One of the great challenges today is to index videos with high-level semantic concepts or features. The basis of our approach is to use a fuzzy decision trees (FDT) to construct the heart of the system in order to reduce the need of human usage in the process of indexation. But when we address large, unbalanced, multiclass data sets, a single classifier - such as the FDT - is insufficient. Therefore we study the use of forests of fuzzy decision trees (FFDT): (a) its effectiveness for a high level feature detection task and (b) the effect on performance from number of classifiers point of view.