Milk fat yield and fatty acid (FA) content in lactating dairy cows can be greatly affected by dietary FA profile. The analysis of individual FA in feedstuffs has traditionally utilized gas chromatography (GC) techniques; however, this technique is costly, labor intensive, and time consuming. The objective of our study was to determine if artificial neural network (ANN) is capable of building improved near-infrared spectroscopy (NIRS) predictive models for feed FA relative to the current model approach using partial least squares (PLS). A total of 324 feed samples were scanned with an NIRS instrument and analyzed by GC (Michigan State University) for the following FA: C12:0; C14:0; C16:0; cis-9 C18:1; cis-9,cis-12 C18:2; cis-9,cis-12,cis-15 C18:3. Two predictive approaches were implemented: PLS and ANN, and all wavelengths were used as predictors in each approach. A random grid search was performed to define the best ANN architecture. Prediction quality was assessed by randomly splitting the data set into training and test sets (70% and 30% of the data set, respectively). Prediction quality of PLS and ANN were similar in terms of R² and Root Mean Squared Error Prediction (RMSEP), for C12:0 (PLS: R² = 0.46, RMSEP = 0.12%, ANN: R² = 0.42, RMSEP = 0.13%), C14:0 (PLS: R² = 0.70, RMSEP = 0.65%, ANN: R² = 0.74, RMSEP = 0.56%), C18:1 (PLS: R² = 0.88, RMSEP = 3.1%, ANN: R² = 0.88, RMSEP = 3.3%), C18:2 (PLS: R² = 0.90, RMSEP = 5.0%, ANN: R² = 0.89, RMSEP = 5.3%), and C18:3 (PLS: R² = 0.83, RMSEP = 6.8%, ANN: R² = 0.76, RMSEP = 8.6%). However, ANN predictions presented lower precision and accuracy for C16:0 (PLS: R² = 0.80, RMSEP = 2.93%, ANN: R² = 0.65, RMSEP = 0.74%) and C18:0 (PLS: R² = 0.58, RMSEP = 1.0%, ANN: R² = 0.44, RMSEP = 1.2%). Overall, results show that NIRS can predict feed FA with reasonable accuracy, but more accurate and precise for the major dietary FA C16:0, C18:1, C18:2, and C18:3. For this data set, the use of ANN did not improve prediction quality when compared with PLS.