A method on the weight of the Chinese term based on minimum classification error
Calculating the weight of feature words is a key step in text classification. A method, based on the traditional method-TFIDF, coupled with feature entropy, introducing minimum classification error to train a parameter of every word is proposed, and then is applied for spam sms filtering. By constructing Chinese sms base, experiments prove the efiectiveness of the method.
Author's Name: Cao, T., Zhong, S.
Volume: Volume 8
Issues: Issue 16
Keywords: Feature weight, MCE, Text representation