Aquaculture optimization method and system based on aquaculture feedback demand mining

Abstract

The present application relates to the field of artificial intelligence technology, and to a breeding optimization method and system based on the mining of breeding feedback needs. By deeply mining the key knowledge points in the breeding feedback needs, the breeding areas of concern, the knowledge points for setting constraints and the target breeding execution strategies are clarified, providing a clear direction and basis for the formulation of breeding strategies. On this basis, the strategy screening mechanism is used to accurately screen out strategies that are highly matched with the target needs from multiple preset breeding execution strategies, ensuring the effectiveness and feasibility of the strategy. Furthermore, by extracting and deeply analyzing the feature data of the target breeding feedback needs and the screened breeding execution strategies, a finely adjusted target optimized breeding strategy is generated. This strategy is not only highly targeted, but also fully considers the various constraints and potential risks in the breeding process, thereby maximizing the breeding benefits.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and specifically, to a breeding optimization method and system based on breeding feedback demand mining. Background Art

[0002] In the existing breeding industry, the efficiency and effectiveness of breeding management are often restricted by many factors, the most prominent of which is the lack of scientificity and precision in the formulation of breeding strategies. Traditional breeding management mostly relies on the experience and judgment of breeders, lacks systematic and data support, resulting in problems such as waste of resources, high costs, and poor disease prevention and control in the breeding process. With the continuous expansion of breeding scale and increasingly fierce market competition, how to scientifically and efficiently formulate breeding strategies and improve breeding benefits has become an urgent problem to be solved in the breeding industry.

[0003] In order to meet the above challenges, aquaculture management systems based on big data and artificial intelligence technologies have gradually emerged in recent years. These systems provide strong support for the formulation of aquaculture strategies by collecting and analyzing various types of data in the aquaculture process. However, existing aquaculture management systems still have shortcomings in mining aquaculture feedback needs and generating targeted aquaculture strategies. Specifically, they often fail to accurately identify the real needs and constraints of farmers, resulting in the lack of pertinence and feasibility of the generated aquaculture strategies. Summary of the invention

[0004] In view of the above-mentioned problems, in combination with the first aspect of the present application, the embodiment of the present application provides a breeding optimization method based on breeding feedback demand mining, the method comprising:

[0005] The breeding feedback mining model is used to mine key knowledge points of the target breeding feedback demand, and the breeding area of ​​concern, the set constraint knowledge points and the target breeding execution strategy of the target breeding feedback demand are generated; the target breeding feedback demand has feedback demand content based on the breeding area;

[0006] Based on the concerned breeding area, the set constraint knowledge point, the target breeding execution strategy and multiple set breeding execution strategy data corresponding to the multiple set breeding execution strategies, the multiple set breeding execution strategies are screened to generate the target screened breeding execution strategy for the target breeding feedback demand;

[0007] Feature data is extracted for the target breeding feedback demand and the target screening breeding execution strategy to generate target feature data, and the breeding feedback mining model is used to make breeding strategy decisions on the target feature data to generate the target optimized breeding strategy.

[0008] On the other hand, an embodiment of the present application also provides a breeding feedback mining system, including a processor and a machine-readable storage medium, wherein the machine-readable storage medium is connected to the processor, the machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the machine-readable storage medium to implement the above method.

[0009] Based on the above aspects, the embodiments of the present application significantly improve the scientificity and accuracy of the formulation of aquaculture strategies, and effectively solve the problems of inefficiency and cost waste caused by decision-making based on experience in traditional aquaculture management. By deeply exploring the key knowledge points in the aquaculture feedback needs, the aquaculture areas to be concerned about, the knowledge points for setting constraints and the target aquaculture execution strategies are clarified, providing a clear direction and basis for the formulation of aquaculture strategies. On this basis, the strategy screening mechanism is used to accurately screen out strategies that are highly matched with the target needs from multiple preset aquaculture execution strategies, ensuring the effectiveness and feasibility of the strategies. Further, by extracting and deeply analyzing the feature data of the target aquaculture feedback needs and the screened aquaculture execution strategies, a finely adjusted target optimization aquaculture strategy is generated. The strategy is not only highly targeted, but also fully considers the various constraints and potential risks in the aquaculture process, and realizes the maximization of aquaculture benefits. As a result, the aquaculture management level can be significantly improved, the aquaculture cost can be reduced, and the aquaculture efficiency can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 It is a schematic diagram of the execution flow of the breeding optimization method based on breeding feedback demand mining provided in an embodiment of the present application.

[0011] Figure 2 It is a schematic diagram of the hardware architecture of the aquaculture feedback mining system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0012] The present application will be described in detail below with reference to the accompanying drawings. Figure 1 It is a flow chart of a breeding optimization method based on breeding feedback demand mining provided by an embodiment of the present application. The breeding optimization method based on breeding feedback demand mining is introduced in detail below.

[0013] Step S110, using the aquaculture feedback mining model to mine key knowledge points of the target aquaculture feedback demand, generating the aquaculture area of ​​interest, setting constraint knowledge points and target aquaculture execution strategy of the target aquaculture feedback demand. The target aquaculture feedback demand has feedback demand content based on the aquaculture area.

[0014] In detail, the breeding feedback mining model is a model based on machine learning or deep learning, which is specifically used to process and analyze feedback information in the breeding process. The breeding feedback mining model can extract key knowledge points from a large amount of breeding data, identify key factors affecting breeding results, and generate targeted breeding strategy recommendations based on this. For example, in the pig breeding scenario, the breeding feedback mining model may learn which factors are highly correlated with key indicators such as pig growth rate and disease incidence by analyzing historical breeding data, expert experience, disease occurrence records and other information. For example, the breeding feedback mining model may find that factors such as feed formula, breeding environment, and stocking density have a significant impact on growth rate, while factors such as sanitary conditions and vaccination are closely related to disease incidence.

[0015] The target breeding feedback requirements are specific requirements that farmers put forward to the system based on the problems or challenges encountered in the current breeding process. They usually have clear optimization goals, such as increasing growth rate, reducing disease incidence, optimizing feed efficiency, etc., and are often targeted at specific breeding areas. For example, the requirements submitted by farmers to the server may be: "I hope to increase the growth rate of pigs in area G while keeping feed costs within a controllable range." Here, "area G" is the breeding area of ​​concern, "increasing growth rate" is the target breeding execution strategy, and "keeping feed costs controllable" is the setting constraint knowledge point.

[0016] The focus breeding area refers to the breeding area that the farmer clearly pointed out in the feedback needs and needs to be focused on and optimized. This focus breeding area may be the entire farm, a specific pig house or a group of pigs. In the above example, "G area" is the focus breeding area. The server will focus on the specific conditions and needs of this area in the subsequent strategy formulation process.

[0017] The constraint knowledge points mentioned above are constraints that are not explicitly stated but implicit in the requirements when farmers put forward feedback requirements, and have an important impact on strategy formulation. These conditions may involve multiple aspects such as cost, manpower, and environment. In the above example, "keeping feed costs controllable" is a constraint knowledge point. When the server screens and optimizes the farming strategy, it needs to ensure that the selected strategy does not significantly increase the feed cost.

[0018] The target farming execution strategy is a specific action plan generated by the farming feedback mining model based on the target farming feedback needs, which is aimed at solving specific problems or achieving specific optimization goals. These strategies may include adjusting feed formulas, improving the farming environment, optimizing feeding management and other measures. For example, in the above example, "increasing growth rate" is a relatively broad target farming execution strategy. The server may generate specific strategies in the subsequent process, such as "adopting XX brand high-efficiency feed formula" or "increasing the amount of exercise for pigs."

[0019] That is, taking the pig farming scenario as an example, in this embodiment, farmers may propose a series of target farming feedback needs based on specific farming areas, such as increasing the growth rate of pigs in a certain area, reducing the incidence of disease, or optimizing feed utilization efficiency, etc. These target farming feedback needs are all targeted at specific problems in specific farming areas.

[0020] In order to respond to these target breeding feedback needs, the server first uses the breeding feedback mining model to mine key knowledge points for these target breeding feedback needs. The breeding feedback mining model may be a machine learning-based model that can extract useful information from a large amount of breeding data. Through mining, the server can identify the breeding areas of interest in the target breeding feedback needs, such as a specific pig house, a breeding area, or the entire breeding farm.

[0021] At the same time, the farming feedback mining model will also mine the knowledge points of setting constraints. These knowledge points of setting constraints are factors that farmers may not explicitly state when they put forward the target farming feedback requirements, but have an important impact on the formulation of farming strategies. For example, farmers may want to increase the growth rate, but at the same time do not want to increase the feed cost too much, or do not want to use certain specific drugs. These constraints are crucial for formulating effective farming strategies.

[0022] Finally, the farming feedback mining model generates a target farming execution strategy, which is a set of specific action plans for the farming area of ​​concern and the set constraint knowledge points. For example, for the need to increase growth rate, the target farming execution strategy may include adjusting feed formula, increasing exercise or improving the farming environment.

[0023] Step S120, based on the concerned breeding area, the set constraint knowledge points, the target breeding execution strategy and multiple set breeding execution strategy data corresponding to the multiple set breeding execution strategies, the multiple set breeding execution strategies are screened to generate a target screened breeding execution strategy for the target breeding feedback demand.

[0024] In detail, the set farming execution strategy data refers to the farming execution strategies and related data for different farming scenarios and needs that have been stored in the system. The strategy feature vector is a representation method for converting the set farming execution strategy into a numerical form. These strategy feature vectors are generated through embedded representation technology and can reflect the similarities and differences between farming execution strategies in a multidimensional space. For example, assuming that the strategy feature vector corresponding to "Strategy A" is [0.5, 0.8, 0.2], this vector represents the specific position and characteristics of Strategy A in the multidimensional space. The server can evaluate the similarity between strategies by calculating the feature distance between different strategy feature vectors.

[0025] That is, after generating the breeding area of ​​concern, setting constraint knowledge points and target breeding execution strategy, the server needs to select the most suitable breeding execution strategy from multiple set breeding execution strategies. These set breeding execution strategies are existing strategies for different breeding scenarios and needs. They all have corresponding breeding execution strategy data, such as historical implementation results, cost input, required resources, etc.

[0026] The screening process may involve multiple steps. First, the server embeds the farming areas of interest, the set constraint knowledge points, and the target farming execution strategy, and converts them into numerical forms that can be processed by machine learning algorithms. Then, the server calculates the feature distance between these embedding representations and multiple set farming execution strategy data to find the strategy that is most similar to the target farming feedback requirements.

[0027] In this process, the server can consider a variety of factors, such as the historical implementation effect of the strategy, the cost-effectiveness ratio, whether it meets the constraints of the farmers, etc. By comprehensively considering these factors, the server can generate a target screening farming execution strategy for the target farming feedback needs.

[0028] Step S130, extracting characteristic data of the target breeding feedback demand and the target screening breeding execution strategy to generate target characteristic data, using the breeding feedback mining model to make breeding strategy decisions on the target characteristic data to generate the target optimized breeding strategy.

[0029] After generating the target screening aquaculture execution strategy, the server needs to further optimize the target screening aquaculture execution strategy. First, the server will extract feature data of the target aquaculture feedback requirements and the target screening aquaculture execution strategy, and generate target feature data containing key information. These target feature data may include specific environmental parameters of the aquaculture area, specific values ​​of the constraint knowledge points, and specific steps of the target aquaculture execution strategy.

[0030] The server then uses the aquaculture feedback mining model to conduct in-depth analysis of these target feature data. The aquaculture feedback mining model may consider more factors, such as seasonal changes, market fluctuations, disease epidemic trends, etc., to evaluate the effectiveness of the target screening aquaculture execution strategy in actual application.

[0031] Finally, based on the analysis results of the farming feedback mining model, the server generates a target optimization farming strategy, which is a set of finely tuned target optimization farming strategies that take into account more practical factors and are more likely to succeed in practical applications. For example, for the need to increase growth rate, the target optimization farming strategy may not only include adjusting feed formula and increasing exercise, but also adjusting the temperature and humidity of the farming environment according to seasonal changes.

[0032] In a more detailed example, the server needs to define a complete set of feature data dimensions to extract key information from target farming feedback requirements and target screening farming execution strategies. These dimensions may include but are not limited to:

[0033] Characteristics of the breeding environment: such as temperature, humidity, ventilation conditions, etc.

[0034] Physiological characteristics of pigs: such as weight, growth rate, disease incidence, etc.

[0035] Feed and nutritional characteristics: such as feed formula, nutritional ingredients, feeding amount, etc.

[0036] Management operation characteristics: such as stocking density, cleaning frequency, disease prevention and control measures, etc.

[0037] Cost and benefit characteristics: such as feed cost, labor cost, expected profit, etc.

[0038] According to the defined feature data dimensions, the server needs to design corresponding feature extraction algorithms, which may involve natural language processing (NLP), image recognition, time series analysis and other technologies.

[0039] NLP algorithm: used to extract keywords and phrases from the text-based target farming feedback requirements, such as "slow growth rate" and "high disease incidence rate", and map them to the corresponding feature dimensions.

[0040] Image recognition algorithm: used to identify the behavior and status of pigs, such as activity level, eating habits, etc., from the surveillance video of the farm to extract physiological characteristic data.

[0041] Time series analysis algorithms: used to analyze trends and cyclical changes in historical farming data, such as the impact of seasonality on growth rate, to extract environmental and management operation characteristic data.

[0042] Next, the server first parses the text content of the target breeding feedback requirements and uses the NLP algorithm to identify key information points. For example, if the feedback requirement mentions "the growth rate of pigs in area G is inconsistent", the server will extract "area G" as the breeding area of ​​concern and "inconsistent growth rate" as part of the target breeding execution strategy that needs to be optimized.

[0043] The parsed key information points are mapped to the predefined feature data dimensions. In this example, "G area" can be mapped to the "specific area" sub-dimension under the breeding environment feature dimension, and "inconsistent growth rate" is mapped to the "growth rate" sub-dimension under the pig physiological feature dimension and marked as abnormal.

[0044] For quantifiable feature data, the server will try to extract specific values ​​from historical farming data, surveillance videos, and other sources. For example, if the server has historical growth rate data for region G, it will extract this data as a basis for quantification. For information that cannot be directly quantified, such as "cleaning frequency" in management operation characteristics, the server will make a subjective assessment or set a default value based on the description in the feedback requirements.

[0045] The server analyzes the text content of the target farming execution strategy and identifies the specific operational measures and expected results. For example, if the strategy mentions "adopting XX brand efficient feed formula", the server will extract "XX brand feed" and "efficient formula" as key information points.

[0046] Similar to the target breeding feedback requirements, the server maps the parsed key information points to the feature data dimension. In this example, "XX brand feed" can be mapped to the "feed formula" sub-dimension under the feed and nutrition feature dimension, and "efficient formula" is used as a quantitative indicator under this sub-dimension (such as the proportion of increased growth rate).

[0047] The server further extracts specific details from the strategy, such as the amount of feed to be fed, the frequency of feeding, the specific operating steps of disease prevention and control measures, etc. These detailed data are crucial for subsequent strategic decisions.

[0048] Therefore, the server integrates the characteristic data of target farming feedback requirements and target screening farming execution strategies to construct a target characteristic data set, which contains all key information points related to farming strategy decisions and their quantitative representation.

[0049] The server inputs the target feature data set into the aquaculture feedback mining model, which has been trained and optimized through a large amount of historical data and expert experience, and can perform complex decision analysis based on the input feature data.

[0050] In the decision-making analysis process, the aquaculture feedback mining model will consider the interaction and influence of multiple factors. For example, machine learning algorithms (such as random forests, neural networks, etc.) can be used to classify, regress or cluster feature data to evaluate the effects and cost-effectiveness of different strategy combinations.

[0051] The aquaculture feedback mining model first evaluates the effect of the target screening aquaculture execution strategy on improving the target aquaculture feedback needs. For example, the changes in key indicators such as growth rate and disease incidence after the implementation of the strategy can be predicted based on historical data and simulation experiments.

[0052] In addition to effect evaluation, the aquaculture feedback mining model will also conduct a cost-benefit analysis. It considers the cost inputs required for strategy implementation (such as feed costs, labor costs, etc.) and the expected benefits (such as additional benefits from increased growth rates). By comparing the cost-benefit ratios of different strategy combinations, the model can select the optimal strategy. During the decision analysis process, the aquaculture feedback mining model will also consider potential risk factors. For example, certain strategies may increase the risk of disease transmission or have a negative impact on the environment. The model will assess the size and likelihood of these risks and adjust the strategy weights or propose risk prevention measures accordingly.

[0053] After comprehensive analysis and evaluation, the aquaculture feedback mining model will generate a target optimized aquaculture strategy, which not only proposes a specific solution for the target aquaculture feedback needs, but also considers multiple aspects such as cost-effectiveness ratio and potential risk factors. For example, it can include a series of finely tuned operational measures and quantitative indicators (such as feeding amount, cleaning frequency, etc.) to ensure the effectiveness and feasibility of the strategy in practical application.

[0054] After generating the target optimization farming strategy, the server feeds it back to the farmers and collects implementation feedback. The farmers can adjust the strategy or put forward new feedback requirements according to the actual situation. The server iteratively optimizes the farming feedback mining model based on the farmers' feedback and new farming data. For example, the model can be retrained, the feature data dimension can be adjusted, or the decision analysis algorithm can be improved to improve the accuracy and adaptability of the model.

[0055] In order to ensure the continued effectiveness of the target optimization farming strategy, the server will continuously monitor and evaluate the farming process. For example, the Internet of Things technology can be used to collect real-time farming data, and data analysis tools can be used to regularly evaluate and adjust the strategy effect.

[0056] Through this process, the server can effectively respond to farmers' feedback needs based on the breeding areas and generate targeted and optimized breeding strategies, which not only helps to improve breeding efficiency and quality, but also reduces breeding costs and risks, bringing greater economic benefits to farmers.

[0057] Based on the above steps, the embodiment of the present application significantly improves the scientificity and accuracy of the formulation of aquaculture strategies, and effectively solves the problems of inefficiency and cost waste caused by decision-making based on experience alone in traditional aquaculture management. By deeply exploring the key knowledge points in the aquaculture feedback needs, the aquaculture areas of concern, the knowledge points for setting constraints and the target aquaculture execution strategies are clarified, providing a clear direction and basis for the formulation of aquaculture strategies. On this basis, the strategy screening mechanism is used to accurately screen out strategies that are highly matched with the target needs from multiple preset aquaculture execution strategies, ensuring the effectiveness and feasibility of the strategy. Further, by extracting and deeply analyzing the feature data of the target aquaculture feedback needs and the screened aquaculture execution strategies, a finely adjusted target optimization aquaculture strategy is generated. The strategy is not only highly targeted, but also fully considers the various constraints and potential risks in the aquaculture process, and realizes the maximization of aquaculture benefits. As a result, the aquaculture management level can be significantly improved, the aquaculture cost can be reduced, and the aquaculture efficiency can be improved.

[0058] In a possible implementation, the plurality of set aquaculture execution strategy data are a plurality of strategy feature vectors. Based on the concerned aquaculture area, the set constraint knowledge point, the target aquaculture execution strategy and the plurality of set aquaculture execution strategy data corresponding to the plurality of set aquaculture execution strategies, step S120 comprises:

[0059] Step A110, embedding the concerned breeding area, the set constraint knowledge points and the target breeding execution strategy to generate a target knowledge point embedding vector of the target breeding feedback requirement.

[0060] Step A120, performing feature distance calculation on the target knowledge point embedding vector and the multiple strategy feature vectors to generate multiple first feature distances between the target knowledge point embedding vector and the multiple strategy feature vectors.

[0061] Step A130, performing strategy screening on the multiple set farming execution strategies based on the multiple first characteristic distances to generate the target screened farming execution strategy.

[0062] In this example, suppose that a specific area of ​​a large pig farm (such as a pig house numbered A) has recently experienced a slowdown in growth, and the farmer hopes to increase the growth rate of the pigs in this area by optimizing the breeding strategy, while keeping the feed cost within a controllable range. Therefore, the farmer submits a feedback request based on the breeding area (area A) to the server.

[0063] The server first analyzes the feedback needs of the farmers, identifies the breeding area of ​​concern as Area A, sets the constraint knowledge point as "keeping feed costs controllable", and the target breeding execution strategy is a broader concept, namely "increasing growth rate".

[0064] Next, the server uses embedding representation technology to convert this text information into a target knowledge point embedding vector in numerical form. This process may involve natural language processing (NLP) technology, such as word embedding or sentence embedding, to map keywords or phrases such as "Area A", "Keep feed costs controllable" and "Increase growth rate" to points in a high-dimensional space. The relative positions of these points reflect the semantic relationship between them.

[0065] The server has pre-stored multiple set farming execution strategies and their corresponding strategy feature vectors. These strategy feature vectors are obtained by analyzing multiple factors such as historical farming data, expert experience, and strategy implementation effects. Each vector represents the position of a certain farming execution strategy in the multidimensional feature space.

[0066] The server now needs to calculate the feature distance between the target knowledge point embedding vector and these strategy feature vectors. There are many ways to calculate the feature distance, such as Euclidean distance, cosine similarity, etc. In this scenario, the server may use cosine similarity as a metric because it can well reflect the directional difference between vectors, and directional differences often reflect the similarity between strategies better than length differences.

[0067] The server traverses all the strategy feature vectors of the set farming execution strategies, calculates the cosine similarity between them and the target knowledge point embedding vector, and generates multiple first feature distances. The smaller these distance values ​​are, the higher the similarity between the corresponding strategy and the target farming feedback demand is.

[0068] The server sorts the multiple set farming execution strategies according to the calculated first characteristic distance, and selects the top N strategies with the highest similarity as candidate strategies. The specific value of N can be adjusted according to the actual situation, for example, it can be determined according to factors such as the server's computing resources, the complexity of strategy implementation, and the farmer's preferences.

[0069] In the candidate strategies, the server also needs to further consider the impact of setting constraint knowledge points. For example, the server will check whether these candidate strategies can improve growth rate while keeping feed costs under control. This may require the server to analyze the specific content of the candidate strategies, including factors such as feed formula, stocking density, and environmental control, as well as their possible impact on feed costs and growth rate.

[0070] After comprehensive evaluation, the server finally generates a target screening breeding execution strategy for the target breeding feedback needs. This strategy is not only highly matched with the feedback needs of the breeders, but also fully considers the impact of the set constraint knowledge points. It is an effective and feasible breeding execution plan.

[0071] In a specific example, assume that the server stores the following three set farming execution strategies and their strategy feature vectors (expressed in simplified form):

[0072] Strategy 1: Adjust feed formula to increase protein ratio (feature vector: [0.5, 0.8, 0.2])

[0073] Strategy 2: Increase exercise and use an automatic feeding system (feature vector: [0.3, 0.4, 0.9])

[0074] Strategy 3: Improve the breeding environment and increase ventilation and light (feature vector: [0.7, 0.5, 0.5])

[0075] The server calculates the cosine similarity between the target knowledge point embedding vector and the three strategy feature vectors and obtains the following results:

[0076] Similarity with strategy 1: 0.85

[0077] Similarity with strategy 2: 0.70

[0078] Similarity with strategy 3: 0.65

[0079] Since Strategy 1 has the highest similarity, and further analysis shows that increasing the protein ratio in feed can help improve growth rate, while the increase in feed cost is within a controllable range, the server selects Strategy 1 as the target screening breeding execution strategy.

[0080] In another possible implementation, the plurality of set aquaculture execution strategy data are a plurality of strategy tag attributes. Based on the concerned aquaculture area, the set constraint knowledge point, the target aquaculture execution strategy and the plurality of set aquaculture execution strategy data corresponding to the plurality of set aquaculture execution strategies, step S120 may further include:

[0081] Step B110, determining the target optimization knowledge points of the target breeding feedback requirements based on the breeding area of ​​interest, the set constraint knowledge points and the target breeding execution strategy.

[0082] Step B120, determining a target label attribute associated with the target optimization knowledge point from among the multiple strategy label attributes.

[0083] Step B130, screening the plurality of set farming execution strategies based on the target tag attribute to generate the target screened farming execution strategy.

[0084] In this example, let's continue to assume that the pigs in area B of a pig farm have a low feed conversion rate. The farmer hopes to improve the feed conversion rate by adjusting the breeding strategy, and at the same time requires that the implementation of the strategy should not significantly increase labor costs. Therefore, the farmer submits a target breeding feedback requirement based on area B to the server.

[0085] The server first analyzes the farmers’ target breeding feedback needs, identifies the breeding area of ​​interest as area B, sets the constraint knowledge point as “not significantly increasing labor costs”, and the target breeding execution strategy is “improving feed conversion rate”.

[0086] Next, the server combines this information to determine the target optimization knowledge point for the target breeding feedback needs. In this scenario, the target optimization knowledge point is "improving the feed conversion rate of pigs in area B while controlling the increase in labor costs within an acceptable range."

[0087] The server has a database containing multiple set farming execution strategies and their corresponding strategy tag attributes. These strategy tag attributes are a high-level summary of the strategy content and are used to quickly retrieve and match related strategies.

[0088] The server now needs to search these strategy label attributes for target label attributes associated with the target optimization knowledge point. For example, you can search for label attributes containing keywords such as "feed conversion rate", "cost control", and "labor cost".

[0089] Through searching, the server found several target label attributes that are closely related to the target optimization knowledge points, such as "efficient feed formula", "automated feeding system", "environmental optimization to reduce human intervention", etc.

[0090] With the target tag attributes, the server can start filtering multiple set farming execution strategies. It will first traverse the entire strategy database to find all strategies that contain these target tag attributes.

[0091] For example, the server might find the following three policies:

[0092] Strategy A: "Use XX brand's highly efficient feed formula, which historical data shows can improve feed conversion by 5%."

[0093] Strategy B: "Introduce an automated feeding system to reduce the number of manual feedings while precisely controlling the amount of feed."

[0094] Strategy C: "Optimize the ventilation and lighting conditions of the pig house to improve the growth environment of the pigs and indirectly increase feed utilization."

[0095] The server will further analyze the specific content of these strategies to ensure that they not only meet the requirements of the target label attributes, but also can improve the feed conversion rate of pigs in area B without significantly increasing labor costs.

[0096] After comprehensive evaluation, the server selects the set farming execution strategy that best meets the target farming feedback requirements as the target screening farming execution strategy. In this scenario, the server may think that the combination of strategy A and strategy B is the optimal solution: on the one hand, the use of efficient feed formula can directly improve feed conversion rate; on the other hand, the introduction of an automated feeding system can reduce manual intervention, which meets the requirements of controlling labor costs.

[0097] Therefore, the target screening farming execution strategy generated by the server may be: "Adopt XX brand high-efficiency feed formula in area B and introduce an automated feeding system to improve feed conversion rate without significantly increasing labor costs."

[0098] In another possible implementation, the plurality of set aquaculture execution strategy data include a plurality of strategy feature vectors and a plurality of strategy label attributes. Based on the concerned aquaculture area, the set constraint knowledge point, the target aquaculture execution strategy and the plurality of set aquaculture execution strategy data corresponding to the plurality of set aquaculture execution strategies, step S120 may further include:

[0099] Step C110, embedding the concerned breeding area, the set constraint knowledge point and the target breeding execution strategy to generate a target knowledge point embedding vector of the target breeding feedback requirement.

[0100] Step C120, performing feature distance calculation on the target knowledge point embedding vector and the multiple strategy feature vectors to generate multiple first feature distances between the target knowledge point embedding vector and the multiple strategy feature vectors.

[0101] Step C130 , screening the multiple set farming execution strategies based on the multiple first characteristic distances to generate a first screened farming execution strategy.

[0102] Step C140, determining the target optimization knowledge points of the target breeding feedback requirements based on the breeding area of ​​interest, the set constraint knowledge points and the target breeding execution strategy.

[0103] Step C150, determining a target label attribute associated with the target optimization knowledge point from among the multiple strategy label attributes.

[0104] Step C160 , screening the plurality of set farming execution strategies based on the target tag attribute, and generating a second screened farming execution strategy.

[0105] Step C170, performing strategy aggregation on the first screening and breeding execution strategy and the second screening and breeding execution strategy to generate the target screening and breeding execution strategy.

[0106] In this example, let's assume that the pigs in area C of a pig farm have slow growth and increased disease incidence. The farmer hopes to improve growth rate and reduce disease incidence by optimizing the breeding strategy, and at the same time requires that the implementation of the strategy should not significantly increase feed costs and labor input. Therefore, the farmer submits a target breeding feedback requirement based on area C to the server.

[0107] The server first analyzes the farmers’ target breeding feedback needs, identifies the breeding area of ​​concern as Area C, sets the constraint knowledge point as “not significantly increasing feed costs and labor input”, and the target breeding execution strategy is “increasing growth rate” and “reducing disease incidence”.

[0108] The server uses embedding representation technology to convert these text information into numerical target knowledge point embedding vectors. This vector represents the specific needs and constraints of farmers in multidimensional space, providing a basis for subsequent strategy screening.

[0109] The server stores multiple set farming execution strategies and their corresponding strategy feature vectors. These strategy feature vectors are obtained by analyzing historical farming data, expert experience and other factors, and can reflect the position of the strategy in the multidimensional feature space. The server calculates the feature distance (such as cosine similarity) between the target knowledge point embedding vector and each strategy feature vector. The smaller the distance, the higher the similarity between the strategy and the target farming feedback demand.

[0110] The server sorts the multiple set breeding execution strategies according to the calculated first feature distance, and selects the top N strategies with the highest similarity as the first screening breeding execution strategies, which are closest to the needs of the breeders in the feature space.

[0111] The server will focus on breeding areas, set constraint knowledge points and target breeding execution strategies to determine the specific optimization goals of target breeding feedback needs, namely "increasing pig growth rate and reducing disease incidence in area C without significantly increasing feed costs and labor input."

[0112] The server searches the policy database for policy label attributes associated with the target optimization knowledge point, and these label attributes may include "efficient growth promotion", "disease prevention and control", "cost-effectiveness optimization", etc.

[0113] The server screens all stored set farming execution strategies according to the target tag attributes, and finds those strategies that explicitly contain these tag attributes, and these strategies are highly relevant to the target optimization knowledge points in terms of content.

[0114] The server now has two sets of screening results: the first screening breeding execution strategy is based on feature distance calculation, and is the strategy that is closest to the needs of farmers in the feature space; the second screening breeding execution strategy is based on strategy label attribute screening, and is the strategy that is highly matched with the target optimization knowledge point in terms of content.

[0115] The server needs to aggregate these two sets of strategies. The aggregation process can include removing duplicate strategies, evaluating the complementarity between strategies, and considering the complexity and cost of strategy implementation. Finally, the server generates a set of target screening and breeding execution strategies that integrate the results of the first and second screenings.

[0116] For example, the target screening farming implementation strategy may include: "In area C, use XX brand high-efficiency growth-promoting feed formula and combine it with YY disease prevention vaccine program; at the same time, introduce an automated cleaning system to reduce manual intervention and reduce the risk of disease transmission; all strategy implementations must consider the cost-effectiveness ratio to ensure that feed costs and labor inputs are not significantly increased."

[0117] In a possible implementation, the method further includes:

[0118] Step D110, mining key knowledge points of the plurality of set breeding execution strategy data to generate a plurality of set strategy knowledge points corresponding to the plurality of set breeding execution strategy data. The plurality of set strategy knowledge points include at least one of a plurality of set breeding areas, a plurality of breeding constraint knowledge points and a plurality of set breeding execution strategies.

[0119] Step D120, based on the multiple set strategy knowledge points, the knowledge points of the concerned breeding area, the set constraint knowledge points and the target breeding execution strategy are matched to generate a matching breeding area for the concerned breeding area, a matching constraint knowledge point for the set constraint knowledge point and a matching breeding execution strategy for the target breeding execution strategy.

[0120] Step S120 specifically includes: based on the matching breeding area, the matching constraint knowledge point, the matching breeding execution strategy and the multiple set breeding execution strategy data, the multiple set breeding execution strategies are screened to generate the target screened breeding execution strategy.

[0121] In this embodiment, let's continue to assume that a pig farm encounters the problem of inconsistent growth rates of pigs in area D, where some pigs grow quickly while others grow slowly. The farmer hopes to find a strategy that can improve the consistency of the growth rate of pigs in area D without increasing too much cost. The farmer submits this target breeding feedback requirement based on area D to the server.

[0122] The server first mines key knowledge points of the multiple set breeding execution strategy data stored in it. These set breeding execution strategy data may include a large amount of historical breeding records, expert advice, success cases, etc.

[0123] The mining process may involve a variety of technologies such as natural language processing and data mining. The server will identify key information in each strategy, such as breeding areas, breeding constraints (such as cost, manpower, environment, etc.), specific breeding implementation measures, etc., and organize this information into structured set strategy knowledge points.

[0124] For example, a set breeding execution strategy may be mined to obtain the following set strategy knowledge points: the breeding area is "outdoor free-range area", the breeding constraint knowledge point is "low-cost operation", and the set breeding execution strategy is "timed and quantitative feeding + natural foraging".

[0125] The server then matches the concerned farming area (area D), set constraint knowledge points (without increasing excessive costs), and target farming execution strategy (improving growth rate consistency) submitted by the farmer with the multiple set strategy knowledge points mined. The matching process may be based on a variety of algorithms such as similarity calculation and rule matching. The server will look for set strategy knowledge points that are similar to the concerned farming area, meet the set constraint conditions, and are related to the target farming execution strategy.

[0126] For example, the server may find that the "indoor temperature-controlled breeding area" has certain similarities with area D in environmental control, but is not a complete match; "low-cost operation" is completely consistent with the farmer's set constraint knowledge points; and "phased and refined feeding" is highly correlated with the target breeding execution strategy of improving growth rate consistency.

[0127] Based on these matching results, the server generates matching breeding areas for the breeding areas of interest (which may be one or more similar areas), matching constraint knowledge points for setting constraint knowledge points (i.e., “low-cost operation”), and matching breeding execution strategies for the target breeding execution strategies (such as “phased refined feeding” and its related strategies).

[0128] With the matching results, the server can more accurately screen multiple set breeding execution strategies. This screening is not only based on the original breeding area of ​​concern, set constraint knowledge points and target breeding execution strategy, but also combined with more specific information obtained from the match.

[0129] The server will first filter out those strategies that do not match the matching breeding area and matching constraint knowledge points. Then, among the remaining strategies, it will further filter out the strategy that is closest to the matching breeding execution strategy of the target breeding execution strategy.

[0130] The screening process may involve multiple aspects such as feature distance calculation, effect evaluation, cost-benefit analysis, etc. The server will comprehensively consider multiple factors and ultimately generate a set of target screening and farming execution strategies that are both in line with the needs of farmers and feasible.

[0131] For example, the server may filter out the following target screening farming execution strategy: "Adopt an indoor temperature-controlled environment in area D, combine low-cost and efficient feed with a phased and refined feeding strategy, and strengthen disease monitoring and prevention to ensure the consistency of pig growth rate while keeping costs within a reasonable range."

[0132] In a possible implementation, the method further includes: embedding the multiple setting strategy knowledge points to generate multiple setting knowledge point embedding vectors corresponding to the multiple setting strategy knowledge points.

[0133] Step D120 includes:

[0134] Step D121, embedding the farming area of ​​interest, the set constraint knowledge points and the target farming execution strategy to generate target farming area features of the farming area of ​​interest, target constraint features of the set constraint knowledge points and target strategy features of the target farming execution strategy.

[0135] Step D122, performing feature distance calculation based on the target breeding area feature, the target constraint feature, the target strategy feature and the multiple set knowledge point embedding vectors, generating multiple second feature distances between the target breeding area feature and the multiple set knowledge point embedding vectors, multiple third feature distances between the target constraint feature and the multiple set knowledge point embedding vectors, and multiple fourth feature distances between the target strategy feature and the multiple set knowledge point embedding vectors.

[0136] Step D123, based on the multiple second feature distances, the multiple third feature distances, the multiple fourth feature distances and the multiple set strategy knowledge points, the knowledge points of the focus breeding area, the set constraint knowledge points and the target breeding execution strategy are matched to generate the matching breeding area, the matching constraint knowledge points and the matching breeding execution strategy.

[0137] In this example, a pig farm in area E encountered the problem of slow growth and frequent diseases of pigs. The farmer hopes to improve the growth rate and reduce the incidence of diseases by optimizing the breeding strategy. The farmer submitted the target breeding feedback requirements based on area E to the server, clarifying the breeding area of ​​concern (area E), setting constraint knowledge points (cost controllable, no excessive increase in manpower investment) and target breeding execution strategy (increase growth rate, reduce disease incidence).

[0138] The server first embeds and represents multiple set strategy knowledge points stored in it. These knowledge points may include but are not limited to the characteristics of different breeding areas, various breeding constraints, and specific breeding execution strategies.

[0139] The server uses embedding technology (such as Word2Vec, BERT, etc.) to convert each set strategy knowledge point into a vector in a high-dimensional space, namely, the set knowledge point embedding vector, which can capture the semantic relationship and similarity between knowledge points.

[0140] The server then embeds the concerned breeding area (E area), set constraint knowledge points (cost controllable, no excessive increase in manpower input) and target breeding execution strategy (increase growth rate, reduce disease incidence) submitted by the farmers. Through similar embedding technology, the server generates the target breeding area feature vector of the concerned breeding area, the target constraint feature vector of the set constraint knowledge points and the target strategy feature vector of the target breeding execution strategy. These vectors represent the representation of the farmers' needs in multidimensional space.

[0141] The server now has a set of set knowledge point embedding vectors and a set of target feature vectors. Next, the server will calculate the feature distance between the target feature vector and each set knowledge point embedding vector.

[0142] In detail, the server calculates the feature distance (such as cosine similarity) between the feature vector of the target breeding area and the embedding vector of each set breeding area, and generates multiple second feature distances, which reflect the similarity between the E area and different set breeding areas.

[0143] Similarly, the server calculates the feature distance between the target constraint feature vector and the embedding vector of each set constraint knowledge point, and generates multiple third feature distances, which represent the degree of matching between the farmer's constraints and different set constraint knowledge points.

[0144] Finally, the server calculates the feature distance between the target strategy feature vector and each set farming execution strategy embedding vector, and generates multiple fourth feature distances, which measure the correlation between the target farming execution strategy and different set farming execution strategies.

[0145] The server performs a comprehensive evaluation based on the calculated second, third, and fourth feature distances and the corresponding set strategy knowledge points.

[0146] For the breeding area of ​​interest (area E), the server selects several set breeding areas with the smallest second feature distance as matching breeding areas. These areas may be most similar to area E in terms of environment, facilities, etc.

[0147] For setting constraint knowledge points (cost controllable and manpower investment does not increase too much), the server considers the set constraint knowledge points with the smallest third feature distance to ensure that the selected strategies can be implemented under the premise of satisfying these constraints.

[0148] For the target farming execution strategy (increasing growth rate, reducing disease incidence), the server focuses on setting farming execution strategies with smaller fourth feature distances, which may directly target the optimization goals of farmers.

[0149] Based on the above information, the server generates matching breeding areas for the breeding areas of interest, matching constraint knowledge points for the set constraint knowledge points, and matching breeding execution strategies for the target breeding execution strategies.

[0150] With the matching results, the server can more accurately screen out the set farming execution strategies that meet the needs of farmers. This process may involve further evaluation and analysis, such as considering the historical implementation effect of the strategy, cost-effectiveness ratio, required resources, etc. Finally, the server generates a set of target screening farming execution strategies for farmers to refer to and implement.

[0151] In a possible implementation, the plurality of set knowledge point embedding vectors include a plurality of set breeding area features, a plurality of set constraint features, and a plurality of set strategy features, and the method further includes: mapping the plurality of set breeding area features in the plurality of set knowledge point embedding vectors to generate labels, and generating breeding area feature labels of the plurality of set breeding area features. Mapping the plurality of set constraint features in the plurality of set knowledge point embedding vectors to generate labels, and generating constraint feature labels of the plurality of set constraint features. Mapping the plurality of set strategy features in the plurality of set knowledge point embedding vectors to generate labels, and generating strategy feature labels of the plurality of set strategy features.

[0152] Step D122 includes:

[0153] Step D1221, performing feature distance calculation on the target breeding area feature and a plurality of set breeding area features corresponding to the breeding area feature label, and generating a plurality of second feature distances between the target breeding area feature and the plurality of set breeding area features.

[0154] Step D1222, performing feature distance calculation on the target constraint feature and a plurality of set constraint features corresponding to the constraint feature label, and generating a plurality of third feature distances between the target constraint feature and the plurality of set constraint features.

[0155] Step D1223, performing feature distance calculation on the target policy feature and a plurality of set policy features corresponding to the policy feature label, and generating a plurality of fourth feature distances between the target policy feature and the plurality of set policy features.

[0156] In this example, a pig farm in area F faces problems such as inconsistent growth rate of pigs, low feed conversion rate, and high pressure of disease prevention and control. The farmer hopes to find a comprehensive breeding strategy that can comprehensively improve the breeding efficiency of area F while controlling costs. Therefore, the farmer submits a target breeding feedback demand based on area F to the server.

[0157] The server has converted multiple set breeding execution strategy data into set knowledge point embedding vectors through embedding representation technology. These vectors include multiple types such as set breeding area features, set constraint features and set strategy features.

[0158] Next, the server further processes these embedding vectors to generate mapping labels:

[0159] Breeding area feature labels: For multiple set breeding area feature embedding vectors, the server generates corresponding breeding area feature labels for each vector based on the specific content of each vector (such as breeding environment, facility conditions, etc.). These labels may include "indoor temperature-controlled breeding area", "outdoor free-range breeding area", "mountain free-range breeding area", etc.

[0160] Constraint feature labels: Similarly, the server generates constraint feature labels for multiple set constraint feature embedding vectors, such as "low-cost operation", "high manpower investment", "strict environmental protection requirements", etc.

[0161] Strategy feature label: For the set strategy feature embedding vector, the server generates corresponding strategy feature labels according to the farming execution strategy content it represents, such as "phased refined feeding", "automated disease monitoring system", "environmentally friendly feed formula", etc.

[0162] The server now has a set of set knowledge point embedding vectors containing mapping labels, as well as target breeding area features, target constraint features, and target strategy features corresponding to the target breeding feedback requirements submitted by the breeders.

[0163] Next, the server performs feature distance calculation:

[0164] Calculate the second characteristic distance: The server first identifies the target breeding area features and searches for the set breeding area feature embedding vectors with the same or similar breeding area feature labels in the set knowledge point embedding vectors. Then, the server calculates the characteristic distance (such as cosine similarity) between the target breeding area features and these vectors, and generates multiple second characteristic distances, which reflect the similarity between the F area and different set breeding areas.

[0165] Calculate the third feature distance: The server then performs a similar operation on the target constraint feature. It looks for the set constraint feature embedding vectors with the same or similar constraint feature labels and calculates the feature distance between the target constraint feature and these vectors, generating multiple third feature distances that measure the degree of match between the farmer's constraints and different set constraint knowledge points.

[0166] Calculate the fourth feature distance: Finally, the server processes the target policy feature. It searches for set policy feature embedding vectors with the same or similar policy feature labels and calculates the feature distance between the target policy feature and these vectors, generating multiple fourth feature distances that evaluate the correlation between the target farming execution strategy and different set farming execution strategies.

[0167] The server performs a comprehensive evaluation based on the calculated second, third, and fourth feature distances and the corresponding mapping labels. It prioritizes those set knowledge points whose feature distances are small and whose mapping labels match the target features.

[0168] For the farming area of ​​interest (area F), the server selects a set farming area with the smallest second feature distance and similar farming area feature labels as the matching farming area.

[0169] For set constraint knowledge points (such as cost controllability, environmental protection requirements, etc.), the server focuses on the set constraint features with a small third feature distance and matching constraint feature labels.

[0170] For the target farming execution strategy (such as increasing growth rate, reducing disease incidence, etc.), the server tends to select the set strategy features with a smaller fourth feature distance and related strategy feature labels.

[0171] Based on the above information, the server generates matching breeding areas for the breeding areas of interest, matching constraint knowledge points for the set constraint knowledge points, and matching breeding execution strategies for the target breeding execution strategies.

[0172] After obtaining the matching results, the server can further screen out the set breeding execution strategies that meet the needs of farmers and make necessary optimization adjustments. This process may involve comprehensive evaluation of multiple matching strategies, cost-benefit analysis, and consideration of factors such as the feasibility and complexity of strategy implementation.

[0173] Finally, the server generated a set of target-optimized breeding strategies for area F, providing farmers with scientific and feasible breeding plans. These strategies can not only help farmers solve the current problems they face, but also provide useful reference and guidance for future breeding management.

[0174] In a possible implementation, step S110 includes: if the target breeding feedback demand has complex feedback demand content based on the breeding area, using the breeding feedback mining model to mine key knowledge points of the target breeding feedback demand to generate the breeding area of ​​concern, the set constraint knowledge points and the target breeding execution strategy.

[0175] In this example, it is assumed that multiple farms under a large pig farming company have recently encountered different farming challenges. Among them, the farms in area G reported that the growth rate of pigs in the area is inconsistent, the pigs in some pig houses grow slowly, and there is a high incidence of diseases. At the same time, due to the remote location of the area, the cost of manpower and material investment is high, so the farmers hope to find a set of farming strategies that can both solve the current problems and control costs.

[0176] After receiving the feedback demand from the farm in region G, the server first parses and identifies it. By analyzing the demand description, the server determines that this is a complex feedback demand based on the breeding area, because it not only involves specific breeding problems (inconsistent growth rate, high disease incidence), but also contains specific regional characteristics (region G) and constraints (cost control).

[0177] The server starts the breeding feedback mining model, which has been trained and optimized through a large amount of historical breeding data, expert experience and successful cases, and can accurately understand the key information in the breeding feedback needs.

[0178] Focus on breeding area mining: The breeding feedback mining model first identifies the "G area" in the demand as the breeding area of ​​concern. It knows that different breeding areas may have different environments, facility conditions and breeding habits, so specific breeding strategies need to be formulated for the area.

[0179] Set constraint knowledge point mining: Next, the farming feedback mining model captures the constraint of "cost control". It understands that farmers not only focus on the effect but also pay great attention to the cost input when seeking solutions. Therefore, when mining the target farming execution strategy, these constraints need to be taken into consideration.

[0180] Mining target breeding execution strategies: Finally, the breeding feedback mining model targets the two specific problems of "inconsistent growth rate" and "high disease incidence", combines the characteristics of region G and the constraints of cost control, and mines possible target breeding execution strategies. These target breeding execution strategies may include adjusting feed formula, optimizing feeding management, strengthening disease prevention and control, and other aspects.

[0181] After in-depth analysis of the farming feedback mining model, the server generated the following key knowledge points:

[0182] Focus on breeding area: Area G

[0183] Setting Constraints Knowledge Point: Cost Control

[0184] Target farming implementation strategies: may include adopting a phased and refined feeding strategy to improve growth rate consistency; introducing an automated cleaning system to reduce the risk of disease transmission; at the same time, selecting cost-effective feed and veterinary drug products to control cost input.

[0185] The server will output the mined key knowledge points in a structured form for use in subsequent strategy screening and optimization processes. These knowledge points not only provide the server with a clear processing direction, but also provide farmers with actionable improvement suggestions.

[0186] In a possible implementation, the method further includes:

[0187] If the target breeding feedback demand has direct feedback demand content based on the breeding area, based on the target breeding feedback demand and the multiple set breeding execution strategy data, the multiple set breeding execution strategies are screened to generate a third screened breeding execution strategy for the target breeding feedback demand.

[0188] The breeding feedback mining model is used to make breeding strategy decisions on the target breeding feedback demand and the third screening breeding execution strategy, and a direct optimized breeding strategy for the target breeding feedback demand is generated.

[0189] In this embodiment, it is assumed that a pig farm finds that the weight of pigs in a specific pig house in area H is growing slowly, and the environmental conditions in the pig house are relatively stable and the breeding management is relatively standardized. Therefore, the farmer proposes a direct feedback demand based on the pig house (i.e., the breeding area), hoping to find a breeding execution strategy that can quickly increase the weight growth rate of pigs.

[0190] After receiving the farmer's feedback demand, the server first analyzes it and identifies that it is a direct feedback demand based on a specific breeding area (a pig house in area H). Different from complex feedback demands, direct feedback demands are usually more specific and direct, pointing to clear breeding problems and optimization goals.

[0191] The server will then screen strategies based on the target farming feedback requirements and the stored multiple set farming execution strategy data. These set farming execution strategy data may include a variety of different farming strategies, each targeting specific farming problems and optimization goals.

[0192] The server will traverse all set farming execution strategies to find those strategies that are directly related to the problem of slow weight gain in the pig house in area H and are likely to be effective. During the screening process, the server can consider multiple factors such as the historical implementation effect of the strategy, applicable conditions, and cost investment.

[0193] After screening, the server will generate a set of third-screen breeding execution strategies for the problem of slow weight gain in pig houses in area H. These strategies may include adjusting feed formulas, increasing exercise, improving breeding environments and other measures.

[0194] After the third screening of the breeding execution strategy, the server will not immediately recommend it to the breeder, but will further use the breeding feedback mining model to make breeding strategy decisions. This step aims to optimize the combination and fine-tune the screened strategies through the model's deep analysis capabilities to ensure that the final generated strategy can meet the needs of the breeder to the greatest extent.

[0195] The server provides the target farming feedback requirements and the third screening farming execution strategy as input information to the farming feedback mining model. The farming feedback mining model will conduct an in-depth analysis of this information, considering the interactions and influences between various factors, such as the impact of feed formula adjustment on weight growth rate, the potential benefits of increased exercise on disease prevention and control, etc. Based on the analysis results, the farming feedback mining model will generate a set of direct optimization farming strategies. These direct optimization farming strategies not only propose specific solutions to the problem of slow weight growth in pig houses in area H, but also consider the cost-effectiveness ratio, feasibility and long-term impact of strategy implementation.

[0196] Finally, the server will generate direct optimization breeding strategies in a structured form for farmers to refer to. These strategies may include detailed feed formula adjustment plans, specific measures to increase exercise volume, suggestions for improving the breeding environment, etc. Farmers can use these strategies to guide actual breeding work in order to achieve the goal of increasing the weight growth rate of pigs.

[0197] Figure 2 The hardware structure of the aquaculture feedback mining system 100 provided in the embodiment of the present application for implementing the above-mentioned aquaculture optimization method based on aquaculture feedback demand mining is shown as follows: Figure 2As shown, the farming feedback mining system 100 may include a processor 110 , a machine-readable storage medium 120 , a bus 130 , and a communication unit 140 .

[0198] In a possible design, the farming feedback mining system 100 can be a single server or a server group. The server group can be centralized or distributed (for example, the farming feedback mining system 100 can be a distributed system). In some embodiments, the farming feedback mining system 100 can be local or remote. For example, the farming feedback mining system 100 can access information and / or data stored in the machine-readable storage medium 120 via a network. For another example, the farming feedback mining system 100 can be directly connected to the machine-readable storage medium 120 to access the stored information and / or data. In some embodiments, the farming feedback mining system 100 can be implemented on a farming feedback mining system. As an example only, the farming feedback mining system can include a private cloud, a public cloud, a hybrid cloud, a community cloud, a distributed cloud, an internal cloud, a multi-layer cloud, etc. or any aggregation thereof.

[0199] The machine-readable storage medium 120 can store data and / or instructions. In some embodiments, the machine-readable storage medium 120 can store data obtained from an external terminal. In some embodiments, the machine-readable storage medium 120 can store data and / or instructions that the farming feedback mining system 100 uses to execute or use to complete the exemplary methods described in this application.

[0200] In the specific implementation process, one or more processors 110 execute computer executable instructions stored in the machine-readable storage medium 120, so that the processor 110 can execute the breeding optimization method based on breeding feedback demand mining in the above method embodiment. The processor 110, the machine-readable storage medium 120 and the communication unit 140 are connected through the bus 130, and the processor 110 can be used to control the sending and receiving actions of the communication unit 140.

[0201] The specific implementation process of the processor 110 can refer to the various method embodiments executed by the above-mentioned breeding feedback mining system 100. The implementation principles and technical effects are similar, and this embodiment will not be repeated here.

[0202] In addition, an embodiment of the present application also provides a readable storage medium, in which computer executable instructions are set. When the processor executes the computer executable instructions, the above-mentioned breeding optimization method based on breeding feedback demand mining is implemented.

[0203] It should be noted that in order to simplify the description disclosed in this application and thus help understand one or more embodiments of the invention, in the foregoing description of the embodiments of the present application, multiple features are sometimes combined into one embodiment, drawings, or descriptions thereof. Similarly, it should be noted that in order to simplify the description disclosed in this application and thus help understand one or more embodiments of the invention, in the foregoing description of the embodiments of the present application, multiple features are sometimes combined into one embodiment, drawings, or descriptions thereof.

Claims

1. A breeding optimization method based on breeding feedback demand mining, characterized in that: The method comprises: The breeding feedback mining model is used to mine key knowledge points of the target breeding feedback demand, and the breeding area of ​​concern, the set constraint knowledge points and the target breeding execution strategy of the target breeding feedback demand are generated; the target breeding feedback demand has feedback demand content based on the breeding area; Based on the concerned breeding area, the set constraint knowledge point, the target breeding execution strategy and multiple set breeding execution strategy data corresponding to the multiple set breeding execution strategies, the multiple set breeding execution strategies are screened to generate the target screened breeding execution strategy for the target breeding feedback demand; Extracting characteristic data of the target breeding feedback demand and the target screening breeding execution strategy to generate target characteristic data, using the breeding feedback mining model to make breeding strategy decisions on the target characteristic data to generate a target optimized breeding strategy; The method further comprises: Performing key knowledge point mining on the plurality of set breeding execution strategy data to generate a plurality of set strategy knowledge points corresponding to the plurality of set breeding execution strategy data; the plurality of set strategy knowledge points include at least one of a plurality of set breeding areas, a plurality of breeding constraint knowledge points and a plurality of set breeding execution strategies; Based on the multiple set strategy knowledge points, the concerned breeding area, the set constraint knowledge points and the target breeding execution strategy are matched to generate a matching breeding area for the concerned breeding area, a matching constraint knowledge point for the set constraint knowledge point and a matching breeding execution strategy for the target breeding execution strategy; The strategy screening of the multiple set breeding execution strategies based on the concerned breeding area, the set constraint knowledge point, the target breeding execution strategy and the multiple set breeding execution strategy data corresponding to the multiple set breeding execution strategies is performed to generate the target screening breeding execution strategy for the target breeding feedback demand, specifically: Based on the matching breeding area, the matching constraint knowledge point, the matching breeding execution strategy and the multiple set breeding execution strategy data, the multiple set breeding execution strategies are screened to generate the target screened breeding execution strategy; Wherein, the method further comprises: Embedding the plurality of setting strategy knowledge points to generate a plurality of setting knowledge point embedding vectors corresponding to the plurality of setting strategy knowledge points; The performing knowledge point matching on the concerned breeding area, the set constraint knowledge point and the target breeding execution strategy based on the multiple set strategy knowledge points to generate a matching breeding area for the concerned breeding area, a matching constraint knowledge point for the set constraint knowledge point and a matching breeding execution strategy for the target breeding execution strategy includes: Embedding the farming area of ​​interest, the set constraint knowledge point and the target farming execution strategy to generate target farming area features of the farming area of ​​interest, target constraint features of the set constraint knowledge point and target strategy features of the target farming execution strategy; Based on the target breeding area feature, the target constraint feature, the target strategy feature and the multiple set knowledge point embedding vectors, feature distance calculation is performed to generate multiple second feature distances between the target breeding area feature and the multiple set knowledge point embedding vectors, multiple third feature distances between the target constraint feature and the multiple set knowledge point embedding vectors, and multiple fourth feature distances between the target strategy feature and the multiple set knowledge point embedding vectors; Based on the multiple second feature distances, the multiple third feature distances, the multiple fourth feature distances and the multiple set strategy knowledge points, knowledge point matching is performed on the focus breeding area, the set constraint knowledge points and the target breeding execution strategy to generate the matching breeding area, the matching constraint knowledge points and the matching breeding execution strategy.

2. The aquaculture optimization method based on aquaculture feedback demand mining according to claim 1 is characterized in that: The multiple set breeding execution strategy data are multiple strategy feature vectors; the multiple set breeding execution strategy data corresponding to the multiple set breeding execution strategies are screened based on the concerned breeding area, the set constraint knowledge points, the target breeding execution strategy and the multiple set breeding execution strategies to generate the target screened breeding execution strategy for the target breeding feedback demand, including: Embedding the concerned breeding area, the set constraint knowledge point and the target breeding execution strategy to generate a target knowledge point embedding vector of the target breeding feedback requirement; Performing feature distance calculation on the target knowledge point embedding vector and the multiple strategy feature vectors to generate multiple first feature distances between the target knowledge point embedding vector and the multiple strategy feature vectors; The multiple set farming execution strategies are screened based on the multiple first characteristic distances to generate the target screened farming execution strategy.

3. The aquaculture optimization method based on aquaculture feedback demand mining according to claim 1 is characterized in that: The multiple set breeding execution strategy data are multiple strategy tag attributes; the multiple set breeding execution strategy data corresponding to the multiple set breeding execution strategies are screened based on the concerned breeding area, the set constraint knowledge point, the target breeding execution strategy and the multiple set breeding execution strategies to generate the target screened breeding execution strategy for the target breeding feedback demand, including: Determining target optimization knowledge points for target breeding feedback requirements based on the breeding area of ​​interest, the set constraint knowledge points and the target breeding execution strategy; Determining a target label attribute associated with the target optimization knowledge point from among the multiple strategy label attributes; The plurality of set farming execution strategies are screened based on the target tag attribute to generate the target screened farming execution strategy.

4. The aquaculture optimization method based on aquaculture feedback demand mining according to claim 1 is characterized in that: The multiple set breeding execution strategy data include multiple strategy feature vectors and multiple strategy label attributes; the multiple set breeding execution strategy data corresponding to the multiple set breeding execution strategies are screened based on the concerned breeding area, the set constraint knowledge point, the target breeding execution strategy and the multiple set breeding execution strategies to generate the target screened breeding execution strategy for the target breeding feedback demand, including: Embedding the concerned breeding area, the set constraint knowledge point and the target breeding execution strategy to generate a target knowledge point embedding vector of the target breeding feedback requirement; Performing feature distance calculation on the target knowledge point embedding vector and the multiple strategy feature vectors to generate multiple first feature distances between the target knowledge point embedding vector and the multiple strategy feature vectors; Performing strategy screening on the plurality of set farming execution strategies based on the plurality of first characteristic distances to generate a first screened farming execution strategy; Determining target optimization knowledge points for target breeding feedback requirements based on the breeding area of ​​interest, the set constraint knowledge points and the target breeding execution strategy; Determining a target label attribute associated with the target optimization knowledge point from among the multiple strategy label attributes; Performing strategy screening on the plurality of set farming execution strategies based on the target tag attribute to generate a second screened farming execution strategy; The first screening and breeding execution strategy and the second screening and breeding execution strategy are aggregated to generate the target screening and breeding execution strategy.

5. The aquaculture optimization method based on aquaculture feedback demand mining according to claim 4 is characterized in that: The plurality of set knowledge point embedding vectors include a plurality of set breeding area features, a plurality of set constraint features, and a plurality of set strategy features, and the method further includes: Mapping labels to multiple set breeding area features in the multiple set knowledge point embedding vectors to generate breeding area feature labels for the multiple set breeding area features; mapping labels to multiple set constraint features in the multiple set knowledge point embedding vectors to generate constraint feature labels for the multiple set constraint features; mapping labels to multiple set strategy features in the multiple set knowledge point embedding vectors to generate strategy feature labels for the multiple set strategy features; The feature distance calculation is performed based on the target breeding area feature, the target constraint feature, the target strategy feature, and the multiple set knowledge point embedding vectors to generate multiple second feature distances between the target breeding area feature and the multiple set knowledge point embedding vectors, multiple third feature distances between the target constraint feature and the multiple set knowledge point embedding vectors, and multiple fourth feature distances between the target strategy feature and the multiple set knowledge point embedding vectors, including: Performing feature distance calculation on the target breeding area feature and a plurality of set breeding area features corresponding to the breeding area feature label to generate a plurality of second feature distances between the target breeding area feature and the plurality of set breeding area features; Performing feature distance calculation on the target constraint feature and a plurality of set constraint features corresponding to the constraint feature label to generate a plurality of third feature distances between the target constraint feature and the plurality of set constraint features; A feature distance calculation is performed on the target policy feature and a plurality of set policy features corresponding to the policy feature label to generate a plurality of fourth feature distances between the target policy feature and the plurality of set policy features.

6. The aquaculture optimization method based on aquaculture feedback demand mining according to claim 1 is characterized in that: The method of using the aquaculture feedback mining model to mine key knowledge points of the target aquaculture feedback demand, generating the aquaculture area of ​​interest, setting constraint knowledge points and the target aquaculture execution strategy of the target aquaculture feedback demand, includes: If the target breeding feedback demand has complex feedback demand content based on the breeding area, the breeding feedback mining model is used to mine key knowledge points of the target breeding feedback demand to generate the breeding area of ​​concern, the set constraint knowledge points and the target breeding execution strategy.

7. The aquaculture optimization method based on aquaculture feedback demand mining according to claim 1 is characterized in that: The method further comprises: If the target breeding feedback demand has direct feedback demand content based on the breeding area, based on the target breeding feedback demand and the multiple set breeding execution strategy data, the multiple set breeding execution strategies are screened to generate a third screened breeding execution strategy for the target breeding feedback demand; The breeding feedback mining model is used to make breeding strategy decisions on the target breeding feedback demand and the third screening breeding execution strategy, and a direct optimized breeding strategy for the target breeding feedback demand is generated.

8. A farming feedback mining system, characterized in that: The breeding feedback mining system includes a processor and a memory, the memory is connected to the processor, the memory is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the memory to implement the breeding optimization method based on breeding feedback demand mining as described in any one of claims 1 to 7.

Images