the probability value is lower than a preset threshold, a text or voice prompt is output to the terminal.

5. According to the method of claim 1, in the step S4, in the training data set of the second training model, the data is classified according to the preset user fishing behavior to obtain different classification categories; the classification categories include normal fishing behavior, abnormal fishing behavior-lifting the rod too early, abnormal fishing behavior-lifting the rod too late, abnormal fishing behavior-lifting the rod too lightly; the user's fishing behavior is identified based on the data input into the second data set, the second training model is based on the LSTM model, including three LSTM branches, and the historical second data set obtained with T1 as the period is used as the training set for training, and the time series features of different time spans in the training set are input into different LSTM branches, and the output results of the three LSTM branches are spliced ​​and fused, and then input into the fully connected layer for feature combination, and finally the output of the fully connected layer is input into the sigmoid layer for final classification output to obtain the classification of user behavior; wherein, The basic unit of LSTM includes output gate, input gate, forget gate and cell state. The cell state stores long-term historical information to ensure the effective flow of information during model training. The gate mechanism limits the amount of information. LSTM uses forget gate and input gate to control the cell state C. t , forget gate f t Determine the cell state C at the previous moment t-1 How many cell states C can be retained to the current moment? t , the forget gate retains information from a long time ago; the input gate i t Determines how many inputs x there are at the current moment t Retain to unit state C t , which not only ensures that valid content is entered into memory, but also avoids irrelevant content from being saved in memory again. The content of the unit state C t From the last unit state C t-1 and the current input cell state and input gate i t And the forget gate f t Joint decision; output gate o t It is used to control the unit state C t How many outputs are there to the LSTM model? The current output value h t , used to control the impact of long-term memory on the current output; LSTM current output value h t From the current cell state C d With the current output gate value o t Joint decision; forget gate t , input gate i t , unit status C t , output gate o t and the current output value h t The calculation formula is as follows: f t =σ(W f ·[h t-1 ,x t ]+b f ) i t =σ(W i ·[h t-1 ,x t ]+b i ) the t =σ(W o ·[h t-1 ,x t ]+b o ) h t =o t .tanh C t Among them, W f , W i , W o Respectively represent the forget gate, input gate, and output gate for the current input x t The weight of f 、b i 、b o They are the bias items corresponding to the forget gate, input gate, and output gate respectively.

6. According to the method of claim 1, in the step S5, the report is in the form of any one of text, voice, animation, or a combination of the above.

7. According to the method of claim 1, in the steps S1-S6, the user can choose to stop receiving information sent by the cloud server at any time.

8. A machine learning based intelligent fishing device for executing the method of claims 1-7, comprising: Environmental data acquisition module, user behavior data acquisition module, data transmission module, environmental data analysis module, user behavior data analysis module, data return module; wherein, the environmental data acquisition module is used to collect environmental data, including climate data, location area and information searched on the Internet with fishing location as the keyword; the user behavior data acquisition module is used to collect user behavior data, and the user behavior data includes the user's rod lifting speed, rod lifting acceleration, and rod lifting frequency; the data transmission module is used to return the collected environmental data and user behavior data to the cloud analysis platform; the environmental data analysis module is used to analyze the uploaded environmental data; the user behavior data analysis module is used to analyze the uploaded user behavior data; the data return module is used to transmit the analysis results of the environmental data analysis module and / or the user behavior data analysis module back to the user terminal.

9. An intelligent fishing system based on machine learning, comprising a sensor, a processor, a memory, an input device and an output device, and a machine learning-based intelligent fishing device according to claim 8.

Images