· KLDP.org · KLDP.net · KLDP Wiki · KLDP BBS ·
cjftn0/2008-03-30

Describe cjftn0/2008-03-30


-¹ÌÆý𣿡 10cvÀÇ Æò±ÕÀÌ 70%Á¤µµ¶ó°í ¸»¾¸µå·È´Âµ¥, À̹øÀÇ °á°ú´Â ¾à°£ ³·°Ô ³ª¿Ô½À´Ï´Ù. ÀÌ·± ÀÌÀ¯´Â ·£´ý »ùÇøµ¿¡ ÀÇÇؼ­ ÃʱâÀÇ »ùÇõéÀ» n1,n2,.....,n10±îÁöÀÇ µ¥ÀÌÅÍ ¼ÂÀ¸·Î ³ª´©°Ô µÇ´Âµ¥, ÀÌ µ¥ÀÌÅÍ ¼ÂÀÇ °¢ ¸ð¼ö(mean, variance)°¡ ¸Å¹ø ¸ðµ¨ ¸¸µé ¶§ ¸¶´Ù º¯°æµÇ°Ô µÇ°í, ÀÌ·¯ÇÑ ¿µÇâÀÌ ¸Å¹ø °á°ú¿¡ ¿µÇâÀ» Áֱ⠶§¹®À¸·Î »ý°¢µË´Ï´Ù. ¾ÕÀ¸·Î ¿©·¯ ¹ø ¸ðµ¨À» ¸¸µé¾î¼­ ¸ðµ¨ Á¤È®µµÀÇ Æò±ÕÀ» »êÃâÇÏ¿© °á°ú¿¡ Æ÷ÇÔ½ÃÅ°°Ú½À´Ï´Ù. ±³¼ö´Ô °¡¸£Ä§À¸·Î Áö±Ý±îÁö °£°ú ÇÏ°í ÀÖ¾ú´ø ºÎºÐÀ» »õ»ï ±ú´Ý°Ô µÇ¾ú½À´Ï´Ù. ^^;;

-ºÐ·ù ¸ðµ¨ ±¸¼º¿¡ »ç¿ëµÈ ¹æ½Ä

1.msc.features.select ·Î ¼Ó¼ºÂ÷¿ø ÁÙÀÓ 2.tree ¸ðµ¨ - »ç¿ëµÈ feature selection ¹æ½ÄÀº gini
  1. 10 cross validation


-½Ãµµ Çغ» ¹æ¹ý

1.¼Ó¼º ¼±ÅÃÇÏ´Â ´Ù¸¥ ¸Þ¼­µå ÀÌ¿ë- relifcat feature selection
°á°ú- 10cvÀÇ Æò±Õ ¾à 58%
2.rpart ÀÌ¿ëÇÏ¿© ¸ðµ¨ ±¸Ãà½Ãµµ -º¤ÅÍÇÒ´ç ¿¡·¯ or too many element specified 3.msc.features.select¿¡¼­ RemCorrcol, keepCol ÆĶó¹ÌÅÍ Á¶Á¤


0.98/0.72 => null vector (¼±ÅÃµÈ ¼Ó¼ºÀÌ ¾ø´Â cv ¹ß»ý) 0.98/0.75 => null vector (¼±ÅÃµÈ ¼Ó¼ºÀÌ ¾ø´Â cv ¹ß»ý) 0.98/0.8 => null vector (¼±ÅÃµÈ ¼Ó¼ºÀÌ ¾ø´Â cv ¹ß»ý)


-½Ãµµ µµÁß ¾òÀº ¾ÆÀ̵ð¾î
1.¿ÀºÐ·ùÀ²ÀÌ 15% ÀÌÇÏÀÎ 10cvµé Áß¿¡¼­ °øÅëµÈ ¼Ó¼ºÀ» ¹ß°ß ÀÌ ¼Ó¼ºÀÌ »ç¿ë ¾È µÈ 10cv´Â ¿ÀºÐ·ùÀ²ÀÌ ³ô´Ù´Â °ÍÀ» ¹ß°ß
=>¿ÀºÐ·ùÀ²ÀÌ 15% ÀÌÇÏÀÎ 10cvµé Áß¿¡¼­ ¼±ÅÃµÈ ¼Ó¼ºµéÀÇ Àüü ÁýÇÕ¿¡¼­ ºó¹ßÇÏ ´Â ¼Ó¼º°ú °øÅë¼Ó¼ºÀ» ÀÌ¿ëÇÏ¿© ¸ðµ¨ ±¸Ãà.´ÙÀ½ÁÖ °èȹ


ºó¹ßÇÏ´Â ¼Ó¼ºÀÇ ºóµµ¼ö ¹× °øÅë¼Ó¼º(5¹øÀÇ ºóµµ¼ö °¡Áø ¼Ó¼º)

5¹øÀÇ 10 cross validationÀ» ÅëÇØ ¸¸µé¾îÁø 10cv Áß¿¡¼­ ¾à15% ÀÌÇÏÀÇ ¿ÀºÐ·ùÀ²À» ³ªÅ¸ ³½ 10cv ¸¸ ¼±ÅÃÇÏ¿©, ÀÌ°ÍÀÇ ¼±Åà µÇ¾îÁø ¼Ó¼ºÀÇ °øÅëµÈ ¼Ó¼º°ú ºóµµ¼ö¸¦ ±¸Çß½À´Ï´Ù.

¼Ó¼º ºóµµ¼ö "X228769_at" °øÅë¼Ó¼º ¡°X227094_at¡± 4 "X219821_s_at" 3 "X1557483_at" 3 "X1564190_x_at" 2 "X221572_s_at" 2 "X227356_at" 2 "X219429_at" 2



Classification tree: tree(formula = class ~ ., data = iter10, na.action = na.pass,
split = c("gini"), x = FALSE, y = TRUE)
Variables actually used in tree construction: 1 "X219821_s_at" "X1557483_at" "X227094_at" "X212707_s_at" 5 "X234668_at" "X220455_at" "X206819_at" "X1564190_x_at" 9 "X228769_at" Number of terminal nodes: 13 Residual mean deviance: 0.5868 = 69.24 / 118 Misclassification error rate: 0.1527 = 20 / 131
a10 <-classError(p10, iter_t10,num) a10
$misclassified 1 7 11 $errorRate 1 0.1538462

Classification tree: tree(formula = class ~ ., data = iter7, na.action = na.pass,
split = c("gini"), x = FALSE, y = TRUE)
Variables actually used in tree construction: 1 "X219821_s_at" "X210387_at" "X221572_s_at" "X1564190_x_at" 5 "X228769_at" "X227094_at" "X227733_at" Number of terminal nodes: 12 Residual mean deviance: 0.4563 = 54.3 / 119 Misclassification error rate: 0.1145 = 15 / 131
a7 <-classError(p7, iter_t7,num) a7
$misclassified 1 3 10 $errorRate 1 0.1538462

Classification tree: tree(formula = class ~ ., data = iter10, na.action = na.pass,
split = c("gini"), x = FALSE, y = TRUE)
Variables actually used in tree construction: 1 "X230134_s_at" "X228769_at" "X204712_at" "X227094_at" "X205695_at" 6 "X227356_at" "X1557483_at" "X223517_at" Number of terminal nodes: 13 Residual mean deviance: 0.4802 = 56.66 / 118 Misclassification error rate: 0.1069 = 14 / 131
a10 <-classError(p10, iter_t10,num) a10
$misclassified 1 7 9 $errorRate 1 0.1538462


Classification tree: tree(formula = class ~ ., data = iter6, na.action = na.pass,
split = c("gini"), x = FALSE, y = TRUE)
Variables actually used in tree construction: 1 "X219821_s_at" "X219429_at" "X243476_at" "X228769_at" "X222283_at" 6 "X201611_s_at" "X1557483_at" "X227094_at" Number of terminal nodes: 13 Residual mean deviance: 0.4657 = 54.96 / 118 Misclassification error rate: 0.1221 = 16 / 131
a6 <-classError(p6, iter_t6,num) a6
$misclassified 1 5 10 $errorRate 1 0.1538462


Classification tree: tree(formula = class ~ ., data = iter8, na.action = na.pass,
split = c("gini"), x = FALSE, y = TRUE)
Variables actually used in tree construction:
1 "X206766_at" "X219429_at" "X221572_s_at" "X226745_at" "X228769_at" 6 "X227356_at" "X236717_at" "X223147_s_at" "X226748_at" "X222657_s_at"
Number of terminal nodes: 13 Residual mean deviance: 0.473 = 55.82 / 118 Misclassification error rate: 0.1145 = 15 / 131
a8 <-classError(p8, iter_t8,num) a8
$misclassified 1 5 11 $errorRate 1 0.1538462


ID
Password
Join
If you make a mistake you right it immediately to the best of your ability.


sponsored by andamiro
sponsored by cdnetworks
sponsored by HP

Valid XHTML 1.0! Valid CSS! powered by MoniWiki
last modified 2008-03-30 19:52:41
Processing time 0.0015 sec