Milk production in dairy cattle is affected by linear and non-linear interactions between genetic and environmental effects. While, conventional methods are based on linear relationships, Artificial Neural Network system also considers non-linear relationships between parameters. In many countries, analysis of milk traits for 305 days in lactation is the foundation of dairy cattle genetic evaluations. Thus, a mathematical model for prediction of second parity milk yield and fat percentage, with the use of first parity information seemed to be helpful as a tool for predicting the yield of prospective producing cows. In this study, a back propagation neural network and multiple linear regression methods were compared based on their prediction differences with observed values. Data from 4 large sized dairy farms in Isfahan were used. From 1880 available records of first and second parities, 1850 records were used for training, testing and evaluation a back propagation artificial neural network system and multiple linear regression model, and 30 randomly chosen records for simulation. The obtained results of the simulation showed that artificial neural network with lower RMSE (817.84 0.336 for milk yield kg and fat% respectively) and S.D. ratio (0.576 0.106 for milk yield kg and fat% respectively) than multiple linear regressions with RMSE (933.93 0.350 for milk yield kg and fat% respectively) and S.D. ratio (0.697 0.544 for milk yield kg and fat% respectively) and also higher adjust coefficient of determination (36.74 72.88 for milk yield kg and fat% respectively ) than multiple linear regression (32.70 40.35 for milk yield kg and fat% respectively) presented the better result . Fitness of both ANN and MLR for %MF relative to MY, which showed that the employed input variables, may be ca good explain the change in %fat than milk yield kg. And also this result shows that A are reliable as a decision support system that helps breeders to choose a cow to be left or culled from herd than MLR. Selection of a proper sample is very important for training an ANN, The efficiency of A will be more improved when samples and variables which are more relevant to the output variables are used, Increase in response accuracy of network with adding input and output variable with high correlation. Thus a flexibility of this method relative to MLR is that it can be further developed for health, fertility, lifetime and other economical traits in dairy industry. In MLR Adding new data requires a new statistical model, whereas a neural network system can update itself with new data and ANN can be improved with more additional input vari