摘要：針對徑向基函數(shù)(Radial Basis Function,RBF)神經(jīng)網(wǎng)絡(luò)算法在無線網(wǎng)絡(luò)室內(nèi)定位中拓?fù)浣Y(jié)構(gòu)和網(wǎng)絡(luò)參數(shù)難以確定,其定位效果不理想的問題,提出了一種用核主成分分析的模糊C均值聚類算法(Fuzzy C-Means clustering algorithm based on Kernel Principal Component Analysis,KPCA-FCM)和模擬退火自適應(yīng)遺傳算法(Simulated Annealing adaptive Genetic Algorithm,SAGA)優(yōu)化RBF神經(jīng)網(wǎng)絡(luò)的無線室內(nèi)定位算法。首先利用KPCA對原始訓(xùn)練數(shù)據(jù)樣本進(jìn)行數(shù)據(jù)預(yù)處理,再通過KPCA-FCM算法計(jì)算出最優(yōu)聚類數(shù)目和聚類中心點(diǎn);其次將聚類數(shù)目和聚類中心點(diǎn)作為隱含層神經(jīng)元個(gè)數(shù)和中心值,創(chuàng)建RBF神經(jīng)網(wǎng)絡(luò),并將其網(wǎng)絡(luò)參數(shù)映射到SAGA算法中;再次由SAGA算法進(jìn)行網(wǎng)絡(luò)參數(shù)尋優(yōu),把最優(yōu)的解映射回RBF神經(jīng)網(wǎng)絡(luò);最后利用樣本數(shù)據(jù)對RBF神經(jīng)網(wǎng)絡(luò)進(jìn)行訓(xùn)練和測試,完成建立RBF神經(jīng)網(wǎng)絡(luò)算法模型。實(shí)驗(yàn)表明,在相同的環(huán)境中,所提算法比傳統(tǒng)RBF神經(jīng)網(wǎng)絡(luò)定位精度提高了48.41%。