基于共同主成分的多元时间序列降维方法.pdfVIP

绗?8 鍗 绗? 鏈 鎺 鍒 涓 鍐 绛 2013 骞 4 鏈 Vol. 28 No. 4 Control and Decision Apr. 2013 鏂囩珷缂栧彿: 1001-0920 (2013) 04-0531-06 鍩轰簬鍏卞悓涓绘垚鍒嗙殑澶氬厓鏃堕棿搴忓垪闄嶇淮鏂规硶 鏉庢娆? 閮缓鑳? 鎯犳檽婊? 瀹嬮椋 (绌哄啗宸ョ▼澶у瑁呭绠＄悊涓庡畨鍏ㄥ伐绋嬪闄紝瑗垮畨710051) 鎽 瑕? 閽堝甯歌鐨勯檷缁存柟娉曢毦浠ユ湁鏁堜繚鐣欏鍏冩椂闂村簭鍒椾富瑕佺壒寰佺殑闂, 鍒嗘瀽浜嗕紶缁熶富鎴愬垎鍒嗘瀽(PCA) 鏂规硶鍦ㄥ 鍏冩椂闂村簭鍒楅檷缁翠腑鐨勫眬闄愭€? 鎻愬嚭涓€绉嶅熀浜庡叡鍚屼富鎴愬垎鍒嗘瀽鐨勫鍏冩椂闂村簭鍒楅檷缁存柟娉? 骞堕€氳繃浠跨湡瀹為獙姣旇緝浜嗕袱绉 鏂规硶鐨勯檷缁存湁鏁堟€у拰璁＄畻澶嶆潅搴? 瀹為獙缁撴灉琛ㄦ槑, 鎵€鎻愬嚭鐨勯檷缁存柟娉曡兘澶熶互鐩稿杈冨皬鐨勮绠椾唬浠? 鏇存湁鏁堝湴瀵瑰鍏冩椂 闂村簭鍒楄繘琛岄檷缁? 鍏抽敭璇岤 闄嶇淮锛涘鍏冩椂闂村簭鍒楋紱涓绘垚鍒嗗垎鏋愶紱鍏卞悓涓绘垚鍒嗗垎鏋愶紱璁＄畻澶嶆潅搴 涓浘鍒嗙被鍙耳 TP311 鏂囩尞鏍囧織鐮? A Dimension reduction method for multivariate time series based on common principal component LI Zheng-xin, GUO Jian-sheng, HUI Xiao-bin, SONG Fei-fei (Institute of Equipment Management and Safety Engineering锛孉ir Force Engineering University 锛孹i鈥檃n 710051 锛孋hina. Correspondent 锛歀I Zheng-xin 锛孍-mail 锛歭izhengxin 2005@163.com) Abstract: Existing dimension reduction method for multivariate time series can鈥檛 preserve their feature effectively. Therefore, the drawback of PCA method is analyzed, when it is used in MTS dimension reduction, and based on common principal component analysis, a dimension reduction method for multivariate time series is proposed. The computational complexity and the validity of dimension reduction are compared between different methods. The results of experiments show that the proposed method can reduce dimension effectively at comparatively low computational cost, and at the same time preserve most feature of multivariate time series. Key words: dimension reduction 锛沵ultivariate time series 锛沺rincipal component analysis 锛沜ommon principal component analysis 锛沜ompu