Analysis of York Pigs Feeding Behavior Using Stepwise Regression and Principal Component Regression

Xuelin FU Yajing CHEN Manting WU Junyong HU Wanghong LIU

Abstract A statistical analysis was conducted on the feeding behavior of 106 York breeding pigs. Pearson correlation analysis, principal component correlation analysis and multiple stepwise regression equation methods were applied to establish regression equations of the York breeding pigs total feed intake per time and average feed intake per time with corrected fat thickness, feed conversion rate, and corrected daily gain. The results showed that: ① there were three peak feed intake periods for the pigs, and the correlation coefficient between the feed intake and the corrected fat thickness of the pigs in the 24 h period was positive or negative, that is, increasing the number of feeding times and the feed intake was not necessarily conducive to the fat thickness accumulation, but the breeding goal of fat thickness could be achieved by controlling the feeding times and feed intake; ② the average feed intake of pigs in the 60-90 kg body weight stage was 30%-50% higher than that of the 30-60 kg body weight stage, but the number of feeding times  decreased, the peak feeding time was more concentrated, and the feeding duration per time was 3.0 min longer, indicating that as the weight of pigs increased, the feed intake increased significantly; and ③ the stepwise regression equations and the principal component equations showed that the feeding behavior of York pigs in the 30-90 kg growth stage was not only affected by the feeding time within 24 h, but also by environmental factors such as temperature and humidity. The feeding behavior of York pigs is a complex process of interaction between environmental factors and animal factors.

Key words Feed intake; Corrected daily weight gain; Feed conversion ratio; Corrected fat thickness; Stepwise regression; Principal component regression

Received: October 3, 2020  Accepted: December 23, 2020

Xuelin FU (1982-), female, P. R. China, inspector, master, devoted to research about animal breeding.

*Corresponding author. E-mail: whliu@mail.hzau.edu.cn.

Feeding is the most important behavior for animals to obtain their own nutrients. At present, the instrument for recording the feeding behavior of pigs is the automatic feeding system ACEMA64 system of France ACEMO. The main principle is to use an automatic feeding station, by which when a pig with an electronic sensor on its ears enters the feeding station, a receiver immediately recognizes its ear tag and transmits it to the computer to record its ear tag and measure the feed intake per time, and the feeding duration, time and weight of feed taken by the pig per time are recorded. There is also a kind of fully automatic feed intake equipment (F.I.R.E) produced by Osborne industrial company in the United States, which is mainly directed to the feeding of growing pigs at the 30-150 kg stage. Its working system consists of a feeding instrument and a computer workstation for collecting information. The feeding instrument is equipped with a radio device that can recognize electronic ear tags. When a pig enters the feeding instrument, the instrument will record the pigs ID, feeding time, feed intake and pig weight. For breeding pig rearing, we must first grasp the pigs feeding laws, such as the daily feed intake of the pigs, the daily number of feeding times, and the time required for feeding each time. Only on this basis can the feed be accurately supplied to achieve precise feeding. Mastering the rules of pig feeding can not only make the pigs fully eat, but also avoid feed waste and achieve better economic benefits. In this study, we collected the feeding data of 106 Duroc pigs, adopted stepwise regression and principal component regression analysis methods to establish models of pig feeding laws, with an attempt to analyze the relationships of pigs feeding with corrected fat thickness, corrected daily gain, and feed conversion rate.

Materials and Methods

Experimental design

The data of 106 Duroc breeding pigs measured by the Breeding Swine Quality Supervision and Testing Center, Ministry of Agriculture (Wuhan) were used. The weight range of test pigs was 25-30 kg at the beginning of the test, and 95-105 kg at the end. After the test pigs were fed in the isolation houses for one week for isolation and observation, they were transferred to and fed in the test houses, and the recording of the feeding data was started. The statistical time of the breeding pig test period was 80 d.

Feeding management and determination conditions

Enclosed houses with cement floor were adopted. Before the test, the determination station was thoroughly cleaned and disinfected, and the ACEMA64 system was overhauled, debugged, and verified. The test pigs were raised with pellet feed under conditions of free intake, free drinking, routine immunization, and the feed was changed after their body weight exceeded 60 kg.

Determination items

The ACEMA64 system, a fully automatic feeding system from France ACEMO, recorded the daily number of feeding times, the feeding duration per time, the feed time, and the feed weight of each pig in the 30-60 and 60-100 kg weight stages. The growth performance determination and calculation of related data were carried out in accordance with the requirements of NY/T822-2019 standard: Feed conversion rate=Total feed consumption (kg)/Total weight gain (kg); Daily weight gain during the determination period (g)=70×1 000/［Age with corrected weight up to 100 kg (d)-Age with corrected weight up to 30 kg (d)］; Target weight backfat thickness (mm) = Measured backfat thickness (cm) + ［Measured weight (kg)-Target weight (kg)］ ×Measured backfat thickness (cm)/(Measured weight-B), where the value of B in the formula referred to 6.1.3 in NY/T 822-2019.

Statistical methods

Data processing

(a) Unified number: The pig number was converted to digital number.

(b) Variable introduction: xi (i=1, 2…12), feed intake, feeding duration per time and total feeding times per day.

The 24-hour feed intake of each pig were divided into once every 2 h, namely 0-2, 3-4, 5-6, 7-8, 9-10, 11-12, 13-14, 15-16, 17-18, 19-20, 21-22, 23-24, that is to say, divided into 12 feeding stages. The statistics included daily feed intake in 12 stages, feeding duration per time and total number of feeding times per day. The feeding time of each pig was from the beginning to the end of the test, and the average time was about 80 d.

(c) Data professing: The feed intake (x) and the number of feeding times (n) of each pig in the 12 stages of the determination period were summarized to form ∑xi, ∑ni, i=1, 2,…12, and the average feed intake per time xi was calculated at each stage xi=∑xi/∑ni. The feed conversion rate, corrected daily weight gain and corrected fat thickness of each pig were calculated.

(d) Data statistics: Two methods, i.e., stepwise regression and principal component regression, were applied.

Results and Analysis

Recording of the feed time and feed intake of 106 pigs, observation of the peak time period and total feed intake of the pigs

The results showed that at the growth stage of 30-90 kg, the feed intake peaks of pigs were mainly concentrated in three time periods, namely, 7:00-8:00, 11:00-12:00, and 17:00-18:00; and the total food intake was between 10 000 and 30 000 g, as shown in Fig. 1. At the 30-60 kg stage, the feed intake peaks of pigs were mainly concentrated in three time periods, namely, 8:00-10:00, 11:00-12:00, and 14:00-19:00; and the total food intake was in the range of 7 000-18 000 g, as shown in Fig. 2. At the growth stage of 60-90 kg, the feed intake peaks of pigs were mainly concentrated in three time periods, namely, 7:00-8:00, 9:00-10:00, and 16:00-18:00; and the total feed intake was in the range of 13 000-30 000 g, as shown in Fig. 3.

Data statistics and correlation analysis and multiple regression analysis

The Pearson correlation analysis (pearson correlation coefficients) was performed on the total feed intake in the 12 time periods, and the results are shown in Table 1. The results of the stepwise regression F test, are shown in Table 2. The results of the multiple regression T-test analysis are shown in Table 3.

Establishment of regression equations by stepwise regression

The correlation coefficient matrixs eigenvalues and proportions (table 4) and eigenvectors (Table 5) were calculated to analyze the corrected daily weight gain and corrected fat thickness, the regression relationship between feed conversion rate and corrected fat thickness, and regression relationships of average total feed intake per time (total feed intake per time/number of feeding times) with corrected fat thickness, corrected daily weight gain and feed conversion rate.

The stepwise regression equations of the average feed intake per time with corrected daily gain, feed conversion rate, and corrected fat thickness were established, as shown in Table 6. The stepwise regression equations of the total feed intake per time with corrected daily weight gain, corrected fat thickness and feed conversion rate were established, as shown in Table 7.

Establishment of principal component regression equations by principal component and multiple regression methods

The average feed intake per time (total feed intake per time/number of feeding times) was used to establish principal component regression equations with feed conversion rate, corrected daily weigh gain and corrected fat thickness, respectively. The results are shown in Table 8, Table 9, and Table 10. The total feed intake per time was used to establish principal component regression equations with feed conversion rate, corrected daily weight gain and corrected fat thickness, respectively. The results are shown in Table 11, Table 12, and Table 13.

Conclusions and Discussion

Feed intake and feed conversion rate

It can be seen from Fig. 1 to Fig. 3 that there were differences in the average feed intake per time and feeding duration of pigs between the weight stages of 30-60 kg and 60-100 kg. The average feed intake per time was 30%-50% higher in the 60-90 kg body weight stage than the 30-60 kg body weight stage, but the number of feeding times was less, the feeding peaks were more concentrated, and the feeding duration per time was 3.0 min longer. It showed that as the weight of pigs increased, the feed intake increased significantly. From the principal component regression equation of the average feed intake per time with feed conversion rate and the principal component regression equation of the total feed intake per time with feed conversion rate, it can be seen that the coefficient of xi was positive or negative, indicating that the feeding of individual pigs was not always conducive to the improvement of feed conversion rate. The reason might be that the stomach needed time for digestion, nutrient absorption, and rest after feeding. This result is similar to that of Xu et al.［1］, who found in their study that the number of feeding times was slightly different between different pig breeds, but the feed intake of pigs at the growth stage of 60-90 kg was 62.2% higher than that of pigs of 30-60 kg, and the feeding duration was 2.4 min longer.

Factors affecting the accuracy of the variables and independent variables in the regression equations

The data recorded by the instrument showed the feeding behavior of the pig herd during the whole determination period. Some pigs did not develop good eating habits at night (22:00 to 6:00 the next day), that is, the pigs did not leave the instrument in time after feeding, but slept in the instrument, but the feeding duration recorded by the instrument was based on the time the pig entered and exited the instrument, so the time the pigs slept in the instrument was also counted into the feeding duration, and consequently, the instrument recorded a too-long feeding time, but the number of feeding times was unchanged, resulting in reduced accuracy of the variable (average feed intake per time = total feed intake per time/number of feeding times). However, from another point of view, the reduced accuracy of the variables affected by poor feeding habits of pigs could objectively reflect the characteristics of pigs feeding behavior during the fattening period: the number of feeding times and the feed intake were irregular, and the feeding duration was random.

This study was conducted from July to October. The weather in Wuhan is hot, and the environment is characterized by high temperature and humidity. Although the pig houses have wet curtain cooling, fan ventilation and other equipment, the feed intake, the number of feeding times and the time required for each feeding may be affected by heat stress. Studies have reported that environmental temperature has an important regulatory effect on the energy balance in animals, and in turn has an important impact on animal feed intake［2-3］. Therefore, if you collect data such as breeding pig feeding in other seasons, and then perform stepwise regression and principal component analysis, the accuracy of the regression equations may be improved.

The selected principal component regression equation factors and eigenvectors themselves were imperfect. The principal component factor was the feed intake (xi) in 12 time periods, without considering environmental factors such as temperature, humidity, and differences in digestive energy of individual pigs. Since the feeding behavior of pigs is the result of the interaction between the animal itself and the surrounding environment［4-7］, it is imperfect to only consider the feed intake of 12 time periods as the principal component analysis equation.

The significance of the regression equations to pig raising

The statistical data of the feeding time of pigs in the 24 h of a day was divided into 12 stages, that is, statistics was made every 2 h. The stepwise regression and principal component regression equations from Table 13 to Table 20 showed that the feeding of pigs at different times of the day had different effects on the corrected fat thickness, corrected daily gain and feed conversion rate. For example, in the regression equations, the xi coefficient was positive or negative, indicating that feeding in some stages of the 12 feeding stages had an accumulative effect on corrected fat thickness, such as the positive coefficient xi, which was corresponding to the feeding stage ti. Similarly, negative coefficients indicated that the feeding in the feeding stages had an unfavorable effect on corrected fat thickness, such as the negative coefficient xj, which was corresponding to the feeding stage tj. When the breeding goal is lower fat and lower feed conversion rate, the pigs should be controlled to eat during the corresponding time period when the coefficient of xi is negative. Similarly, when the breeding goal is a higher corrected daily gain, the pigs should be controlled to eat during the corresponding time period when the coefficient of xi is positive.

For example, for Ysl-0 (average feed intake per time with feed conversion rate)=0.000 364 826x1+0.000 035 669x2-0.000 058 192x3+0.000 231 263x4-0.000 220 332x5-0.000 056 076x6-0.000 219 737x7+0.000 292 654x8+0.000 199 109x9+0.000 067 604x10-0.000 151 360x11-0.000 216 461x12+1.975 54, the positive independent variable coefficients were x1, x2, x4, x8, x9, and x10, and the negative independent variable coefficients were x3, x5, x6, x7, x11, and x12. The equation showed that when the pigs ate at 5-6 oclock, 9-14 oclock, 22-24 oclock, etc., lower fat thickness and lower feed conversion rate could be obtained, and when the pigs ate at 0-4 oclock, 7-8 oclock, 15-16 oclock, and 21-22 oclock, pigs could get a higher daily gain. These regression equations can provide certain reference opinions for pig breeders in pig feeding.

References

［1］ XU RH, HU JP, WENG JQ, et al. Observation of swine feeding laws［J］. Swine Production, 2007(2): 18. (in Chinese)

［2］ YANG GM. Research progress on swine feeding behaviors［J］. Swine Industry Science, 2006(3): 14-16. (in Chinese)

［3］ WANG FY. Feed factors affecting pig feed intake［J］. Henan Journal of Animal Husbandry and Veterinary Medicine, 2009, 30(9): 27. (in Chinese)

［4］ LIU QH, CHENG W. Cause analysis and countermeasures of pig feed intake changes［J］. china Animal Husbandry & Veterinary Medicine, 2004, 31(9): 21-23. (in Chinese)

［5］ GAO HY, LI ZJ, DONG JH, et al. Research on comprehensive application technology to improve pig feed intake［J］. Animal Science & Ueterlnary Medlne, 2005(2): 42-43. (in Chinese)

［6］ ZHU L, LUO FZ, TIAN Z. Observation and research on feeding behaviors of York growing pigs［J］. Hunan Feed, 2010(3): 27-28. (in Chinese)

［7］ TAN XQ, LUO FZ, ZHU L. Study on feeding behaviors of York growing pigs［J］. Animal Husbandry & Veterinary Medicine, 2011,43(3): 45-47. (in Chinese)