38 2 2010 4 Journal of Fuzhou University Natural Science Vol 38 No 2 Apr 2010 1000-2243 2010 02-0213 - 06 MLP SVM 1 1 2 1 350108 2 350108 MIP SVM OA MLP - SVM TP391 72 A Research of dialectical classification on traditional Chinese medicine prescription for OA based on SVM combined with MLP LIAO Bin 1 YE Shao - zhen 1 ZHENG Chun - song 2 1 College of Mathematics and Computer Science Fuzhou University Fuzhou Fujian 350108 China 2 Fujian Academy of Integrative Medicine Fujian University of Traditional Chinese Medicine Fuzhou Fujian 350108 China Abstract The multi - layer feed - forward neural network and SVM methods were used to study the classification of traditional Chinese medicine which treats Osteoarthritis and get the classification of different drugs and the selection of key drugs At the same time combining the pattern classification rules to get the classification of the most similar drugs they can provide important basis datas for the study of compatibility and new herbal - drug for OA The comparison between classification results and clinical trials results shows that classification results are consistent with Chinese medicine theory Keywords neural network MLP - SVM pattern classification rules Osteoarthritis traditional Chinese medicine prescription 1 Osteoarthritis OA 2 OA 2009-05 - 05 1963 - E - mail yeshzh@ vip sina com 2009J01282 200710018
214 38 3 4 SVM OA 1 SVM MLP 1 1 4 1 Tab 1 1 Drug feature data quantization table 10 7 4 2 0-2 - 4-7 - 10 10 7 10 7 10 7 10 7 10 6 3 0 10 10 10 10 10 10 10 1 2 2 2 1 10 8 4 10 8 6 0 5 1 2 9 6 110 30 140 4 3 140 2 140 1 3 1 3 1 SVM SVM
2 MLP SVM 215 1 1 Fig 1 Quantitative result 1 3 2 MLP SVM 2 x = x 1 x 2 x d y = sgn ( N i = 1 α i y i K x i x + b ) 1 α i y i α i 2 Fig 2 SVM SVM model ψ Rd - > H H ψ x i ψ x j ψ x i K K x i x j = ψ x i ψ x j Mercer 7 8 MLP K x y = tanh κ x y - δ SVM 1 3 3 2 1 1 SVM 0 1 2 2 1 3 80% MLP SVM 2
216 38 3 Fig 3 Output result 2 2 SVM 9 10 2 2 1 2 2 2 4 0 5 0 5 1 0 33 0 33 0 33 0 4 Fig 4 Quantitative database 2 2 3 & k 2 1 2 3 r ij = m 槡 m & * k x * k y k & k * xk 2 * m & k * yk x k y k r ij R R 1 2 2 3
2 MLP SVM 217 2 Tab 2 Similarity matrix table 1 0 632 5 0 559 0 0 599 3 0 477 6 0 477 6 0 119 9 0 119 9 0 421 6 0 0 632 5 1 0 589 3 0 756 8 0 100 7 0 100 7 0 126 3 0 126 3 0 444 4 0 0 559 0 0 589 3 1 0 735 7 0 318 2 0 318 2 0 266 7 0 399 4 0 766 0 0 25 0 599 3 0 756 8 0 735 7 1 0 206 9 0 206 9 0 259 6 0 165 7 0 630 5 0 199 7 0 477 6 0 100 7 0 318 2 0 206 9 1 1 0 264 1 0 264 1 0 300 0 0 264 8 0 477 6 0 100 7 0 318 2 0 206 9 1 1 0 264 1 0 264 1 0 300 0 0 264 8 0 119 9 0 126 3 0 266 7 0 259 6 0 264 1 0 264 1 1 0 331 4 0 251 4 0 199 7 0 119 9 0 126 3 0 399 4 0 165 7 0 264 1 0 264 1 0 331 4 1 0 251 4 0 199 7 0 421 6 0 444 4 0 766 0 0 630 5 0 300 0 0 300 0 0 251 4 0 251 4 1 0 471 4 0 0 0 25 0 1997 0 2648 0 2648 0 1997 0 1997 0 4714 1 2 2 4 4 R r ij R R G g ij g ij = n r ik r kj R R = R G = R G R G R G R G = R G R R r ij = r ji r ij g ij = n r ik r kj g ij 2 = n r jk r ki = n r kj r ik = g ij g ji R R R R R = R 10 3 4
218 38 Tab 3 3 Equivalence matrix table 1 0 632 5 0 632 5 0 632 5 0 477 6 0 477 6 0 331 4 0 399 4 0 632 5 0 471 4 0 632 5 1 0 735 7 0 756 8 0 477 6 0 477 6 0 331 4 0 399 4 0 735 7 0 471 4 0 632 5 0 735 7 1 0 735 7 0 477 6 0 477 6 0 331 4 0 399 4 0 76 6 0 471 4 0 632 5 0 756 8 0 735 7 1 0 477 6 0 477 6 0 331 4 0 399 4 0 735 7 0 471 4 0 477 6 0 477 6 0 477 6 0 477 6 1 1 0 331 4 0 399 4 0 477 6 0 471 4 0 477 6 0 477 6 0 477 6 0 477 6 1 1 0 331 4 0 399 4 0 477 6 0 471 4 0 331 4 0 331 4 0 331 4 0 331 4 0 331 4 0 331 4 1 0 331 4 0 331 4 0 331 4 0 399 4 0 399 4 0 399 4 0 399 4 0 399 4 0 399 4 0 331 4 1 0 399 4 0 399 4 0 632 5 0 735 7 0 76 6 0 735 7 0 477 6 0 477 6 0 331 4 0 399 4 1 0 471 4 0 471 4 0 471 4 0 471 4 0 471 4 0 471 4 0 471 4 0 331 4 0 399 4 0 471 4 1 2 2 5 1 10 5 OA 3 ~ 5 3 Fig 5 5 Dynamic pattern classification MLP SVM OA 1 J 2000 7 8 21-22 2 J 2005 9 38 133-135 3 J 1996 23 2 28-31 4 J 2009 26 1 9-16 5 J 1997 22 3 186 6 J 2000 41 2 116-117 7 J 2003 34 1 89-91 8 RBF MLP D 2007 9 J 2007 22 8 882-888 10 M 1983