Aim:

The aim of this work was to estimate the efficiency of milk production in 1 168 cases in Ecuadoran Sierra Sur Andina, with the implementation of neural networks with multilayer perceptrons.

Materials and Methods:

These cases were collected from secondary samples provided by the Official Institute of National Statistics of Ecuador, in 2016. The variables chosen for the model were total milk production on the previ-ous day (P), as dependent variable; and total cattle heads (CH), total laborers in the field (E), and total surface at-tended by laborer (S), as independent variables. The selection criteria were the existence of data from individual cas-es, and their impact on the dependent variable.

Results:

The average efficiency was 8.11 %, from which the total cases detected efficiently (> 0.70) accounted for 11 (0.9 % of the sample). Later, the cases studied were classified into three groups, depending on the efficiency calculated: Group 1 (≤ 0.4 efficiency); Group 2 (> 0.4 - ≤ 0.7 efficiency); and Group 3 (> 0.7 efficiency).

Conclusion:

A comparison produced several statistical differences (P < 0.01) for variables total milk production/year on the farm, total field laborers, farm size, total cows, total cattle heads, calvings, pregnant cows, and served cows.