Artificial Neural Networks are processing computer techniques that use a mathematical model that is capable to acquire knowledge through the experience. The intelligent behavior of the network come from the interactions among units of processing that are denominated artificial neurons. The objective of this work was elaborate a neural network capable to foresee the stability of vegetable oils by using data of chemical compositions of the oils, by seeking a model to forecast the shelf-life of vegetable oils, by taking as parameters just some data of chemical compositions of the oils. The first step to develop a neural network was based on the collection of the data related to the problem, and the separation of these data in one training set and another set for testing. The variables of these groups were data of chemical composition, that included the total in fatty acids, phenols, tocopherols, and the individual composition in fatty acids. On the following step was done the training, where the input pattern presented to the network, as training parameter, was the peroxide index, that was experimentally obtained under absence of light at 65ºC, during the period of 16 days. Following it was tested the network capacity of forecast that was acquired during the training process, in relation to the parameter of stability chose, by using a new group of oils. In the final step, the linear correlation was determined among the values of stability foreseen by the network and those determined experimentally. Through the obtained results, it can be confirmed, by the neural network, the viability of forecast the stability of vegetable oils, by using data of chemical composition of the oils, by using the peroxide index as parameter of stability.
Neural networks; Vegetable oils; Stability
