{"id":2162,"date":"2020-04-08T02:55:54","date_gmt":"2020-04-07T17:55:54","guid":{"rendered":"http:\/\/141.164.34.82\/?p=2162"},"modified":"2020-04-08T07:50:40","modified_gmt":"2020-04-07T22:50:40","slug":"lm-rf-%ea%b4%9c%ec%b0%ae%ec%9d%80-%ed%98%bc%ec%a2%85","status":"publish","type":"post","link":"http:\/\/ds.sumeun.org\/?p=2162","title":{"rendered":"lm + rf : \uad1c\ucc2e\uc740 \ud63c\uc885"},"content":{"rendered":"

\ub79c\ub364\ud3ec\ub808\uc2a4\ud2b8\uc758 \ud55c\uacc4<\/h2>\n

\ubd80\uc2a4\ud305\uc758 \uacbd\uc6b0\ub3c4 \ub9c8\ucc2c\uac00\uc9c0\uc9c0\ub9cc, \uc758\uc0ac\uacb0\uc815\ub098\ubb34\ub97c \uae30\ubc18\uc73c\ub85c \ud55c \ub79c\ub364\ud3ec\ub808\uc2a4\ud2b8\ub294 \ubc18\uc751\uace1\uc120(response curve) \ub610\ub294 \ubc18\uc751\uace1\uba74(response surface)\uac00 \uc218\ud3c9\uc120, \uc218\ud3c9\uba74\uc77c \uc218 \ubc16\uc5d0 \uc5c6\ub2e4. \ubb3c\ub860 \uba40\ub9ac\uc11c \ubcf4\uba74 \uace1\uc120\uc73c\ub85c \ubcfc \uc218\ub3c4 \uc788\uaca0\uc9c0\ub9cc, \uc544\uc8fc \uc790\uc138\ud788 \ub4e4\uc5b4\uac00\ubcf4\uba74 \uc791\uc740 \uc218\ud3c9\uc120 \uad6c\uac04\uc758 \ubaa8\uc74c\uc774\ub2e4.<\/p>\n

\uc774\ub294 \ud68c\uadc0: \ub0b4\uc0bd\uacfc \uc678\uc0bd<\/a>\uc5d0\uc11c\ub3c4 \uc124\uba85\ud588\ub4ef\uc774 \uc678\uc0bd<\/strong>\uc5d0\uc11c\ub294 \uc5c4\uccad\ub09c \ud398\ub110\ud2f0\ub85c \uc791\uc6a9\ud560 \uc218\ubc16\uc5d0 \uc5c6\ub2e4. \ub2e4\uc2dc \ub9d0\ud574 \uae30\uc6b8\uae30\uac00 0\uc774 \uc544\ub2cc \uc9c1\uc120\uc744 \ub79c\ub364\ud3ec\ub808\uc2a4\ud2b8\ub85c \ud559\uc2b5\ud560 \uc218\ub3c4 \uc788\uaca0\uc9c0\ub9cc, \uc120\ud615\ud68c\uadc0\ub97c \uc0ac\uc6a9\ud55c\ub2e4\uba74 10\uac1c \ub0b4\uc678\uc758 \ub370\uc774\ud130\ub85c \ucda9\ubd84\ud788 \ud559\uc2b5\ud560 \uc218 \uc788\ub294 \ubaa8\ud615\uc744 1000\uac1c \uc774\uc0c1\uc758 \ub370\uc774\ud130\ub97c \uc368\ub3c4 \uc798 \ud559\uc2b5\ud558\uc9c0 \ubabb\ud560 \uc218 \uc788\ub294 \uac83\uc774\ub2e4. (\uc608. \ud53c\ucc98 \uc5d4\uc9c0\ub2c8\uc5b4\ub9c1 2<\/a>)<\/p>\n

\uc774\ub860\uc801\uc73c\ub85c \ub4b7\ubc1b\uce68\ub41c \uc120\ud615 \ud68c\uadc0\uc758 \ub180\ub77c\uc6b4 \uc678\uc0bd<\/strong> \ub2a5\ub825\uacfc \ub79c\ub364\ud3ec\ub808\uc2a4\ud2b8\uc758 \uc548\uc815\uc801\uc778 \ub0b4\uc0bd<\/strong>\ub2a5\ub825\uc744 \ud569\uce60 \uc21c \uc5c6\uc744\uae4c?<\/p>\n

\ubaa8\ud615<\/h3>\n

\ub370\uc774\ud130 \uc0dd\uc131 \ubaa8\ud615\uc740 \ub2e4\uc74c\uacfc \uac19\ub2e4.<\/p>\n

\\[y = x_1 + \\sin(x_2) + \\cos(x_3 + \\pi) + y_0(x_1, x_2, x_3) + e, \\ \\ e \\sim \\mathcal{N}(0,0.5^2)\\]<\/p>\n

library(randomForest)\nlibrary(dplyr)\nlibrary(tidyr)\nlibrary(forcats)\nlibrary(ggplot2)\n\nx1 = rnorm(1000)\nx2 = runif(1000, -3, 3)\nx3 = 6*(rbeta(1000, 2, 1)-1\/2)\n#e = runif(1000, -1, 1)\ne = rnorm(1000, 0, 0.5)\n\ndat = data.frame(x1, x2, x3)\ndat$y0 = with(dat, ifelse(x2 < -0.5, 1,\n               ifelse(x1 > 0.5 & x3 < -0.5, -2,\n                      ifelse(x1 > 1 & x2 > 1.1 & (x3 > 1 & x3 <1.4), 0.5, 0))))\n\ndat$y = with(dat, 1*x1 + sin(x2) + cos(x3+pi) + dat$y0 + e)\n<\/code><\/pre>\n
\uadf8\ub798\ud504\ub97c \uadf8\ub824\ubcf4\uba74 \ub2e4\uc74c\uacfc \uac19\ub2e4.<\/p>\n
pairs(dat %>% select(x1, x2, x3, y))\n<\/code><\/pre>\n
\"plot<\/p>\n
\uc120\ud615 \ud68c\uadc0 + \ub79c\ub364 \ud3ec\ub808\uc2a4\ud2b8<\/h2>\n
\ub69c\ub837\ud55c \uad00\uacc4 \ub610\ub294 \uc774\ub860\uc801 \uad00\uacc4<\/h3>\n
\uc704\uc758 \uadf8\ub798\ud504\ub97c \ubcf4\uba74 \\(y\\) \uc640 \\(x_1\\) \uc758 \uc120\ud615\uc801 \uad00\uacc4\ub97c \ub69c\ub837\uc774 \uad00\ucc30\ud560 \uc218 \uc788\ub2e4. \uc190\uc27d\uac8c \uc0dd\uac01\ud560 \uc218 \uc788\ub294 \ubaa8\ud615\uc740 \ub2e4\uc74c\uacfc \uac19\uc740 \uc120\ud615 \ubaa8\ud615\uc774\ub2e4.<\/p>\n
\\[y = \\beta_0 + \\beta_1 x_1 + e\\]<\/p>\n
summary(lm(y ~ x1, dat))\n<\/code><\/pre>\n
## \n## Call:\n## lm(formula = y ~ x1, data = dat)\n## \n## Residuals:\n##     Min      1Q  Median      3Q     Max \n## -3.5220 -0.6975  0.0033  0.6982  3.0868 \n## \n## Coefficients:\n##             Estimate Std. Error t value Pr(>|t|)    \n## (Intercept)  0.29561    0.03193   9.258   <2e-16 ***\n## x1           0.92066    0.03177  28.979   <2e-16 ***\n## ---\n## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1\n## \n## Residual standard error: 1.009 on 998 degrees of freedom\n## Multiple R-squared:  0.457,  Adjusted R-squared:  0.4564 \n## F-statistic: 839.8 on 1 and 998 DF,  p-value: < 2.2e-16\n<\/pre>\n
\uc774\uc81c \ubcf8\uaca9\uc801\uc73c\ub85c ML\uc744 \ud558\uae30 \uc804\uc5d0, \ubaa8\ud615 \uc131\ub2a5\uc744 \ud3c9\uac00\ud558\ub294 \ubc29\ubc95\uc744 CV(C<\/strong>ross V<\/strong>alidation)\ub85c \ud558\uaca0\ub2e4(\ucc38\uace0<\/a>). <\/p>\n
createfolds = function(N, k) {\n  isamp <- sample(N)\n  split(isamp, cut(1:N,\n                   breaks=N*seq(0, 1, length.out=k+1),\n                   include.lowest=TRUE))\n}\n\nfolds <- createfolds(1000,5)\n<\/code><\/pre>\n
\uc2e4\ub840<\/h2>\n
\uc774\uc81c \uccab \ubc88\uc9f8 \ud6c8\ub828-\uac80\uc99d\uc14b\uc5d0 \ub300\ud574 \uae30\uacc4\ud559\uc2b5\uc744 \ud574\ubcf4\uc790. \ud3c9\uac00\ub294 MAE(M<\/strong>ean A<\/strong>bsolute E<\/strong>rror)\ub97c \uc0ac\uc6a9\ud558\uaca0\ub2e4. <\/p>\n
1. \uc624\uc9c1 \ub79c\ub364\ud3ec\ub808\uc2a4\ud2b8<\/h3>\n
Xtrain = dat[-folds[[1]], ] %>% select(x1:x3) %>% as.matrix\nytrain = dat[-folds[[1]],] %>% pull(y)\n\nXtest = dat[folds[[1]], ] %>% select(x1:x3) %>% as.matrix\nytest = dat[folds[[1]],] %>% pull(y)\n\nfitRF <- randomForest(x=Xtrain, y=ytrain)\n\nypred <- predict(fitRF, newdata=Xtest)\ncaret::MAE(ypred, ytest)\n<\/code><\/pre>\n
## [1] 0.4937447\n<\/pre>\n
2. \uc120\ud615\ud68c\uadc0 + \ub79c\ub364\ud3ec\ub808\uc2a4\ud2b8<\/h3>\n
\uc55e\uc11c \uc598\uae30\ud588\ub4ef\uc774 \ub79c\ub364\ud3ec\ub808\uc2a4\ud2b8\ub294 \uc120\ud615\uc801 \uad00\uacc4\uc5d0 \ub454\uac10\ud558\ub2e4. \uc774\uc81c \ub69c\ub837\uc774 \ub098\ud0c0\ub09c \uc120\ud615 \uad00\uacc4\ub97c \uc120\ud615 \ud68c\uadc0\ub97c \uc0ac\uc6a9\ud558\uc5ec \ubaa8\ub378\ub9c1\ud558\uace0, \ub098\uba38\uc9c0 \uc794\ucc28\ub97c \ub79c\ub364\ud3ec\ub808\uc2a4\ud2b8\ub85c \uc608\uce21\ud574\ubcf4\uc790. \uad6c\uccb4\uc801\uc73c\ub85c \ub2e4\uc74c\uacfc \uac19\ub2e4. \ud559\uc2b5\uc740 lm<\/code>, randomForest<\/code>\ub85c \uc9c4\ud589\ud558\uace0, \uc608\uce21\uc740 predict<\/code>, predict<\/code>\uc758 \uac12\uc744 \ud569\ud558\ub294 \uac83\uc73c\ub85c \uc774\ub8e8\uc5b4\uc9c4\ub2e4. 1\ubc88\ubcf4\ub2e4 \uc131\ub2a5\uc774 \uc88b\uc740\uac00?<\/p>\n
fitLM <- lm(y ~ x1, data=dat[-folds[[1]],])\nfitRF2 <- randomForest(x=Xtrain[, c('x2', 'x3')], y=resid(fitLM))\n\nypredLM <- predict(fitLM, newdata = dat[folds[[1]],])\nresidLM <- predict(fitRF2, newdata = Xtest[, c('x2', 'x3')])\nypred2 <- ypredLM + residLM\ncaret::MAE(ypred2, ytest)\n<\/code><\/pre>\n
## [1] 0.5009684\n<\/pre>\n
CV(1 vs 2)<\/h3>\n
\ud55c\ub450 \ubc88\uc758 \uacb0\uacfc\ub9cc\uc73c\ub85c \ub450 \ubc29\ubc95\uc744 \ube44\uad50\ud558\uae30\uc5d0\ub294 \ubb34\ub9ac\uac00 \uc788\uc73c\ubbc0\ub85c CV\ub97c \ud574\ubcf8\ub2e4. <\/p>\n
dfMAE <- data.frame(rf = rep(NA, 5), lmrf = rep(NA,5))\nfor (i in 1:5) {\n  Xtrain = dat[-folds[[i]], ] %>% select(x1:x3) %>% as.matrix\n  ytrain = dat[-folds[[i]],] %>% pull(y)\n\n  Xtest = dat[folds[[i]], ] %>% select(x1:x3) %>% as.matrix\n  ytest = dat[folds[[i]],] %>% pull(y)\n\n  fitRF <- randomForest(x=Xtrain, y=ytrain)\n\n  ypred <- predict(fitRF, newdata=Xtest)\n  dfMAE[i, 1] <- caret::MAE(ypred, ytest)\n\n  fitLM <- lm(y ~ x1, data=dat[-folds[[i]],])\n  fitRF2 <- randomForest(x=Xtrain[, c('x2', 'x3')], y=resid(fitLM))\n  ypredLM <- predict(fitLM, newdata = dat[folds[[i]],])\n  residLM <- predict(fitRF2, newdata = Xtest[, c('x2', 'x3')])\n  ypred2 <- ypredLM + residLM\n  dfMAE[i, 2] <- caret::MAE(ypred2, ytest)\n}\n\n\ndfMAE %>% gather(key='key', value='value', rf:lmrf) %>%\n  ggplot(aes(x=key, y=value, col=key)) +\n  geom_boxplot() + \n  geom_point()\n<\/code><\/pre>\n
\"plot<\/p>\n
\uc0c1\ud638\uc791\uc6a9\uc744 \ud3ec\ud568\ud558\uae30<\/h2>\n
\uc6b0\ub9ac\ub294 \ucc38\ubaa8\ud615\uc744 \uc54c\uace0, \ucc38\ubaa8\ud615\uc5d0\uc11c \\(x_1\\) \uc740 \ub2e4\ub978 \uc608\uce21 \ubcc0\uc218\uc640 \uc0c1\ud638\uc791\uc6a9\uc744 \uc54a\uc544\ubcf4\uc774\uc9c0\ub9cc, \uc0ac\uc2e4 \\(y_0(x1, x2, x3)\\) \uc5e3 \uc0c1\ud638\uc791\uc6a9\uc774 \uc788\uc74c\uc744 \uc54c \uc218 \uc788\ub2e4. \ub530\ub77c\uc11c \uc120\ud615 \ud68c\uadc0 \uc774\ud6c4 \uc794\ucc28 \uc608\uce21\uc5d0 \\(x_1\\) \uc744 \ub2e4\uc2dc \uc0ac\uc6a9\ud55c\ub2e4\uba74 \uc131\ub2a5\uc774 \uc99d\uac00\ud560 \uc218 \uc788\ub2e4. \uadf8\ub9ac\uace0 \uc815\ub9d0 \uadf8\ub807\ub2e4!<\/p>\n
dfMAE <- data.frame(rf = rep(NA, 5), lmrf = rep(NA,5), `lmrf+I` = rep(NA,5))\nfor (i in 1:5) {\n  Xtrain = dat[-folds[[i]], ] %>% select(x1:x3) %>% as.matrix\n  ytrain = dat[-folds[[i]],] %>% pull(y)\n\n  Xtest = dat[folds[[i]], ] %>% select(x1:x3) %>% as.matrix\n  ytest = dat[folds[[i]],] %>% pull(y)\n\n  fitRF <- randomForest(x=Xtrain, y=ytrain)\n\n  ypred <- predict(fitRF, newdata=Xtest)\n  dfMAE[i, 1] <- caret::MAE(ypred, ytest)\n\n  fitLM <- lm(y ~ x1, data=dat[-folds[[i]],])\n  fitRF2 <- randomForest(x=Xtrain[, c('x2', 'x3')], y=resid(fitLM))\n  ypredLM <- predict(fitLM, newdata = dat[folds[[i]],])\n  residLM <- predict(fitRF2, newdata = Xtest[, c('x2', 'x3')])\n  ypred2 <- ypredLM + residLM\n  dfMAE[i, 2] <- caret::MAE(ypred2, ytest)\n\n  fitLM <- lm(y ~ x1, data=dat[-folds[[i]],])\n  fitRF2 <- randomForest(x=Xtrain[, c('x1', 'x2', 'x3')], y=resid(fitLM))\n  ypredLM <- predict(fitLM, newdata = dat[folds[[i]],])\n  residLM <- predict(fitRF2, newdata = Xtest[, c('x1', 'x2', 'x3')])\n  ypred2 <- ypredLM + residLM\n  dfMAE[i, 3] <- caret::MAE(ypred2, ytest)\n}\n\n\ndfMAE %>% gather(key='key', value='value', rf:lmrf.I) %>%\n  ggplot(aes(x=key, y=value, col=key)) +\n  geom_boxplot() + \n  geom_point()\n<\/code><\/pre>\n
\"plot<\/p>\n
dfMAE$i = 1:5\n\ndfPlot <- dfMAE %>% gather(key='key', value='value', rf:lmrf.I)\n\ndfPlot$key = dfPlot$key %>% \n  fct_reorder(dfPlot$value)\n\ndfPlot %>%\n  ggplot(aes(x=key, y=value, col=key)) +\n  geom_line(aes(group=i), col='gray20') + \n  geom_point(size=5, col='white') + \n  geom_point(size=4)\n<\/code><\/pre>\n
\"plot<\/p>\n
\uc120\ud615 \ud68c\uadc0 \ud55c \ubc88 \ub354!<\/h2>\n
\uc6b0\ub9ac\ub294 \uc0b0\uc810\ub3c4 \ud589\ub82c\uc744 \uadf8\ub824\ubcf8 \ud6c4, \uc120\ud615 \ud68c\uadc0\ub97c \ud558\uace0, \ub79c\ub364\ud3ec\ub808\uc2a4\ud2b8\ub97c \ud588\ub2e4. \uadf8\ub9ac\uace0 \uc120\ud615 \ud68c\uadc0\ub97c \ud558\uae30 \uc704\ud574 \ubb54\uac00 \ud2b9\ubcc4\ud55c \uc774\ub860\uc774 \uc788\uc5c8\ub358 \uac83\uc740 \uc544\ub2c8\ub2e4. \uadf8\ub0e5 \uc0b0\uc810\ub3c4 \ud589\ub82c\uc744 \ubcf4\uace0 \ucd94\uce21\ud588\ub2e4. \ub2e4\uc2dc \ub9d0\ud574 \uc608\uce21\ud558\uace0\uc790 \ud558\ub294 \\(y\\) \uc640 \uc608\uce21 \ubcc0\uc218 \\(x_1\\) , \\(x_2\\) , \\(x_3\\) \uc758 \uad00\uacc4\uc5d0 \ub300\ud574 \uadf8\ub7f4\uc2f8\ud55c \uac00\uc124\uc744 \ub9cc\ub4e4\uc5c8\ub2e4\ub294 \uac83\uc774\ub2e4. <\/p>\n
\ub79c\ub364\ud3ec\ub808\uc2a4\ud2b8\ub294 \uccab \ubc88\uc9f8 \uc120\ud615 \ud68c\uadc0 \uacb0\uacfc\ub85c \uc5bb\uc740 \uc794\ucc28\ub97c \\(x_1\\) , \\(x_2\\) , \\(x_3\\) \ub85c \uc608\uce21\ud558\ub824\uace0 \ud55c\ub2e4. \uc774\ub4e4\uc5d0 \ub300\ud574 \uc0b0\uc810\ub3c4 \ud589\ub82c\uc744 \ub2e4\uc2dc \uadf8\ub824 \ubcf4\uc790! <\/p>\n
\ud3ec\ubb3c\uc120\uc774 \ubcf4\uc774\uc9c0 \uc54a\uc740\uac00?<\/p>\n
fitLM_all <- lm(y~x1, dat)\ndat$resid <- resid(fitLM_all)\npairs(dat %>% select(resid, x1, x2, x3))\n<\/code><\/pre>\n
\"plot<\/p>\n
fitLM_all2 <- lm(resid ~ I(x3^2), dat)\ndat$resid2 <- resid(fitLM_all2)\npairs(dat %>% select(resid2, x1, x2, x3))\n<\/code><\/pre>\n
\"plot<\/p>\n
\uadf8\ub798\uc11c \ub79c\ub364 \ud3ec\ub808\uc2a4\ud2b8\ub97c \ud558\uae30 \uc804\uc5d0 \uc120\ud615\ud68c\uadc0\ub97c \ud55c \ubc88 \ub354 \ud558\uae30\ub85c \ud588\ub2e4. \uc774\ubc88\uc5d0 \ub2e4\ud56d\ud68c\uadc0\ub97c \ud558\uaca0\ub2e4.<\/p>\n
\uacb0\uacfc\ub294…<\/p>\n
dfMAE <- data.frame(rf = rep(NA, 5), lmrf = rep(NA,5), \n                    `lmrf+I` = rep(NA,5),\n                    'lm2rf' = rep(NA,5))\nfor (i in 1:5) {\n  Xtrain = dat[-folds[[i]], ] %>% select(x1:x3) %>% as.matrix\n  ytrain = dat[-folds[[i]],] %>% pull(y)\n\n  Xtest = dat[folds[[i]], ] %>% select(x1:x3) %>% as.matrix\n  ytest = dat[folds[[i]],] %>% pull(y)\n\n  fitRF <- randomForest(x=Xtrain, y=ytrain)\n\n  ypred <- predict(fitRF, newdata=Xtest)\n  dfMAE[i, 1] <- caret::MAE(ypred, ytest)\n\n  fitLM <- lm(y ~ x1, data=dat[-folds[[i]],])\n  fitRF2 <- randomForest(x=Xtrain[, c('x2', 'x3')], y=resid(fitLM))\n  ypredLM <- predict(fitLM, newdata = dat[folds[[i]],])\n  residLM <- predict(fitRF2, newdata = Xtest[, c('x2', 'x3')])\n  ypred2 <- ypredLM + residLM\n  dfMAE[i, 2] <- caret::MAE(ypred2, ytest)\n\n  fitLM <- lm(y ~ x1, data=dat[-folds[[i]],])\n  fitRF2 <- randomForest(x=Xtrain[, c('x1', 'x2', 'x3')], y=resid(fitLM))\n  ypredLM <- predict(fitLM, newdata = dat[folds[[i]],])\n  residLM <- predict(fitRF2, newdata = Xtest[, c('x1', 'x2', 'x3')])\n  ypred2 <- ypredLM + residLM\n  dfMAE[i, 3] <- caret::MAE(ypred2, ytest)\n\n  datTrain = dat[-folds[[i]],]\n  fitLM <- lm(y ~ x1, data=datTrain)\n  datTrain$resid = resid(fitLM)\n  fitLM2 <- lm(resid ~ I(x3^2), datTrain)\n  datTrain$resid2 = resid(fitLM2)\n  fitRF3 <- randomForest(resid2 ~ x1 + x2 + x3, datTrain)\n\n  datTest = dat[folds[[i]],]\n  ypred <- predict(fitLM, newdata = datTest)\n  residpred <- predict(fitLM2, newdata = datTest)\n  resid2pred <- predict(fitRF3, newdata = datTest)\n\n  ypred3 <- ypred + residpred + resid2pred\n  dfMAE[i, 4] <- caret::MAE(ypred3, ytest)\n}\n\ndfMAE %>% gather(key='key', value='value', rf:lm2rf) %>%\n  ggplot(aes(x=key, y=value, col=key)) +\n  geom_boxplot() + \n  geom_point()\n<\/code><\/pre>\n
\"plot<\/p>\n
dfMAE$i = 1:5\n\ndfPlot <- dfMAE %>% gather(key='key', value='value', rf:lm2rf)\n\ndfPlot$key = dfPlot$key %>% \n  fct_reorder(dfPlot$value)\n\ndfPlot %>%\n  ggplot(aes(x=key, y=value, col=key)) +\n  geom_line(aes(group=i), col='gray20') + \n  geom_point(size=5, col='white') + \n  geom_point(size=4)\n<\/code><\/pre>\n
\"plot<\/p>\n
\ub098\uc058\uc9c0 \uc54a\ub2e4!<\/p>\n
\uacb0\ub860<\/h2>\n
    \n
  1. \uc120\ud615\ud68c\uadc0\ub294 \ube44\uc120\ud615\uc131<\/strong>\uc5d0 \ucde8\uc57d\ud558\uace0, \ub79c\ub364 \ud3ec\ub808\uc2a4\ud2b8\ub294 \uc678\uc0bd<\/strong>\uc5d0 \ucde8\uc57d\ud558\ub2e4.<\/li>\n
  2. \uc774 \ub458\uc744 \ud569\uce60 \uc218 \uc788\ub2e4!<\/li>\n<\/ol>\n
    \uc8fc\uc758!<\/h3>\n
      \n
    • \ud558\uc9c0\ub9cc \uc704\uc758 \ubc29\ubc95\uc740 \uc804\uccb4 \ub370\uc774\ud130\ub97c \ud1b5\ud574 \\(y\\) \uc640 \\(x_1\\) , \\(x_2\\) , \\(x_3\\) \uc758 \uad00\uacc4\ub97c \ucd94\uc815\ud588\uc73c\ubbc0\ub85c, \ubc18\uce59\uc774\ub77c\uace0 \ubcfc \uc218 \uc788\ub2e4. \ub418\ub3c4\ub85d \uc774\ub860<\/strong>\uc801\uc778 \uadfc\uac70\ub97c \ud1b5\ud574 \uc678\uc0bd \uac00\ub2a5\ud55c \uc120\ud615 \uad00\uacc4\ub97c \uc0dd\uac01\ud558\ub294 \uac83\uc774 \uc88b\uc744 \uac83 \uac19\ub2e4.<\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"
      \ub79c\ub364\ud3ec\ub808\uc2a4\ud2b8\uc758 \ud55c\uacc4 \ubd80\uc2a4\ud305\uc758 \uacbd\uc6b0\ub3c4 \ub9c8\ucc2c\uac00\uc9c0\uc9c0\ub9cc, \uc758\uc0ac\uacb0\uc815\ub098\ubb34\ub97c \uae30\ubc18\uc73c\ub85c \ud55c \ub79c\ub364\ud3ec\ub808\uc2a4\ud2b8\ub294 \ubc18\uc751\uace1\uc120(response curve) \ub610\ub294 \ubc18\uc751\uace1\uba74(response surface)\uac00 \uc218\ud3c9\uc120, \uc218\ud3c9\uba74\uc77c \uc218 \ubc16\uc5d0 \uc5c6\ub2e4. \ubb3c\ub860 \uba40\ub9ac\uc11c \ubcf4\uba74 \uace1\uc120\uc73c\ub85c \ubcfc \uc218\ub3c4 \uc788\uaca0\uc9c0\ub9cc, \uc544\uc8fc \uc790\uc138\ud788 \ub4e4\uc5b4\uac00\ubcf4\uba74 \uc791\uc740 \uc218\ud3c9\uc120 \uad6c\uac04\uc758 \ubaa8\uc74c\uc774\ub2e4. \uc774\ub294 \ud68c\uadc0: \ub0b4\uc0bd\uacfc \uc678\uc0bd\uc5d0\uc11c\ub3c4 \uc124\uba85\ud588\ub4ef\uc774 \uc678\uc0bd\uc5d0\uc11c\ub294 \uc5c4\uccad\ub09c \ud398\ub110\ud2f0\ub85c \uc791\uc6a9\ud560 \uc218\ubc16\uc5d0 \uc5c6\ub2e4. \ub2e4\uc2dc \ub9d0\ud574 \uae30\uc6b8\uae30\uac00 0\uc774 \uc544\ub2cc \uc9c1\uc120\uc744 \ub79c\ub364\ud3ec\ub808\uc2a4\ud2b8\ub85c \ud559\uc2b5\ud560 \uc218\ub3c4 \uc788\uaca0\uc9c0\ub9cc, \uc120\ud615\ud68c\uadc0\ub97c […]<\/p>\n","protected":false},"author":1,"featured_media":2165,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":[],"categories":[390,28],"tags":[368,447,321,446],"jetpack_featured_media_url":"http:\/\/ds.sumeun.org\/wp-content\/uploads\/2020\/04\/lm_plus_rf.png","_links":{"self":[{"href":"http:\/\/ds.sumeun.org\/index.php?rest_route=\/wp\/v2\/posts\/2162"}],"collection":[{"href":"http:\/\/ds.sumeun.org\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/ds.sumeun.org\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/ds.sumeun.org\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"http:\/\/ds.sumeun.org\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=2162"}],"version-history":[{"count":4,"href":"http:\/\/ds.sumeun.org\/index.php?rest_route=\/wp\/v2\/posts\/2162\/revisions"}],"predecessor-version":[{"id":2167,"href":"http:\/\/ds.sumeun.org\/index.php?rest_route=\/wp\/v2\/posts\/2162\/revisions\/2167"}],"wp:featuredmedia":[{"embeddable":true,"href":"http:\/\/ds.sumeun.org\/index.php?rest_route=\/wp\/v2\/media\/2165"}],"wp:attachment":[{"href":"http:\/\/ds.sumeun.org\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=2162"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/ds.sumeun.org\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=2162"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/ds.sumeun.org\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=2162"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}