R. Lacroix et al., EFFECTS OF DATA PREPROCESSING ON THE PERFORMANCE OF ARTIFICIAL NEURALNETWORKS FOR DAIRY YIELD PREDICTION AND COW CULLING CLASSIFICATION, Transactions of the ASAE, 40(3), 1997, pp. 839-846
The effect of data preprocessing on the learning ability of artificial
neural networks was investigated with regard to the impact of distrib
uting the input vectors uniformly with respect to the output categorie
s in the training data set. The analyses were performed for neural net
works dedicated to (1) dairy cow culling classification and (2) milk y
ield prediction. The two types of neural network used for culling clas
sification were backpropagation and learning vector quantization. For
yield prediction, backpropagation was used. The study was repeated wit
h several architectures for both types of network. Preprocessing of da
ta did not have a large impact on the general performance of the netwo
rks, but did affect the results for each output category. The effects
were more pronounced in the categories containing less frequent events
, for which the results always improved. For the categories with large
r number of records, balancing the data degraded the results. The resp
ective improvements and degradation of the results occurred for both p
rediction and classification, with the two types of neural networks, a
nd with all architectures tested. However the magnitude of the effects
varied with the type of neural network and with the architecture. The
results of this study indicate that, in general, the distribution of
outputs influences the learning process of neural networks for both ty
pes of application. The results also suggest that the types of output
distribution required for the training of neural nets may depend on th
e specifics of each problem.