Optimizing Feeding Schedule and Live-Weight Prediction for Native Chicken Based on Machine Learning

To meet market supply and demand, producers must accurately schedule processing dates to ensure optimal pricing. This study developed a practical feeding program system for local Taiwanese chicken breeds, including Guzao males, Huangjin females, and Red Junglefowl males. The system integrates daily predictions of cage-level body weight to guide each flock toward a target weight before the planned processing date. Four prediction models were evaluated, including random forest, XGBoost, Extra Trees, and an artificial neural network. The best-performing model was embedded into the system, and an Extra Trees model was used to estimate the total remaining ration and update daily feed allocations under standard feeding conditions. A validation experiment was conducted using a 54-day batch of Guzao males, during which cage-level data were collected. The feed conversion ratio of birds managed under the feeding program was compared with that of conventional feeding. The results provide preliminary support for the feasibility of a data-guided feeding program system with potential agricultural application value, although additional batches and cross-farm evaluations are needed to confirm generalizability and operational performance.