기계학습 모델을 이용한 성장단계별 사료 섭취량 예측

초록

This study aims to estimate the average feed intake (FI) of pigs at different growth stages, select variables through correlation analysisbetween parameters, and develop a machine learning-based regression model to predict pigs' feed intake (FI). The experiment was conductedover 93 days, from September 14, 2023, to December 15, 2023. Feed was provided twice a day at 09:00 and 17:00, with the amountgiven being 5% of the pigs' average body weight. The pigs' body weight (PBW) was measured daily at 09:00 using a portable pigscale. Temperature (RT), relative humidity (RH), and NH3 levels inside the barn were collected every 5 minutes using LivestockEnvironment Management System (LEMS) sensors. The growth stages were divided into three phases, referred to as GS1, GS2, andGS3. The average feed intake and standard deviation for each growth stage were calculated to analyze the significance and trends infeed intake across different stages. For model performance evaluation (, RMSE, MAPE), the data was split in an 8:2 ratio for accuracyvalidation. The results showed a significant difference in feed intake across the growth stages (p < 0.05), confirming that pigs consumea consistent amount of feed as they grow. Additionally, a strong correlation was found between FI and PBW (R > 0.94) during thecorrelation analysis. In terms of model performance evaluation, the Random Forest Regression (RFR) model demonstrated the highestaccuracy ( = 0.959, RMSE = 195.9, MAPE = 5.739)

키워드