Figure 3

Multiple location prediction for protein P . First, SVMs S V M₁,…,S V M _q are used to obtain the location indicator estimates ${\widehat{l}}_1^P,\dots ,{\widehat{l}}_q^P$. The Bayesian network classifiers C₁,…,C _q are then used to predict the actual location indicators $l_1^P,\dots ,l_q^P$. The Bayesian network classifiers use the location-indicator estimates as well as with inter-dependencies among the locations.