特征提取在提高分類(lèi)的準(zhǔn)確性中起著非常關(guān)鍵的作用.對(duì)時(shí)序特征提取的方法進(jìn)行歸納分類(lèi),將有利于對(duì)特征提取整體性,全面性的認(rèn)識(shí).回顧現(xiàn)有的時(shí)間序列中特征提取的方法,將其總結(jié)為四大類(lèi),它們分別是基于基本統(tǒng)計(jì)方法的特征提取、基于模型的特征提取、基于變換的特征提取、基于分形維數(shù)的特征提取.針對(duì)每一類(lèi)的特征提取方法,進(jìn)一步研究了它相應(yīng)的分類(lèi)方法和它在時(shí)間序列數(shù)據(jù)中的應(yīng)用鄰域.%The main contributions of this paper are: 1) The main feature extraction approaches are classified into four categories; 2) The main idea of each category is analyzed, the advantages and disadvantages are pointed out; 3) The guidelines of choosing suitable feature extraction approach is suggested.
展開(kāi)▼