System for assessing the correlation between the health of the pig's intestinal tract and trace element absorption

Abstract

The present application relates to the technical field of animal husbandry, in particular to a pig intestinal health and trace element absorption correlation evaluation system, which comprises a data storage and input module, a multi-source information fusion and feature extraction module, an intelligent evaluation core module, an adaptive decision support module and an output response module. The data storage and input module stores pig population history and real-time multi-modal data and receives external real-time data. The multi-source information fusion and feature extraction module extracts multi-dimensional feature vectors from multi-modal data using a feature system. The intelligent evaluation core module trains a model using an ensemble learning algorithm and outputs an intestinal health index, an absorption efficiency index and a correlation score. The adaptive decision support module generates a ration adjustment, a prebiotic addition and an economic benefit prediction report based on the evaluation results and a rule base. The output response module visually presents the report and the core indicators. The present application overcomes the traditional lag, isolation and empirical defects, improves production performance, reduces breeding costs and environmental pollution, and achieves comprehensive benefits such as dynamic quantitative evaluation and precise management.

Description

Technical Field

[0001] This invention relates to the field of animal husbandry technology, and more specifically, to a system for assessing the correlation between intestinal health and trace element absorption in pigs. Background Technology

[0002] Currently, in large-scale pig farming, the precise supply of trace elements (such as zinc, copper, and selenium) is a crucial link in ensuring animal health and improving production performance. However, traditional management models have many limitations in this regard:

[0003] First, the assessment methods are outdated and isolated: Currently, the assessment of pig intestinal health relies heavily on empirical visual observation (such as fecal scoring) or post-mortem dissection, while the assessment of trace element absorption efficiency is mainly conducted through expensive, in vitro laboratory tests (such as atomic absorption spectrometry to determine the content in feces or blood). These methods are not only time-consuming and labor-intensive, but more importantly, they assess in isolation from the two highly related physiological processes of intestinal health and trace element absorption, failing to reflect the intrinsic connection between the two in real time and dynamically.

[0004] Secondly, the reliance on experience-based decision-making lacks precision: most micronutrient supplementation plans in diets are based on fixed nutritional standards and production stages, or rely on extensive adjustments made based on the experience of farmers. This "one-size-fits-all" approach cannot respond to individual differences in pig herds, fluctuations in health status, and environmental changes. It is highly likely that micronutrients will either be insufficiently added, affecting health and production, or excessively added, causing resource waste and environmental pollution (such as soil pollution caused by high zinc and high copper manure).

[0005] Third, the problem of data silos is serious: farms accumulate a large amount of multimodal data on a daily basis, including production performance records, feed formulations, environmental monitoring data, etc., but this data is usually scattered in different systems or logbooks, independent of each other, forming "data silos". There is a lack of effective technical means to integrate, analyze and extract insightful data from these multi-source heterogeneous data to make decisions, and a large amount of data value is buried.

[0006] Fourth, the causal relationship is difficult to quantify: Although practitioners generally recognize that gut health affects the absorption of trace elements, and vice versa, the strength of this interaction, and specifically which trace elements are most closely associated with which gut health indicators, is difficult to quantify precisely. Managers cannot answer precise questions such as "How much will improving the absorption efficiency of trace elements increase if gut health is improved by 10%?", and decision-making lacks data support.

[0007] In view of this, a correlation assessment system for pig gut health and trace element absorption is proposed. Summary of the Invention

[0008] The purpose of this invention is to provide a system for assessing the correlation between pig gut health and trace element absorption, so as to achieve real-time, dynamic and quantitative assessment of the correlation between pig gut health and trace element absorption, and to provide precise and personalized nutritional regulation programs based on the assessment results.

[0009] To solve the above-mentioned technical problems, the present invention provides a system for assessing the correlation between porcine intestinal health and trace element absorption, comprising:

[0010] The data storage module is used to store historical and real-time multimodal data of the pig herd;

[0011] The input module, connected to the data storage module, is used to receive real-time multimodal pig herd data from an external data source;

[0012] A multi-source information fusion and feature extraction module, connected to the input module and the data storage module respectively, is used to standardize and normalize the received multimodal data, and extract multidimensional feature vectors for evaluating the correlation from the multimodal data based on a preset feature system for quantifying the dynamic correlation between intestinal health and trace element absorption; the feature system includes trace element apparent absorption rate features, trace element bioaccumulation features, intestinal physiological stress features, and production performance features;

[0013] The intelligent assessment core module is connected to the multi-source information fusion and feature extraction module, and it contains a correlation dynamic assessment model. The correlation dynamic assessment model is trained by an ensemble learning algorithm on historical multidimensional feature vectors and label data verified by in vitro detection. It is used to synchronously calculate and output the gut health index, the trace element absorption efficiency index and the correlation score that characterizes the correlation strength between the two based on the input real-time multidimensional feature vector.

[0014] The adaptive decision support module, connected to the intelligent assessment core module, is used to generate a structured report containing personalized micronutrient diet adjustment suggestions, prebiotic addition schemes, and expected economic benefit predictions based on correlation scores, gut health index, micronutrient absorption efficiency index, and a preset decision rule base.

[0015] The output response module, connected to the adaptive decision support module, is used to visualize the structured report and core evaluation indicators through a graphical user interface.

[0016] As a further improvement to this technical solution, the input module includes:

[0017] A standardized data interface unit is used to receive the multimodal data of the pig herd from the production management record system, feed formula database, in vitro detection equipment and manual observation record terminal;

[0018] Edge computing units, deployed at data acquisition terminals, are used to perform preliminary cleaning, redundancy removal, and formatting of initial data.

[0019] As a further improvement to this technical solution, the multimodal data of the pig herd includes production performance data, dietary trace element composition data, fecal trace element residue content data, hair trace element enrichment content data, intestinal health visual score data, and pig house environmental monitoring data.

[0020] As a further improvement to this technical solution, the correlation dynamic evaluation model is an ensemble learning model built based on the gradient boosting decision tree algorithm;

[0021] The real-time multidimensional feature vector input includes: dietary zinc level, fecal copper residue rate, hair selenium enrichment concentration, diarrhea rate, average daily weight gain, and intestinal visual score; the model output includes the predicted intestinal health index and the comprehensive utilization rate of trace elements.

[0022] The correlation dynamic assessment model quantifies the contribution of each input's real-time multidimensional feature vector to the gut health index and trace element absorption efficiency index, and reveals the correlation strength between different trace elements and different gut health indicators in the form of a heat map.

[0023] As a further improvement to this technical solution, the system also includes:

[0024] The model continuous learning module is connected to the intelligent evaluation core module and the data storage module, respectively. It is used to periodically use newly added multimodal data of pig herds and corresponding validation label data to incrementally train and optimize the correlation dynamic evaluation model in order to update the model parameters.

[0025] As a further improvement to this technical solution, the preset decision rule base in the adaptive decision support module combines production rules with case-based reasoning, and the rules in the decision rule base are in the following form:

[0026] IF correlation score < threshold T1 AND trace element absorption efficiency index ∈ [range A] AND gut health index ∈ [range B];

[0027] THEN Implementation: Adjust the organic zinc supplementation level in the diet to X mg / kg and recommend the use of prebiotic Y, which is expected to improve absorption efficiency by Z% and reduce diarrhea rate by W%; where the specific values ​​of threshold T1, range A, range B, parameters X, Z, and W, as well as the type of Y, are dynamically determined based on statistical analysis of historical datasets and case reasoning results.

[0028] As a further improvement to this technical solution, the visualization interface generated by the output response module includes a correlation analysis dashboard, which displays the correlation strength between different trace elements and different intestinal health indicators as revealed by the correlation dynamic assessment model in the form of a heat map, and displays the historical changes of core key performance indicators and the predicted trajectory based on current decision recommendations in the form of trend lines.

[0029] As a further improvement to this technical solution, the system also includes an early warning module, which is connected to the intelligent assessment core module and the adaptive decision support module, respectively.

[0030] The early warning module is configured to automatically trigger multi-level early warning signals when the correlation score is lower than a preset threshold, and push the early warning information and corresponding structured decision suggestions to the user terminal of the designated manager.

[0031] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0032] 1. This system for assessing the correlation between pig gut health and trace element absorption integrates data fusion, intelligent assessment, and decision support, enabling dynamic, quantitative assessment and precise management of the correlation between pig gut health and trace element absorption. It effectively overcomes the lag, isolation, and empirical deficiencies of traditional models, thereby generating significant comprehensive benefits in improving production performance, reducing breeding costs, and reducing environmental pollution.

[0033] 2. In this system for assessing the correlation between pig gut health and trace element absorption, the data storage module, input module, and multi-source information fusion and feature extraction module can systematically access, clean, standardize, and integrate multimodal data from production management, feed formulation, in vitro testing, manual observation, and environmental monitoring. This breaks down data silos and extracts high-value feature vectors for assessing correlation based on a specially designed feature system, laying a solid data foundation for accurate assessment.

[0034] 3. In this system for assessing the correlation between porcine gut health and trace element absorption, a dynamic correlation assessment model trained within the intelligent assessment core module can simultaneously and rapidly calculate the gut health index, the trace element absorption efficiency index, and, most importantly, the correlation score, which characterizes the strength of the correlation between the two. This changes the previous reliance on subjective experience and delayed detection, achieving a technological leap from "qualitative inference" to "quantitative assessment."

[0035] 4. In this system for assessing the correlation between swine gut health and trace element absorption, the adaptive decision support module and its pre-defined decision rule base can automatically generate a structured report containing specific adjustment parameters, prebiotic addition plans, and expected economic benefit predictions based on real-time assessed indices and scores. This allows trace elements to move from "general-purpose" to "personalized customization," significantly improving the accuracy of nutrient utilization. Attached Figure Description

[0036] Figure 1 This is an overall system block diagram of the present invention. Detailed Implementation

[0037] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0038] Example 1: In traditional aquaculture management, gut health and trace element absorption are two isolated assessment links. Decision-making and delayed detection rely on experience, which fails to reveal their inherent dynamic relationship. This results in either insufficient addition of trace elements affecting production or excessive addition causing waste and pollution.

[0039] In view of this, please refer to Figure 1 As shown, the purpose of this invention is to provide a system for assessing the correlation between porcine intestinal health and trace element absorption, which includes:

[0040] The data storage module stores historical and real-time multimodal data of the pig herd. This module can be a central server configured with a MySQL database and an HDFS distributed file system. MySQL is used to store structured production performance data and diet formulation data, while HDFS is used to store unstructured images (such as gut visual scoring images) and time-series data (such as environmental monitoring data).

[0041] The input module, connected to the data storage module, is used to receive real-time multimodal pig herd data from external data sources. The input module connects to various data sources via wired / wireless networks, as detailed below:

[0042] A RESTful API interface is used to receive production data (daily weight gain, feed intake) in JSON format from production management record systems (such as PigWin).

[0043] An ODBC / JDBC interface is used to periodically retrieve dietary trace element composition data from the feed formulation database; a dedicated data parser is used to parse test result files (including fecal and hair trace element content) transmitted via USB or Bluetooth from in vitro testing equipment (such as an ICP-MS spectrometer).

[0044] The mobile app serves as a terminal for manual observation and recording, allowing keepers to upload photos of diarrhea rates, mental status scores, and intestinal visual scores.

[0045] The multi-source information fusion and feature extraction module is connected to the input module and the data storage module, respectively. It is implemented using a Python data processing service deployed on a server, and its workflow is as follows:

[0046] The received multimodal data is standardized and normalized (first, the received multimodal data is cleaned (missing values ​​and outliers are handled), and then all numerical features are Z-score standardized to eliminate the influence of units).

[0047] Based on a pre-defined feature system for quantifying the dynamic correlation between gut health and trace element absorption, multidimensional feature vectors for assessing the correlation are extracted from multimodal data. The feature system includes trace element apparent absorption rate features, trace element bioaccumulation features, gut physiological stress features, and production performance features. For example, the apparent absorption rate of zinc is calculated from "dietary zinc level" and "fecal zinc residue"; the bioaccumulation coefficient of selenium is calculated from "hair selenium concentration"; the gut physiological stress index is extracted from "diarrhea rate" and "gut visual score"; and the production performance score is extracted from "average daily weight gain" and "feed ratio".

[0048] The intelligent assessment core module is connected to the multi-source information fusion and feature extraction module, and it contains a correlation dynamic assessment model. The correlation dynamic assessment model is trained by an ensemble learning algorithm (XGBoost algorithm, a gradient boosting decision tree library) on historical multidimensional feature vectors and label data verified by in vitro detection (verified "real gut health status" and "real trace element absorption rate" through in vitro detection). It is used to synchronously calculate and output the gut health index (GHI, 0-100), the trace element absorption efficiency index (TEI, 0-100), and the correlation score (CS, 0-1) that characterizes the correlation strength between the two based on the input real-time multidimensional feature vector. This score is derived from the sum of feature importance calculated internally by the model, which quantifies the correlation strength between GHI and TEI.

[0049] The adaptive decision support module, connected to the intelligent assessment core module, is used to generate a structured report containing personalized micronutrient diet adjustment suggestions, prebiotic supplementation plans, and expected economic benefit predictions based on correlation scores, gut health indices, micronutrient absorption efficiency indices, and a pre-set decision rule base (loaded using the Drools rule engine, which contains dozens of production rules). Specifically, the process of generating the structured report is as follows:

[0050] The adaptive decision support module receives three indices output by the intelligent evaluation core module and matches them against conditions in the rule base. For example, if a rule is matched, a decision is triggered, generating a structured report in JSON format containing: "It is recommended to increase the organic zinc in the diet from 80ppm to 100ppm," "It is recommended to add 500g / ton of prebiotic fructooligosaccharides (FOS)," "The expected zinc absorption rate will increase by 15%," and "The expected diarrhea rate will decrease by 5% next week."

[0051] The output response module, connected to the adaptive decision support module, is a web application developed based on the Vue.js framework. It serves as a graphical user interface (GUI) and mainly includes data dashboards, trend charts, and detailed results of this evaluation. It is used to visualize the structured report and core evaluation indicators through the graphical user interface.

[0052] By constructing the aforementioned system, multimodal data, feature engineering, machine learning models, and rule engines are integrated into an automated closed-loop system for the first time. This achieves end-to-end automation from data to decision-making, solving the long-standing problems of "data silos" and "decision lag." A model is trained to quantify the "correlation strength" between gut health and micronutrient absorption efficiency, and this abstract score is directly used as a trigger for precise nutritional regulation. This cross-disciplinary technological integration (livestock farming + data science + decision science) has generated a synergistic effect, resulting in significant comprehensive benefits in improving production performance, reducing breeding costs, and minimizing environmental pollution.

[0053] Example 2: Considering the inconsistent formats of external data sources (such as detection equipment and manual terminals) and the presence of noise in the raw data (such as outliers and duplicates), direct transmission to the backend would lead to low processing efficiency and decreased model accuracy; simultaneously, traditional systems lack edge processing, and all data is uploaded to the cloud, which can easily cause network congestion; therefore, the input module includes:

[0054] A standardized data interface unit is used to receive the multimodal data of the pig herd from the production management record system, feed formulation database, in vitro testing equipment, and manual observation and recording terminal; specifically:

[0055] The standardized data interface unit is a universal data access gateway, written in Java, integrating the aforementioned RESTful API, JDBC driver, file listening service, and an MQTT message broker (for receiving data from IoT environment sensors). All access data is first converted into a unified internal data format (Apache Avro format).

[0056] Edge computing units, deployed at data acquisition terminals, are used to perform preliminary cleaning, redundancy removal, and formatting of initial data; specifically:

[0057] Lightweight edge computing programs are deployed on data acquisition terminals (such as tablets with an app installed or industrial control computers connected to testing equipment). For example, on a tablet, the app calls a locally running TensorFlow Lite model to perform preliminary analysis on intestinal images before uploading them, automatically generating a preliminary visual score, compressing the image size, completing preliminary cleaning and formatting, and then uploading it to the server, greatly reducing server load and network bandwidth pressure.

[0058] Through the "cloud-edge" collaborative architecture, standardized and lightweight preprocessing of data was achieved, ensuring high quality and consistency of input data and laying a solid foundation for subsequent accurate analysis.

[0059] Example 3: Considering that traditional data collection only focuses on "production performance (e.g., ADG)" and "dietary composition," ignoring "micronutrient bioaccumulation (e.g., hair content)" and "direct indicators of gut health (e.g., visual scores)," resulting in a single data dimension that cannot support the correlation assessment between the two (e.g., it cannot determine whether "low selenium in hair is due to poor absorption or insufficient diet"). Therefore, multimodal data for pig herds includes production performance data, dietary micronutrient composition data, fecal micronutrient residue data, hair micronutrient bioaccumulation data, gut health visual score data, and pig house environmental monitoring data. Among these, production performance data includes average daily gain (ADG), average daily feed intake (ADFI), and feed conversion ratio (FCR); dietary micronutrient composition data includes Zn, Cu, and F. The data includes the following: the form (inorganic / organic) and specific content (mg / kg) of added elements such as e, Mn, and Se; fecal trace element residue data, obtained by ICP-MS detection of the dry matter content (mg / kg) of the above elements in feces; hair trace element enrichment data, obtained by ICP-MS detection of the content (mg / kg) of the above elements in pig back hair, reflecting a stable indicator of medium- to long-term absorption; intestinal health visual scoring data, where veterinarians or trained breeders score the appearance (color, elasticity, contents, lymphoid follicles) of the intestines after slaughter on a 5-point scale according to standard procedures, and upload images via an app for verification; and pigsty environmental monitoring data, including real-time collection of indoor temperature, humidity, and ammonia concentration (ppm) via IoT sensors.

[0060] By introducing hair enrichment data as a biomarker and environmental data as a stressor, the assessment dimensions are greatly enriched. The non-obviousness lies in combining the medium- to long-term indicator of "hair trace elements" with the short-term indicator of "fecal residue" and linking it with environmental stress, thus constructing a more comprehensive assessment system.

[0061] Example 4: Because machine learning models are "black boxes," even if a prediction is made, users cannot understand which factors influence the prediction result or to what extent, making it difficult to trust and adopt the decision. Simultaneously, the model has few input features (only 2-3) and outputs only a single indicator (such as absorption efficiency), making it impossible to quantify the correlation strength, and the results are not intuitive. Therefore, the correlation dynamic evaluation model is an ensemble learning model built based on the gradient boosting decision tree algorithm. The real-time multidimensional feature vector input includes: dietary zinc level, fecal copper residue rate, hair selenium enrichment concentration, diarrhea rate, average daily weight gain, and gut visual score. The model output includes the predicted gut health index and trace element comprehensive utilization rate. The correlation dynamic evaluation model quantifies the contribution of each real-time multidimensional feature vector input to the gut health index and trace element absorption efficiency index, revealing the correlation strength between different trace elements and different gut health indicators in the form of a heatmap. Specifically:

[0062] First, the correlation dynamic evaluation model is a gradient boosting decision tree (GBDT) model built using Python's XGBoost library;

[0063] Secondly, the input features are the multi-dimensional feature vectors extracted in Example 1, including but not limited to: dietary zinc level, fecal copper residue rate, hair selenium enrichment concentration, diarrhea rate, average daily weight gain, intestinal visual score and more than 20 other features.

[0064] Then, the correlation dynamic assessment model has two output heads, which predict the gut health index (GHI) and the comprehensive utilization rate of trace elements (TEU), respectively.

[0065] Finally, after training, the SHAP (SHapley Additive exPlanations) library is used to calculate the contribution of each input feature to the two prediction results (SHAP value). Then, the contribution relationship between trace element-related features and gut health-related features is visualized in the form of a heatmap. The color intensity in the heatmap directly reveals the correlation strength between different trace elements and different gut health indicators, thus visualizing the correlation strength.

[0066] By utilizing the aforementioned dynamic correlation assessment model and employing SHAP interpretability technology, the model's black box is opened, making the abstract "correlation score" interpretable and verifiable. This significantly enhances users' trust in the system's decisions. Advanced interpretable AI (XAI) technology is then applied to the aquaculture field to quantify correlation strength.

[0067] Example 5: Considering that traditional models have fixed parameters after training, when the pig breed is updated (e.g., from Large White to Duroc-Landrace-Large White crossbred) or the feed formula is adjusted (e.g., from inorganic zinc to organic zinc), the deviation between the model and the actual scenario increases, requiring manual retraining (long cycle, high cost), leading to a decrease in evaluation accuracy. Therefore, the system also includes:

[0068] The model continuous learning module, connected to both the intelligent evaluation core module and the data storage module, is used to periodically perform incremental training and optimization of the correlation dynamic evaluation model using newly added multimodal data of pig herds and corresponding validation label data, in order to update the model parameters. The specific scheme is as follows:

[0069] The system is set up with a scheduled task (e.g., once a month). The model continuous learning module will automatically collect data that has been newly added in the past month and has corresponding validation labels (e.g., a new batch of in vitro test reports). Then, it will use this new data to perform incremental training (or fine-tuning) on ​​the original model.pkl. At the same time, the module will monitor the model's performance metrics (e.g., MSE) on the latest validation set. If the performance drops below the threshold, it will automatically trigger the retraining process and update the new model in the production environment with the better-performing model.

[0070] Through the aforementioned technical means, the system acquires the ability to self-evolve and continuously optimize, adapting to changes in the farm and maintaining long-term assessment accuracy. This is crucial in dynamically changing farming environments. Furthermore, it enables automated model iteration without human intervention.

[0071] Example 6: Considering that traditional decision-making rules are based on "fixed threshold + experience-based suggestions" (e.g., "add zinc if diarrhea rate > 5%)", there is no quantitative basis, and they do not combine "relevance score" and "economic benefits", leading to blind decision-making (e.g., the zinc addition amount is set to 150mg / kg based on experience, but in reality 120mg / kg would be sufficient, resulting in wasted costs); at the same time, there are no case references, making it unable to cope with complex scenarios (e.g., "low correlation + low absorption efficiency + moderate gut health"), therefore, the preset decision rule base in the adaptive decision support module is a combination of production rules and case reasoning. The rules in the decision rule base are in the following form:

[0072] IF correlation score < threshold T1 AND trace element absorption efficiency index ∈ [range A] AND gut health index ∈ [range B]; for example: IF correlation score CS < 0.7 AND trace element absorption efficiency index TEI ∈ [30, 50] AND gut health index GHI ∈ [40, 60];

[0073] THEN Implementation: Adjust the organic zinc supplementation level in the diet to X mg / kg and recommend the use of prebiotic Y, which is expected to improve absorption efficiency by Z% and reduce diarrhea rate by W%; For example: THEN Implementation: Adjust the organic zinc supplementation level in the diet to 120 mg / kg and recommend the use of prebiotic (fructooligosaccharides), which is expected to improve zinc absorption efficiency by 18% and reduce diarrhea rate by 8%.

[0074] The specific values ​​of the threshold T1, value range A, value range B, parameters X, Z, and W, and the type of Y in the rules are not fixed. The system backend maintains a case database storing historical successful intervention cases. When new data is input, the system simultaneously performs a case search to find the most similar historical cases. The thresholds and parameters in the above rules are dynamically fine-tuned based on the results of similar cases (for example, adjusting 120mg / kg to 115mg / kg), making the decision more accurate.

[0075] The above design combines rule-based rapid response with case-based precise optimization, achieving a balance between "principle-based and flexibility." This hybrid decision-making mechanism makes the system's recommendations not only reliable but also increasingly personalized, and its effects (improved absorption rate, reduced diarrhea rate) are predictable and verifiable.

[0076] Example 7: Considering that complex evaluation results and decision reports, if presented in text form, are not intuitive and fail to quickly grasp the key points; and that abnormal situations require constant manual monitoring of the screen to detect, the visualization interface generated by the output response module includes a correlation analysis dashboard. This dashboard displays the correlation strength between different trace elements and different gut health indicators revealed by the correlation dynamic evaluation model in the form of a heatmap, and shows the historical changes of core key performance indicators and the predicted trajectory based on current decision recommendations in the form of trend lines. The system also includes an early warning module, connected to the intelligent evaluation core module and the adaptive decision support module, respectively. The early warning module is configured to automatically trigger multi-level early warning signals when the correlation score is lower than a preset threshold, and push the early warning information and corresponding structured decision recommendations to the user terminal of the designated manager. Specifically:

[0077] Develop correlation analysis dashboards using the ECharts library. Core components include:

[0078] Heatmap: Directly displays the SHAP value heatmap generated in Example 4, visually showing the correlation strength of pairings such as selenium-diarrhea rate and zinc-intestinal score.

[0079] Trend lines: Show the historical change curves of key indicators (such as GHI, TEI, CS) over the past 90 days, and draw the predicted future trajectory with dashed lines based on current decision-making recommendations;

[0080] The early warning module monitors the correlation score (CS) output by the intelligent evaluation core module. Two threshold levels are set: when CS < 0.6, a yellow warning is triggered, and a pop-up notification appears in the GUI; when CS < 0.4, a red alert is triggered, and in addition to the GUI notification, the alert information and key decision-making suggestions (“Urgent adjustment of the zinc-copper scheme is needed!”) are automatically sent to the site manager's mobile phone via SMS.

[0081] By employing the aforementioned technologies, data is transformed into intuitive graphics, decision-making into forward-looking predictions, and passive queries into proactive early warnings, significantly improving human-computer interaction efficiency and system usability. This proactive, multi-channel early warning design ensures that critical issues are addressed promptly, demonstrating the system's intelligence and practicality.

[0082] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A system for assessing the correlation between the health of the intestinal tract of a pig and the absorption of trace elements, characterized in that it comprises: The application relates to a pig intestinal health and trace element absorption dynamic correlation evaluation system, which comprises the following parts: a data storage module for storing historical and real-time pig group multi-modal data; an input module connected with the data storage module for receiving real-time pig group multi-modal data from external data sources; a multi-source information fusion and feature extraction module connected with the input module and the data storage module, used for standardizing and normalizing the received multi-modal data, and extracting a multi-dimensional feature vector for evaluating correlation from the multi-modal data based on a preset feature system for quantifying the dynamic correlation between intestinal health and trace element absorption; the pig group multi-modal data comprises production performance data, ration trace element composition data, fecal trace element residual content data, hair trace element enrichment content data, intestinal health visual score data and pig house environment monitoring data; the feature system comprises trace element apparent absorption rate features, trace element biological enrichment features, intestinal physiological stress features and production performance features; an intelligent evaluation core module connected with the multi-source information fusion and feature extraction module, which is provided with a correlation dynamic evaluation model; the correlation dynamic evaluation model is an integrated learning model constructed based on a gradient boosting decision tree algorithm, is obtained by training historical multi-dimensional feature vectors and label data verified by in-vitro detection, and is used for synchronously calculating and outputting an intestinal health index, a trace element absorption efficiency index and a correlation degree score representing the correlation strength between the two indexes according to input real-time multi-dimensional feature vectors; the correlation dynamic evaluation model quantifies the contribution of each input real-time multi-dimensional feature vector to the intestinal health index and the trace element absorption efficiency index by SHAP interpretive technology, and reveals the correlation strength between different trace elements and different intestinal health indexes in the form of a heat map; an adaptive decision support module connected with the intelligent evaluation core module, used for generating a structured report containing individualized trace element ration adjustment suggestions, prebiotic addition schemes and expected economic benefit forecasts according to the correlation degree score, the intestinal health index, the trace element absorption efficiency index and a preset decision rule base; the preset decision rule base is a combination of production rules and case-based reasoning, and the rule form in the decision rule base is: IF correlation degree score < threshold value T1 AND trace element absorption efficiency index [value range A] AND intestinal health index [value range B]; THEN execute: adjust the organic zinc addition level in the ration to X mg / kg, and suggest using prebiotic Y, which is expected to improve the absorption efficiency by Z% and reduce the diarrhea rate by W%; wherein the specific values of the threshold value T1, the value range A, the value range B, the parameters X, Z and W and the type of Y are dynamically determined based on statistical analysis of a historical data set and case-based reasoning results; an output response module connected with the adaptive decision support module, used for visualizing the structured report and core evaluation indexes through a graphical user interface. The system further comprises a model continuous learning module connected with the intelligent evaluation core module and the data storage module respectively, configured to periodically perform incremental training and optimization on the correlation dynamic evaluation model using newly added pig group multi-modal data and corresponding verification label data, so as to update the model parameters.

2. The system for assessing the correlation between the intestinal health of a pig and the absorption of trace elements according to claim 1, characterized in that, The input module comprises: a standardized data interface unit configured to receive the pig group multi-modal data from a production management record system, a feed formula database, an in-vitro detection device and an artificial observation record terminal; an edge computing unit deployed at a data acquisition terminal and configured to perform preliminary cleaning, redundancy removal and formatting processing on the initial data.

3. The system for assessing the correlation between the intestinal health of a pig and the absorption of trace elements according to claim 1, characterized in that, The output response module generates a visual interface comprising a correlation analysis dashboard, which displays the correlation strength between different trace elements and different intestinal health indicators revealed by the correlation dynamic evaluation model in the form of a heat map, and displays the historical changes of the core key performance indicators and the predicted trajectory based on the current decision suggestion in the form of a trend line.

4. The system for assessing the correlation between the intestinal health of a pig and the absorption of trace elements according to claim 1, characterized in that: The system further comprises a warning module connected with the intelligent evaluation core module and the adaptive decision support module respectively; The warning module is configured to automatically trigger a multi-level warning signal when the correlation score is lower than a preset threshold, and push the warning information and corresponding structured decision suggestions to the user terminal of a designated manager.

Images