This paper exploits the visual differences of five video genres, presents a combined model of editing, color, texture and motion features that could best distinguish one from the other, and uses the modified directed acyclic graph support vector machine (DAGSVM) model as the classifier. Experiment shows that: the features extracted have improved the identifiability of different genres, and computational complexity has been reduced; by introducing the DAG policy, the performance of the classifier has been enhanced; result demonstrates the precision and effectiveness of this approach, comparing with two other methods.