To optimize the proposed prediction model
ork, we need to intensively select the input fea
arning methods. In the case of features, the cu
features is 12. Only the most significant subse
atures might be optimal for a particular learnin
itable way to select a subset of significant featu
nimum classification error (MCE) algorithm
plied. Apart from the current features, one o
atures is the window size of POS sequences a
oven to highly affect the performance of p
ediction for English [1]. If the window s
quences is suitable, the learning method can
ucial information from the input text.