Non-English title

This study aimed to describe the cocoa beans behavior, and to predict the cocoa beans moisture content, during an intermittent microwave drying at 3 power levels (450 w, 600 w and 700 w). Experiments were carried out on fermented cocoa beans using a domestic microwave oven. The data obtained were adjusted using five semi-empirical models of drying on thin layer and a recurrent artificial neural network. Among the semi empirical models used, Page model was observed the most appropriate one for describing the cocoa beans behavior. For the various power levels, it presented respectively R super(2) of 0.9993, 0.9971 and 0.9967. The recurrent artificial neural network used (R super(2) > 0.999), presented a good ability to predict the moisture content of cocoa beans.