Method for evaluating stability of pig feeding behavior based on circadian rhythm and seasonal shift and application thereof

Abstract

This invention discloses a method for assessing the stability of pig feeding behavior based on diurnal rhythms and seasonal shifts, and its application. The method includes the following steps: Step S1, data collection of the experimental group and original feeding time series; Step S2, quality control and correction of the original feeding time series; Step S3, construction of a 24-hour diurnal feeding rhythm quantitative index; Step S4, quantification of monthly seasonal feeding hotspot shifts; Step S5, breed feeding strategy classification and labeling; Step S6, construction of a comprehensive feeding behavior stability scoring model. The application of this invention is to screen highly stable feed-saving boars or optimize pig farm feeding programs. This stability assessment method integrates two dimensions—24-hour diurnal feeding rhythm and annual monthly seasonal feeding shifts—to complete a quantitative evaluation of stability. It can be used for marker-assisted breeding of feed-saving boars and can also guide large-scale pig farms to adjust daily feeding strategies according to seasonal feeding patterns, achieving precision feeding.

Description

Technical Field

[0001] This invention belongs to the fields of precise phenotypic determination of livestock and poultry, data quality control of intelligent breeding, and animal genetic breeding technology. Specifically, it relates to a method for assessing the stability of pig feeding behavior based on diurnal rhythm and seasonal shift, and its application. Background Technology

[0002] Feed costs account for 60% to 70% of the total cost of pig farming. Remaining feed intake and feed conversion efficiency are core selection indicators for pig breeding. The diurnal fluctuations in pigs' daily feed intake and seasonal shifts in feeding time directly affect feed utilization efficiency. Current industry technologies have several significant shortcomings: First, when using commercial electronic feeding equipment to collect feeding data, the same feeding session can be split into multiple records due to repeated ear tag recognition. Additionally, weight sensor drift and abnormal feeding times can introduce a large amount of invalid noise data. Existing conventional analysis methods simply remove extremely large or small values; there is no standardized end-to-end data processing solution. The raw feeding data, without standardized correction, directly distorts the feeding rhythm and feeding time statistics.

[0003] Second, current evaluations of foraging behavior only use static averages such as the average number of feedings per day and the average weight of food consumed per feeding. They do not quantify the fluctuations in foraging time throughout the day or the deviations in foraging time under different temperature conditions throughout the year, making it impossible to accurately distinguish the stability of foraging among different individuals. The industry lacks a unified numerical standard for judgment, relying on manual experience to classify the foraging patterns of different species, which is highly subjective and has poor repeatability.

[0004] Third, there is currently no heritable comprehensive stability scoring system that simultaneously incorporates diurnal and seasonal fluctuations, making it impossible to develop quantifiable breeding indicators. The lack of standardized scoring hinders large-scale breeding of pigs.

[0005] Fourth, existing technologies have not formed a complete closed-loop operation process from raw data cleaning, rhythm quantification, seasonal shift calculation, feeding pattern classification, comprehensive scoring to breeding / feeding application, and cannot be adapted to large-scale genetic improvement work involving tens of thousands of animals.

[0006] In summary, existing technologies have multiple shortcomings, including low standardization of raw feed data cleaning, lack of quantitative evaluation standards for feed fluctuations, subjective classification of breed feed patterns, and lack of a heritable comprehensive scoring system. These shortcomings limit the breeding of feed-saving pigs and the implementation of refined feeding practices in pig farms. Summary of the Invention

[0007] The purpose of this invention is to overcome the shortcomings of existing electronic feeding data processing methods, such as single error identification, low correction accuracy, easy loss of breed biological characteristics, and inability to adapt to large-scale breeding populations. This invention provides a method for assessing the stability of pig feeding behavior based on diurnal rhythms and seasonal shifts, and its application. This stability assessment method relies on long-term feeding time-series data collected by fully automated electronic feeding stations, and integrates two dimensions: 24-hour diurnal feeding rhythm and annual monthly seasonal feeding shifts to complete a quantitative evaluation of stability. It can be used for marker-assisted breeding of feed-saving boars and can also guide large-scale pig farms to adjust daily feeding strategies according to seasonal feeding patterns, achieving precision feeding.

[0008] To achieve the above objectives, this invention provides a method for assessing the stability of pig feeding behavior based on diurnal rhythms and seasonal shifts, comprising the following steps: A method for assessing the stability of pig feeding behavior based on diurnal rhythms and seasonal shifts includes the following steps: Step S1: Data collection of the experimental population and the original feeding time sequence; Complete raw feeding time data of boars to be evaluated throughout the year were collected, excluding records of the first three days of acclimatization. The raw data included individual ear tags, feeding start time, feeding end time, single feed intake, weight at feeding time, and age on the day of feeding. Step S2: Quality control and correction of the original feeding sequence; Multi-level quality control and missing weight imputation were carried out on the original feeding time sequence, and the corrected single feeding weight was calculated using a multivariate linear correction formula to output a standardized feeding dataset. Step S3: Construct quantitative indicators for the 24-hour diurnal feeding rhythm; The daily feeding time sequence was divided into twenty-four one-hour windows. The relative feeding intensity of each window was calculated, the diurnal peak index was extracted, and the diurnal dispersion was calculated. Step S4: Quantification of monthly seasonal feeding hotspot shift; A nuclear density heatmap of feeding frequency is plotted monthly, and the monthly offset duration, annual average offset duration, and seasonal dispersion are calculated based on the average feeding time in March and April. Step S5: Classification and labeling of cultivar foraging strategies Based on the average number of feedings per day, the amount of food consumed per feeding, and the feeding rate, three types of feeding strategies are identified: frequent feedings, small feedings, and rapid feeding. Step S6: Construction of a comprehensive stability scoring model for feeding behavior; A weighted comprehensive stability scoring model was constructed to calculate the individual's comprehensive feed intake score and classify the stability level.

[0009] In the preferred technical solution, the refined quality control and missing data filling correction of the original feeding sequence in step S2 specifically includes the following process: Step S21: Merging feeding segmentation records: When the corresponding adjacent feeding time difference is <45min and the adjacent feeding time difference is <single feeding duration, it is determined that the same feeding was repeatedly scanned and segmented by the device, and the corresponding multiple segmentation records are merged as a single complete feeding. Step S22: Identify abnormal feeding records: Identify abnormal feeding times (too long or too short), feeding amounts (too large), and feeding rates (too small) using the threshold identification method; if abnormal records are found, mark all feeding times for that day as missing. Step S23, Robust weight regression denoising: Construct a robust regression model using age in days and age squared. Weight records with a model weight < 0.5 are identified as abnormal and marked as missing. Linear interpolation is used to fill in the missing weights. S2.4 The single feeding amount is corrected using an interactive regression model: Corrected single feeding weight = constant intercept of regression equation + weight correction coefficient × weight at feeding time + feeding duration correction coefficient × single feeding duration + interaction term correction coefficient × product of weight at feeding time and single feeding duration + random residual.

[0010] In the preferred embodiment, the formula for calculating the relative feeding intensity of the window in step S3 is: Window relative feeding intensity = window feeding index value ÷ daily average of corresponding index; diurnal dispersion is the standard deviation of the relative feeding intensity of all twenty-four one-hour windows, and the larger the value, the stronger the daily feeding fluctuation.

[0011] In the preferred technical solution, the monthly feeding hotspot in step S4 is the time corresponding to the highest density in the monthly kernel density map, the monthly feeding offset duration is the absolute value of the time difference between the monthly feeding hotspot and the baseline time in March and April; the annual average offset duration is the sum of the monthly offset durations of the twelve months ÷ twelve; and the seasonal dispersion is the standard deviation of the monthly offset durations of the twelve months.

[0012] The preferred technical solution, step S6, is as follows: The discrete indicators are converted by normalization formula, and then the individual comprehensive stability score is calculated and the stability level is divided by weighted comprehensive scoring formula. The normalization calculation formula is: Normalized score = (maximum value of group index - individual index value) ÷ (maximum value of group index - minimum value of group index), and substituting the day-night dispersion and seasonal dispersion respectively, we can obtain a positive score of 0 to 1.

[0013] The preferred technical solution uses a weighted comprehensive scoring formula to calculate the individual's comprehensive stability score and classify stability levels. The comprehensive stability score calculation formula is as follows: Comprehensive stability score = 0.6 × diurnal normalized score + 0.4 × seasonal normalized score; Classification criteria: Comprehensive stability score ≥ 0.85 is high stability, 0.60 ≤ comprehensive stability score < 0.85 is medium stability, and comprehensive stability score < 0.60 is low stability.

[0014] In the preferred technical solution, the three feeding strategies in step S5—feeding in small amounts frequently, feeding in small amounts frequently, and rapid feeding—are specifically as follows: Frequent but infrequent feeding: The total number of feedings per day is ≤8 times / day, the average weight of each feeding is ≥0.35 kg, and the duration of each feeding is ≥9 minutes; Frequent small feedings: The total number of feedings per day is ≥10 times / day, and the average weight of each feeding is ≤0.28 kg; Rapid feeding: 8 times / day < total daily feeding frequency < 10 times / day, single feeding rate ≥ 0.044 kg / minute.

[0015] Another objective of this invention is to provide an application of a method for assessing the stability of pig feeding behavior based on diurnal rhythms and seasonal shifts, specifically: applying it to screen highly stable feed-saving boars or optimize pig farm feeding programs; it also includes: Step S7: Genetic association analysis; Genetic association analysis was performed between stability scores and feed efficiency traits to screen for highly stable feed-saving boars or to optimize pig farm feeding programs based on seasonal shifts.

[0016] In the preferred technical solution, step S7 uses the restricted maximum likelihood method to calculate the genetic correlation value between the comprehensive stability score and the remaining feed intake weight and feed conversion efficiency. Individuals with a genetic correlation value ≤ -0.3 are judged to be excellent breeding individuals.

[0017] The preferred technical solution is that step S7 optimizes the pig farm feeding plan based on seasonal shifts as follows: the monthly feeding shift time is ≥120 minutes from June to September each year, the afternoon feeding is delayed by 1.5 to 2 hours, and the total amount of feed per feeding is increased by 15% to 20%.

[0018] The present invention, by adopting the above technical solution, has at least the following beneficial effects: (1) The method of the present invention can control the total error of the original feeding data within 5.20%, and the accuracy of the weight loss interpolation filling is stable at 99.90%; the determination coefficient of the multivariate correction model of feeding amount is in the range of 0.91~0.94, which eliminates the temporal noise caused by the electronic feeding equipment to the greatest extent.

[0019] (2) This invention constructs dual quantitative indicators of diurnal dispersion and seasonal dispersion to quantify the fluctuation of daily feeding time and the shift in feeding time caused by seasonal changes throughout the year. The magnitude of the fluctuation can be intuitively quantified by the standard deviation. By using two sets of quantitative indicators, diurnal dispersion and seasonal dispersion, the evaluation standard is unified and highly repeatable, completely eliminating the subjective bias of manual classification. Fixed numerical thresholds verified by group experiments are set for Duroc feeding less frequently and more often, Large White feeding less frequently and more often, and Landrace feeding rapidly. This enables automatic classification by computer, completely eliminating the drawbacks of relying on manual experience to classify feeding patterns. The evaluation standard is unified and repeatable, and different operators and different pig farms can obtain consistent judgment results.

[0020] (3) The comprehensive stability score of feed intake was obtained through a normalized formula and a diurnal and seasonal weighted scoring formula. This score showed a significant negative genetic correlation with the remaining feed intake and feed conversion efficiency, with the genetic correlation stable in the range of -0.31 to -0.43. Individuals with higher stability have better feed utilization efficiency and can be directly used as a specific selective trait for molecular marker-assisted breeding, realizing the targeted breeding of stable feed-saving boars under cold and heat stress, and helping to improve feed-saving pig breeding. The heritability of the comprehensive stability score showed a significant negative genetic correlation with the remaining feed intake and feed conversion efficiency, and can be directly used as a standardized auxiliary selection indicator for molecular marker-assisted breeding.

[0021] (4) The entire method supports two major application directions: on the one hand, it can be used for large-scale genetic evaluation of breeding pigs and screening for high-quality sires in large-scale breeding farms; on the other hand, it can accurately adjust the feeding time and amount in summer based on the monthly feeding hotspot shift. Field trials have confirmed that after optimizing the feeding according to this scheme, the feed conversion rate of pig farms decreased by 0.04, and the daily fluctuation of feeding decreased by 12.6%, which can effectively reduce the cost of breeding feed and take into account both the scientific research and breeding value and the actual economic benefits of large-scale pig farms. Attached Figure Description

[0022] Figure 1 This is a flowchart illustrating a method for assessing the stability of pig feeding behavior based on diurnal rhythms and seasonal shifts, as proposed in this invention.

[0023] Figure 2 is a scatter plot of the distribution of pigs' original weight versus age recorded by the automatic feeding system; where, Figure 2 (a) is a scatter plot of the distribution of original body weight and age of pigs recorded by an automated feeding system with normal growth trend; Figure 2 (b) is a scatter plot of the distribution of raw body weight and age of pigs recorded by an automated feeding system with missing data records; Figure 2 (c) is a scatter plot of the distribution of raw body weight and age of pigs recorded by an automated feeding system with random outliers; Figure 2 In the middle (d), there is a scatter plot of the distribution of the original weight and age of pigs recorded by an automated feeding system with data stratification.

[0024] Figure 3 is a scatter plot comparing abnormal and normal weight records at the feeding station; among them, Figure 3 (a) is a scatter plot of pig age and weight at the same measuring station when the station is operating normally and the records are stable. Figure 3 (b) is a scatter plot of age and weight of pigs at the same measuring station when the group synchronization is abnormal due to mechanical failure at the measuring station.

[0025] Figure 4 The 24-hour feeding rhythm distribution curves for three varieties: Duroc, Big White, and Changbai.

[0026] Figure 5 This is a heatmap of feeding frequency and nuclear density for the twelve natural months of the year. Detailed Implementation

[0027] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the invention as detailed in the appended claims.

[0028] Example 1

[0029] like Figure 1 As shown, a method for assessing the stability of pig feeding behavior based on circadian rhythms and seasonal shifts includes the following steps: Step 1: Collect complete raw foraging timeline data for the entire year; The experimental animals were selected from Duroc, Large White, Landrace or mixed breeds of boars for growth testing. Each boar was fitted with a unique electronic ear tag. The feeding details were continuously collected for twelve calendar months using the Ruibaole fully automatic feeding measurement equipment. The feeding records for the first three days after the experimental boars were transferred to the testing pen were excluded. The effective continuous feeding measurement days for a single experimental boar were no less than thirty days.

[0030] The basic information collected by the equipment includes five items: individual ear tag number, the time of entering the equipment for a single feeding, the end time of leaving the equipment for a single feeding, the original weight of the single feeding directly read by the equipment, the live weight of the pig at the moment of feeding, and the pig's age on that day.

[0031] Step 2: Refined quality control of the original feeding time sequence, missing data filling, and multivariate linear correction of feeding weight; Determination of split feeding records: If the interval between two consecutive feedings is less than 45 minutes and the interval is less than the duration of a single feeding, it can be determined that the feeding behavior has been split into multiple independent records by the device through multiple scans. These multiple split records need to be merged and organized into a complete feeding record.

[0032] Screening for Eleven Types of Abnormal Feeding Records: This identifies eleven types of data that do not conform to the normal physiological patterns of pig feeding. Specifically, it includes eleven sub-abnormalities derived from four core abnormalities: excessively long single feeding sessions, excessively short single feeding sessions, original single feeding weight exceeding the reasonable range, and abnormal feeding rates. If any feeding record on a single day is determined to be abnormal, all feeding time sequences for that calendar day are uniformly marked as missing data.

[0033] The eleven categories of abnormal feeding records are as follows: 1) The duration of a single feeding session is significantly shorter than the normal minimum feeding time for pigs, indicating an excessively short feeding time; 2) The duration of a single feeding session is significantly longer than the normal maximum feeding time for pigs, indicating an excessively long feeding time; 3) The original single feeding weight recorded by the equipment is significantly lower than the normal minimum feeding weight, indicating an insufficient feeding weight; 4) The original single feeding weight recorded by the equipment is significantly higher than the normal maximum feeding weight, indicating an excessively high single feeding weight; 5) The feeding rate calculated from a single feeding session is far lower than the minimum normal feeding rate for pigs, indicating an excessively slow feeding rate. The feeding rate calculated from a single feeding session is much higher than the maximum normal feeding rate of pigs, indicating that the feeding speed is too fast; multiple consecutive feeding records for the same pig within a short period show zero feeding weight, indicating no feeding but a feeding record is generated; the device simultaneously records two or more feeding details with completely overlapping start and end times within the same feeding period, indicating time duplication and conflict; the feeding end time is earlier than the feeding start time, indicating a reversed time logic; the interval between two adjacent feeding records is extremely short and the total feeding weight is significantly higher than the physiological limit for a single feeding session for pigs; the feeding duration value is normal, but the combination of feeding weight and feeding rate violates physiological common sense, indicating a serious mismatch between duration and feed intake.

[0034] Weight gap filling by linear interpolation: A two-way squared weighted robust regression model was constructed using the pig's age in days and the quadratic term of age in days. Weight values ​​with a model operation weight less than 0.5 were judged as abnormally missing. Linear interpolation was used to fill in all missing weight values.

[0035] Multivariate linear correction formula for single feed intake weight: Corrected single-feed weight = constant intercept + weight correction factor × instantaneous live body weight at feeding + feeding duration correction factor × duration of single feeding + interaction correction factor × product of instantaneous body weight at feeding and duration of single feeding + random residual Step 3: Quantify the 24-hour diurnal rhythm and calculate the diurnal dispersion. The entire 24-hour period from midnight to 12:00 each day is divided into 24 consecutive one-hour independent statistical windows. The total feed intake, total feed intake frequency, and total feed intake duration within each window are calculated. The relative feed intake intensity of a window is equal to the sum of the feed intake indicators for a single window divided by the total average value of the corresponding indicators for the entire day.

[0036] The peak feeding intervals were fixed, with the morning peak from 7:00 to 9:00 AM and the afternoon peak from 2:00 to 5:00 PM. Five rhythmic indicators were extracted in sequence: relative intensity of the morning peak, relative intensity of the afternoon peak, absolute value of the difference between the upper and lower peaks, duration of the morning peak, and duration of the afternoon peak.

[0037] Diurnal dispersion is calculated as follows: Diurnal dispersion is equal to the standard deviation of all relative feed intensity values ​​within a 24-hour window. The larger the diurnal dispersion value, the more significant the fluctuation in the daily feed intake time of a single pig, and the worse the diurnal stability of feed intake.

[0038] Step 4: Calculation of monthly feeding hotspot shift and seasonal dispersion; Foraging frequency kernel density heatmaps were plotted for each of the twelve natural months, and the time point with the highest foraging frequency density in each natural month was extracted as the foraging hotspot for that month. The arithmetic mean of the foraging hotspots in March and April of each year was selected as the benchmark for foraging time in the mild spring and autumn seasons.

[0039] Monthly feeding offset duration = the absolute value of the time difference between the peak feeding time of the month and the baseline feeding time in spring and autumn; The average annual offset duration = the sum of the offset durations of all twelve calendar months, divided by twelve. Seasonal dispersion is equal to the standard deviation of the monthly offset durations over the twelve calendar months. The larger the seasonal dispersion value, the greater the variation in pig feeding time during the transition between hot and cold seasons, and the worse the seasonal stability of feeding.

[0040] Step 5: Determine the fixed threshold for variety feeding patterns; Fixed judgment thresholds are set based on four indicators: total number of feedings per day, average weight of feedings per feeding, duration of feeding per feeding, and feeding rate per feeding. The three feeding patterns do not overlap. Frequent feeding pattern, suitable for Duroc boars: the total number of feedings per day is less than or equal to eight, the average weight of each feeding is greater than or equal to 0.35 kg, and the duration of each feeding is greater than or equal to nine minutes. Multiple small feedings mode, suitable for Large White boars: The total number of feedings per day is greater than or equal to ten, and the average weight of each feeding is less than or equal to 0.28 kg. Rapid feeding mode, suitable for Landrace boars: the average total number of feedings per day is greater than eight times and less than ten times per day, and the single feeding rate is greater than or equal to 0.044 kg per minute. Step Six: Normalization calculation and overall stability score calculation; Formula for forward normalization of discreteness: Normalized score = (Maximum value of the same indicator in the group - Individual indicator value) ÷ (Maximum value of the same indicator in the group - Minimum value of the same indicator in the group) The diurnal dispersion and seasonal dispersion of each individual are substituted into the formula for calculation, and finally a positive score in the range of zero to one is obtained; the lower the dispersion of the index, the higher the normalized score is calculated.

[0041] Overall stability weighted calculation formula: Overall stability score = 0.6 × day-night normalized score + 0.4 × seasonal normalized score The weight of the day / night dimension is fixed at 0.6, the weight of the seasonal dimension is fixed at 0.4, and the final range of the overall stable score is between 0 and 1.

[0042] Stability rating criteria: Individuals with a comprehensive stability score greater than or equal to 0.85 are classified as highly stable individuals; individuals with a comprehensive stability score greater than or equal to 0.60 and less than 0.85 are classified as moderately stable individuals; and individuals with a comprehensive stability score less than 0.60 are classified as lowly stable individuals.

[0043] Step 7: Two sets of implementation paths; Breeding screening application: Using GCTA software (a software for analyzing complex traits of the whole genome) with the restricted maximum likelihood method, the genetic correlation values ​​between the comprehensive stability score and the remaining feed intake and feed conversion efficiency were calculated. Individuals with genetic correlation values ​​less than or equal to -0.3 were identified as high-quality feed-saving breeding boars.

[0044] Application of pig farm feeding optimization: Based on the monthly feeding frequency kernel density heat map throughout the year, the monthly feeding offset time is extracted; from June to September each year, as long as the monthly feeding offset time in any month is greater than or equal to 120 minutes, the feeding time in the afternoon of the pig farm is delayed by 1.5 to 2 hours, and the total weight of a single feeding is increased by 15% to 20%.

[0045] Example 2

[0046] 1. All basic experimental conditions; The experimental site was the Chifeng National Core Breeding Farm for Pigs in Inner Mongolia. A total of 9,705 boars were used for growth testing, subdivided into three breeds: Duroc (3,097), Large White (3,871), and Landrace (2,737). All boars were fed a standardized growing boar diet with a uniform formula throughout the experiment, and the temperature and humidity of the rearing environment were routinely controlled. Complete feeding records were continuously collected for twelve calendar months using the Ruibaole fully automated feeding equipment, resulting in a total of 6 million original valid feeding records. A 50K pig SNP gene chip was used to complete genotyping of the entire herd for subsequent verification of genetic parameters.

[0047] Step 1: Complete raw data collection; All 9,705 boars were individually tagged with unique electronic ear tags. The system collected all five basic fields: ear tag number, start and end times of feeding, original single feed intake weight, instantaneous body weight at feed intake, and age. Feed intake data from the first three days of environmental adaptation for all individuals were uniformly excluded. The effective feeding duration for all individuals was greater than thirty days, with no individual failing to meet the effective duration requirement, and there were no large-scale missing data in the original data collection.

[0048] Step 2: Perform a complete set of quality control and feed intake calibration procedures item by item; The system traversed all 6 million original feeding details, processed the equipment split records according to the merging rules, and found that Duroc split records accounted for 2.893%, Big White split records accounted for 2.154%, and Long White split records accounted for 1.991%. All split records were merged.

[0049] Eleven types of abnormal feeding data were identified. If any abnormal record was found on a single day, all feeding time sequence markers for that day were missing. The total error of the original uncorrected data for the three varieties was calculated as follows: Duroc 5.198%, Large White 3.608%, and Changbai 3.872%.

[0050] Combination Figure 2 Weight age distribution pattern Figure 3 To investigate the characteristics of weight fluctuations in cases of abnormal missing weights, a robust regression model was constructed using age in days and the square of age in days. Weights with a weight less than 0.5 were identified as abnormally missing, and linear interpolation was used to fill in all missing weights. Five validation gradients were set up, and 10%, 20%, 30%, 40%, and 50% of the actual weight data were randomly covered to test the filling effect. The weight filling accuracy was ≥99.90% under all five gradients.

[0051] The cleaned and valid data were substituted into the multiple linear correction formula to calculate the corrected single-feeding weight; the model fit goodness was 0.93 for Duroc, 0.94 for Large White, and 0.91 for Long White. All subsequent feeding statistics for the entire population were calculated using the corrected feeding weight.

[0052] Step 3: Complete calculation of day-night dispersion; like Figure 4 The 24-hour circadian feeding rhythm distribution curves for the Duroc, Large White, and Landrace breeds are shown in the diagram. Based on these curves, each day was divided into 24 one-hour windows, and the relative feeding intensity was calculated for each window. Peak feeding times were uniformly defined as 7:00-9:00 AM and 2:00-5:00 PM. The peak morning intensity, peak afternoon intensity, peak difference, and duration of the two peak periods were calculated for each individual in the entire population. The diurnal dispersion was calculated for all individuals. Population averages: Duroc 0.082, Large White 0.147, and Landrace 0.121. The daily feeding fluctuation of Large White was significantly higher than that of the other two breeds.

[0053] Step 4: Complete calculation of seasonal dispersion; like Figure 5 As shown, based on the monthly feeding heatmap throughout the year, feeding hotspot times for each of the 12 months were extracted. The average of the hotspot times in March and April was used to obtain the baseline feeding time of 15:02 for spring and autumn. The monthly feeding offset duration was calculated, and the annual average offset duration and seasonal dispersion were calculated by summing the data for the twelve months. Duroc had an annual average offset of 71 minutes and a seasonal dispersion of 0.063; Large White had an annual average offset of 103 minutes and a seasonal dispersion of 0.117; and Long White had an annual average offset of 116 minutes and a seasonal dispersion of 0.132. The average monthly offset duration for all months from June to September in summer was 128 minutes, and the average offset duration for November to February of the following year in winter was 82 minutes.

[0054] Step 5: Batch determination of the entire group's foraging pattern; The system batch calculates the total number of feedings per day, the average weight of each feeding, the duration of each feeding, and the feeding rate for all individuals, and automatically classifies them according to fixed thresholds. 91.2% of the Duroc population met all the criteria for frequent, infrequent feedings; 94.7% of the Large White population met all the criteria for frequent, small feedings; and 89.5% of the Long White population met all the criteria for rapid feeding.

[0055] Step Six: Batch Calculation and Classification of Stable Scores; All individuals in the entire population were sequentially substituted into the normalization formula to perform forward normalization conversion of diurnal and seasonal dispersion, resulting in two sets of scores in the zero-to-one interval. These scores were then substituted into the comprehensive stability score formula, with weights of 0.6 and 0.4 to complete the weighted summation. The Duroc population's overall stability average score was 0.87, with 72.3% of individuals having a comprehensive score ≥ 0.85, indicating high stability. The average score for the Large White individuals was 0.69, with most exhibiting moderate stability. The average score for the Long White individuals was 0.65, with the highest proportion of moderately stable individuals. All individual classification results strictly corresponded to the score interval standards, with no misclassifications.

[0056] Step 7: Genetic verification and feeding trials; At the breeding level: Genetic correlations were calculated in batches using the restricted maximum likelihood method with GCTA software. The genetic correlation between the overall stability score and remaining feed intake was -0.37 to -0.43, and between the overall stability score and feed conversion efficiency was -0.31 to -0.39. All individuals with a genetic correlation of less than or equal to -0.3 were marked as superior breeding material. The average remaining feed intake of highly stable individuals in the entire population decreased by 0.15 kg per day, and the overall feed conversion efficiency decreased by 0.07.

[0057] Feeding aspects: During the summer months (June to September), if the monthly feed deviation was ≥120 minutes, a feed adjustment experiment was conducted at the pig farm; the afternoon feeding time was generally delayed by two hours, and the weight of each feeding was increased by 18%. After the experiment, the overall feed conversion efficiency of the entire farm decreased by 0.04, and the daily average feed intake fluctuation decreased by 12.6%.

[0058] Precise full-process assessment of a single Duroc individual: Individual DDCFJYA23604065, with a total of 362 days of complete and valid feeding records throughout the year. Complete original feeding details for all twelve months of the year were collected, and invalid feeding records from the first three days of acclimatization were manually removed. All fields, including ear tag, feeding start and end time, original feeding weight, feeding body weight, and age, were complete and without missing data. Twelve equipment-split feeding records were identified and merged; three abnormal weight records were identified and their weights were filled using linear interpolation; all valid feeding data were substituted into a multivariate linear correction formula to calculate the corrected single feeding weight. Twenty-four one-hour windows were used to calculate the window feeding intensity, and peak indicators were extracted. The final calculated relative intensity of the morning peak was 1.12, the relative intensity of the afternoon peak was 2.07, the peak difference was 0.95, and the diurnal dispersion of this individual was 0.071. The monthly feeding hotspots for each of the twelve months were compiled, and the monthly deviation duration was calculated for each month. The average deviation duration for the twelve months was taken as the annual average deviation duration of 64 minutes. The standard deviation of the twelve-month deviation duration was calculated, yielding a seasonal dispersion of 0.058. The individual's average daily feeding frequency was 6.8 times, with an average feeding weight of 0.39 kg per feeding and a single feeding duration of 9.6 minutes. All indicators met the Duroc criteria for frequent, infrequent feeding, and the feeding pattern was ultimately determined to be frequent, infrequent feeding. Substituting the diurnal dispersion and seasonal dispersion of this individual into the normalization formula, the normalized diurnal score was 0.91 and the seasonal score was 0.93. Substituting these into the weighted formula, the overall stability score was 0.6 × 0.91 + 0.4 × 0.93 = 0.918. The score is greater than 0.85, classifying the individual as a highly stable feeder. The individual's remaining feed intake was found to be -0.18 kg per day, with an average daily weight gain of 1.12 kg per day. The individual's comprehensive score and genetic correlation with feed efficiency traits met the breeding criteria of less than or equal to -0.3, which meets the requirements for excellent feed-saving breeding pigs. Therefore, it can be included in the core breeding group for breeding.

[0059] Application Example: Complete Field Trial of Pig Farm Feeding Program Basic experimental information: The experimental animals were two hundred three-way crossbred boars used for growth testing, fed a standardized growing boar diet throughout the experiment; the total experimental period was ninety days. The experimental guidelines are attached. Figure 4 The monthly feeding frequency kernel density heatmap throughout the year is used to extract the feeding offset duration for each month.

[0060] Preliminary data confirmation: The monthly feeding deviation time for the four natural months of June, July, August and September of the whole year was statistically analyzed. The deviation time for all four months was greater than or equal to 120 minutes, which met the triggering conditions for feeding adjustment.

[0061] Complete feeding adjustment plan: The original afternoon feeding time of 2 PM in the pig farm will be uniformly postponed to 4 PM, and the overall afternoon feeding time will be delayed by two hours; the total weight of feed per feeding for all experimental boars will be uniformly increased by 18%. During the trial period, the rest of the feeding environment, feeding frequency, and diet formula will remain completely unchanged.

[0062] Full-cycle performance testing: After the 90-day trial, the feed intake-related indicators of the entire farm were statistically analyzed. The average feed conversion efficiency of the entire farm decreased from 2.42 at the beginning of the trial to 2.38. The daily average feed intake fluctuation of the entire group decreased by 12.6% overall, and the feeding optimization effect was obvious.

[0063] The complete operation was verified by comparing and verifying the data: the monthly feed intake deviation data for November and December of the current year, and January and February of the following year were retrieved. The deviation duration for all four months was less than 120 minutes. The original 14 feeding times and the original feeding weight were kept unchanged. The feed conversion efficiency and feed intake fluctuation of the group in winter were not significantly improved. This proves that the deviation duration threshold setting of the present invention is reasonable and the feeding adjustment rules are effective.

[0064] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for assessing the stability of pig feeding behavior based on circadian rhythms and seasonal shifts, characterized in that, Includes the following steps: Step S1: Data collection of the experimental population and the original feeding time sequence; Complete raw feeding time data of boars to be evaluated throughout the year were collected, excluding records of the first three days of acclimatization. The raw data included individual ear tags, feeding start time, feeding end time, single feed intake, weight at feeding time, and age on the day of feeding. Step S2: Quality control and correction of the original feeding sequence; Multi-level quality control and missing weight imputation were carried out on the original feeding time sequence, and the corrected single feeding weight was calculated using a multivariate linear correction formula to output a standardized feeding dataset. Step S3: Construct quantitative indicators for the 24-hour diurnal feeding rhythm; The daily feeding time sequence was divided into twenty-four one-hour windows. The relative feeding intensity of each window was calculated, the diurnal peak index was extracted, and the diurnal dispersion was calculated. Step S4: Quantification of monthly seasonal feeding hotspot shift; A nuclear density heatmap of feeding frequency is plotted monthly, and the monthly offset duration, annual average offset duration, and seasonal dispersion are calculated based on the average feeding time in March and April. Step S5: Classification and labeling of cultivar foraging strategies; Based on the average number of feedings per day, the amount of food consumed per feeding, and the feeding rate, three types of feeding strategies are identified: frequent feedings, small feedings, and rapid feeding. Step S6: Construction of a comprehensive stability scoring model for feeding behavior; A weighted comprehensive stability scoring model was constructed to calculate the individual's comprehensive feed intake score and classify the stability level.

2. The method for assessing the stability of pig feeding behavior based on circadian rhythm and seasonal shift as described in claim 1, characterized in that, The detailed quality control and missing-filling correction of the original feeding sequence in step S2 specifically includes the following process: Step S21: Merging feeding segmentation records: When the corresponding adjacent feeding time difference is <45min and the adjacent feeding time difference is <single feeding duration, it is determined that the same feeding was repeatedly scanned and segmented by the device, and the corresponding multiple segmentation records are merged as a single complete feeding. Step S22: Identify abnormal feeding records: Identify abnormalities such as excessively long or short single feeding duration, abnormal single feeding amount, and abnormal feeding rate using the threshold identification method; If there are abnormal records, all feeding times for that day will be marked as missing; Step S23, Robust weight regression denoising: Construct a robust regression model using age in days and age squared. Weight records with a model weight < 0.5 are identified as abnormal and marked as missing. Linear interpolation is used to fill in the missing weights. S2.4 The single feeding amount is corrected using an interactive regression model: Corrected single feeding weight = constant intercept of regression equation + weight correction coefficient × weight at feeding time + feeding duration correction coefficient × single feeding duration + interaction term correction coefficient × product of weight at feeding time and single feeding duration + random residual.

3. The method for assessing the stability of pig feeding behavior based on circadian rhythm and seasonal shift as described in claim 1, characterized in that, The formula for calculating the relative feeding intensity of the window in step S3 is as follows: Window relative feeding intensity = window feeding index value ÷ daily average of corresponding index; diurnal dispersion is the standard deviation of the relative feeding intensity of all twenty-four one-hour windows, and the larger the value, the stronger the daily feeding fluctuation.

4. The method for assessing the stability of pig feeding behavior based on circadian rhythm and seasonal shift as described in claim 1, characterized in that, In step S4, the monthly feeding hotspot is the time corresponding to the highest density in the monthly kernel density map. The monthly feeding offset duration is the absolute value of the time difference between the monthly feeding hotspot and the baseline time in March and April. The annual average offset duration is the sum of the monthly offset durations of the twelve months divided by twelve. Seasonal dispersion is the standard deviation of the twelve-month monthly offset duration.

5. The method for assessing the stability of pig feeding behavior based on circadian rhythm and seasonal shift as described in claim 1, characterized in that, The process of step S6 is as follows: The discrete indicators are converted by normalization formula, and then the individual comprehensive stability score is calculated and the stability level is divided by weighted comprehensive scoring formula. The normalization calculation formula is: Normalized score = (maximum value of group index - individual index value) ÷ (maximum value of group index - minimum value of group index), and substituting the day-night dispersion and seasonal dispersion respectively, we can obtain a positive score of 0 to 1.

6. The method for assessing the stability of pig feeding behavior based on diurnal rhythm and seasonal shift as described in claim 5, characterized in that, The formula for calculating the overall stability score and classifying stability levels using the weighted comprehensive scoring formula is as follows: Overall stability score = 0.6 × diurnal normalized score + 0.4 × seasonal normalized score; Classification criteria: Overall stability score ≥ 0.85 is high stability, 0.60 ≤ overall stability score < 0.85 is medium stability, and overall stability score < 0.60 is low stability.

7. The method for assessing the stability of pig feeding behavior based on circadian rhythm and seasonal shift as described in claim 1, characterized in that, The three feeding strategies in step S5—feeding in small, frequent meals, feeding in small, frequent meals, and rapid feeding—are specifically as follows: Frequent but infrequent feeding: The total number of feedings per day is ≤8 times / day, the average weight of each feeding is ≥0.35 kg, and the duration of each feeding is ≥9 minutes; Frequent small feedings: The total number of feedings per day is ≥10 times / day, and the average weight of each feeding is ≤0.28 kg; Rapid feeding: 8 times / day < total daily feeding frequency < 10 times / day, single feeding rate ≥ 0.044 kg / minute.

8. The application of the method for assessing the stability of pig feeding behavior based on circadian rhythm and seasonal shift as described in any one of claims 1 to 7, characterized in that, It can be applied to screen highly stable feed-saving boars or to optimize pig farm feeding programs; It also includes: Step S7: Genetic association analysis; Genetic association analysis was performed between stability scores and feed efficiency traits to screen for highly stable feed-saving boars or to optimize pig farm feeding programs based on seasonal shifts.

9. The application of the method for assessing the stability of pig feeding behavior based on diurnal rhythm and seasonal shift as described in claim 8, characterized in that, In step S7, the restricted maximum likelihood method is used to calculate the genetic correlation between the overall stability score and the remaining feed intake and feed conversion efficiency. Individuals with a genetic correlation value ≤ -0.3 are considered to be excellent breeding individuals.

10. The application of the method for assessing the stability of pig feeding behavior based on diurnal rhythm and seasonal shift as described in claim 8, characterized in that, Step S7 optimizes the pig farm feeding plan based on seasonal shifts as follows: the monthly feeding shift duration is ≥120 minutes from June to September each year, the afternoon feeding is delayed by 1.5 to 2 hours, and the total amount of feed per feeding is increased by 15% to 20%.

Images