Leveraging artificial intelligence and evolutionary algorithms for optimising cow supplementation and milk production

Efficient allocation of grain-based concentrate is essential for maximising milk yield and improving profitability in dairy farming. This study optimised concentrate allocation for dairy cows by integrating machine learning with evolutionary algorithms to enhance milk yield. Data from 1105 cows were analysed to model milk yield using predictor variables including breed, days in milk, stage of lactation, lactation number, daily concentrate allocation and climatic factors. For optimisation, records for 165 cows within a 30-day window were identified using a Genetic Algorithm. Three machine learning models (Gradient Boosting Machines, Extreme Gradient Boosting and Random Forest) were evaluated, with GBM showing superior predictive accuracy (R 2 = 0.75 training, 0.61 testing; RMSE = 3.02 L/cow·day −1 training, 3.47 L/cow·day −1 testing). The pre-trained GBM was integrated into four evolutionary algorithms (NSGA-II, SPEA-II, SMS-EMOA and RVEA) to simultaneously maximise herd-level daily milk yield and minimise deviations in daily concentrate allocation. Concentrate levels per cow were constrained between 6 and 11 kg per day, with adaptive bounds allowing no more than a 2 kg change per cow per day from the previous allocation. Without increasing herd-level supplementation allowance, statistical analysis based on 10 independent runs per algorithm revealed NSGA-II achieved the highest mean milk yield increase of 8.64 ± 0.10%, significantly outperforming SPEA-II (7.94 ± 0.13%), RVEA (7.41 ± 0.10%), and SMS-EMOA (6.63 ± 0.24%). One-way ANOVA confirmed significant differences among all algorithms (F = 303.24, p < 0.001), with NSGA-II also demonstrating superior computational efficiency (74.9 ± 0.7 s per day). These findings demonstrate the potential of combining machine learning and evolutionary algorithms to optimise feed use efficiency, highlighting promising practical applications of individualised feeding approaches to enhance milk production in dairy systems.