焦瑋 楊雪寒 孟潔 張倩



摘 要: 為了利用電子醫(yī)療檔案實(shí)現(xiàn)對患者疾病的智能診斷,提出了一種結(jié)合模糊C均值聚類和區(qū)間二型小腦模型關(guān)節(jié)神經(jīng)網(wǎng)絡(luò)(FCM-IT2CMAC)的兩層分類算法。該算法使用了兩個(gè)分類器,其中小腦模型神經(jīng)網(wǎng)絡(luò)是主分類器,模糊C均值算法是預(yù)分類器。首先,使用預(yù)分類器將樣本數(shù)據(jù)分組,然后應(yīng)用主分類器確定樣本是否處于健康或患病狀態(tài)。此外還采用梯度下降法自適應(yīng)訓(xùn)練主分類算法的參數(shù),并使用李雅普諾夫穩(wěn)定性理論證明了算法的收斂性。最后通過實(shí)驗(yàn)證明該分類算法的有效性。
關(guān)鍵詞: 分類問題; 小腦模型神經(jīng)網(wǎng)絡(luò); 模糊C均值聚類算法; 醫(yī)學(xué)診斷
中圖分類號: TP391 ? ? ?文獻(xiàn)標(biāo)志碼: A
Abstract: In order to realize the intelligent diagnosis of patients diseases by using electronic medical files, this paper proposes a two-layer classification algorithm combining fuzzy C-means clustering and interval type II cerebellar model joint neural network (FCM-IT2CMAC). The algorithm uses two classifiers, in which the cerebellar model neural network is the main classifier and the fuzzy C-means algorithm is the pre-classifier. First, the sample data are grouped using a pre-classifier, and then the main classifier is applied to determine if the sample is in a healthy or diseased state. In addition, the gradient descent method is used to adaptively train the parameters of the main classification algorithm, and the convergence of the algorithm is proved by Lyapunov stability theory. Finally, the effectiveness of the classification algorithm is proved by experiments.
Key words: classification problem; cerebellar model neural network; fuzzy C-means clustering algorithm; medical diagnosis
0 引言
將數(shù)據(jù)分析算法應(yīng)用于電子醫(yī)療檔案的數(shù)據(jù)分析能夠?qū)崿F(xiàn)對是否患病的智能診斷。已有研究提出一些針對電子醫(yī)療數(shù)據(jù)集的數(shù)據(jù)二分類算法[1-3]。文獻(xiàn)[4]提出了一種基于決策樹模型的疾病診斷模型。文獻(xiàn)[5]提出了一種用于肝病早期診斷的神經(jīng)網(wǎng)絡(luò)分類算法。為此本文提出一種模糊C均值聚類算法(FCM)[6]和區(qū)間二型模糊小腦模型神經(jīng)網(wǎng)絡(luò)算法(IT2CMAC)[7]相結(jié)合的兩層醫(yī)療數(shù)據(jù)分類算法,以期實(shí)現(xiàn)基于電子醫(yī)療檔案的疾病準(zhǔn)確診斷。該算法在參數(shù)訓(xùn)練過程中,首先利用模糊C均值聚類算法將訓(xùn)練數(shù)據(jù)劃分為nc組,然后利用這些數(shù)據(jù)組訓(xùn)練區(qū)間二型模糊小腦模型神經(jīng)網(wǎng)絡(luò)算法。其中區(qū)間二型模糊小腦模型神經(jīng)網(wǎng)絡(luò)算法是基于區(qū)間二型模糊神經(jīng)網(wǎng)絡(luò)(IT2FNN)和小腦模型神經(jīng)網(wǎng)絡(luò)(CMAC)所提出的改進(jìn)分類算法,兼具兩種算法的優(yōu)點(diǎn)。……