DEVELOPMENT OF BROILER FEED FORECASTING MODEL BASED ON MACHINE LEARNING APPROACH

• Main Author: Rasyid Pramusakti, Adrishta
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/68657
• Institution: Institut Teknologi Bandung

Cikareo Poultry Farm is a poultry farm that carry out the proses of nurturing broiler chickens from day-old-chick (DOC), that were received from a supplier partner, until they are ready to be harvested and sold back to the supplier partner. This poultry farm nurtures 40.000 to 50.000 chickens each nurturing period and carry out 6 periods each year. In the nurturing process, improvements can be made regarding the feeding process. In each period, loss from feed overstock or overfeeding is experienced by the farm. The objective of this research is to identify the feeding pattern in order to minimize the loss received from overfeeding or feed overstock in the farm using machine learning. Two models are developed in this research, which are chicken growth curve prediction model and feeding prediction model. Chicken growth curve prediction was done using regression analysis model which was done using four empirical equations, which were linear, second-order polynomial, third-order polynomial, and Gompertz curve equations. Feeding prediction modelling was done using recursive linear model, which was tested regarding the combination of window size and prediction horizon that can produce the best predictive value. Based on the calculations that were made, it was determined that the best growth curve prediction model was the Gompertz curve regression and the best feeding prediction model was the combination 7-day window size and 1-day prediction horizon. The performance measure values produced were MRPE value of 5.483 and R2 value of 0.991 for the growth curve prediction model and MRPE LOOCV value of 6.766 and R2 LOOCV value of 0.966 for the feeding prediction model. Calculations were also made on the amount of loss that could be saved and resulted an average of Rp957,263 saved every period and a total of Rp30,632,400 saved through 16 periods from two farmhouses.