{"id":27639,"date":"2024-02-01T17:25:40","date_gmt":"2024-02-01T09:25:40","guid":{"rendered":"http:\/\/www.biocloudservice.com\/wordpress\/?p=27639"},"modified":"2024-02-01T17:25:42","modified_gmt":"2024-02-01T09:25:42","slug":"%e7%94%9f%e5%ad%98%e5%88%86%e6%9e%90%e5%88%ab%e5%8f%aa%e7%94%a8cox%e5%9b%9e%e5%bd%92%e4%ba%86%ef%bc%8c%e8%b7%9f%e7%9d%80%e5%b0%8f%e6%9e%9c%e8%af%95%e8%af%95%e9%9a%8f%e6%9c%ba%e6%a3%ae%e6%9e%97","status":"publish","type":"post","link":"http:\/\/www.biocloudservice.com\/wordpress\/?p=27639","title":{"rendered":"\u751f\u5b58\u5206\u6790\u522b\u53ea\u7528Cox\u56de\u5f52\u4e86\uff0c\u8ddf\u7740\u5c0f\u679c\u8bd5\u8bd5\u968f\u673a\u68ee\u6797\u751f\u5b58\u5206\u6790\uff01-randomForestSRC"},"content":{"rendered":"

\u5c0f\u4f19\u4f34\u505a\u751f\u5b58\u5206\u6790\u7684\u65f6\u5019\u907f\u514d\u4e0d\u4e86\u6784\u5efaCox\u56de\u5f52\u98ce\u9669\u6a21\u578b\uff0c\u5c0f\u679c\u544a\u8bc9\u5927\u5bb6\u4e00\u4e2a\u65b0\u7684\u65b9\u5f0f\uff0c\u5c0f\u4f19\u4f34\u53ef\u4ee5\u4e0d\u518d\u7528Cox\u56de\u5f52\u4e86\uff0c\u6211\u4eec\u53bb\u8bd5\u8bd5\u673a\u68b0\u5b66\u4e60\u7684\u65b9\u6cd5\u4e4b\u4e00\u968f\u673a\u68ee\u6797\u7684\u751f\u5b58\u5206\u6790\u4e5f\u5c31\u662frandomForestSRC\u3002\u6211\u4eec\u5148\u4e86\u89e3\u4e00\u4e0b\u65b9\u6cd5\uff0c<\/p>\n

\u968f\u673a\u751f\u5b58\u68ee\u6797\u5462\uff0c\u5176\u5b9e\u662f\u901a\u8fc7\u8bad\u7ec3\u5927\u91cf\u751f\u5b58\u6811\uff0c\u4ee5\u8868\u51b3\u7684\u5f62\u5f0f\uff0c\u4ece\u4e2a\u4f53\u6811\u4e4b\u4e2d\u52a0\u6743\u9009\u4e3e\u51fa\u6700\u7ec8\u7684\u9884\u6d4b\u7ed3\u679c\u3002<\/p>\n

\u6784\u5efa\u968f\u673a\u751f\u5b58\u68ee\u6797\u5462\u4e00\u822c\u6709\u4e0b\u9762\u51e0\u4e2a\u6d41\u7a0b<\/p>\n

\u2160. \u6a21\u578b\u901a\u8fc7\u201c\u81ea\u52a9\u6cd5\u201d\uff08Bootstrap\uff09\u5c06\u539f\u59cb\u6570\u636e\u4ee5\u6709\u653e\u56de\u7684\u5f62\u5f0f\u968f\u673a\u62bd\u53d6\u6837\u672c\uff0c\u5efa\u7acb\u6837\u672c\u5b50\u96c6\uff0c\u5e76\u5c06\u6bcf\u4e2a\u6837\u672c\u4e2d37%\u7684\u6570\u636e\u4f5c\u4e3a\u888b\u5916\u6570\u636e\uff08Out-of-Bag Data\uff09\u6392\u9664\u5728\u5916\uff1b<\/p>\n

\u2161. \u5bf9\u6bcf\u4e00\u4e2a\u6837\u672c\u968f\u673a\u9009\u62e9\u7279\u5f81\u6784\u5efa\u5176\u5bf9\u5e94\u7684\u751f\u5b58\u6811\uff1b<\/p>\n

\u2162. \u5229\u7528Nelson-Aalen\u6cd5\u4f30\u8ba1\u968f\u673a\u751f\u5b58\u68ee\u6797\u6a21\u578b\u7684\u603b\u7d2f\u79ef\u98ce\u9669\uff1b \u2163. \u4f7f\u7528\u888b\u5916\u6570\u636e\u8ba1\u7b97\u6a21\u578b\u51c6\u786e\u5ea6\u3002<\/p>\n

\u4e0b\u9762\u6211\u4eec\u5f00\u59cb\u5427\uff0c\u6211\u4eec\u5148\u6309\u7167\u4e00\u4e0bR\u5305\uff0c\u6ca1\u6709\u7684\u5c0f\u4f19\u4f34\u5148\u5b89\u88c5\u4e00\u4e0b\u54e6<\/p>\n

if (!require(randomForestSRC)) install.packages(“randomForestSRC”)<\/p>\n

if (!require(survival)) install.packages(“survival”)<\/p>\n

library(randomForestSRC)<\/p>\n

library(survival)<\/p>\n

\u6211\u4eec\u5148\u8bfb\u53d6\u4e0b\u793a\u4f8b\u6570\u636e\u770b\u770b\uff0c\u8fd9\u662fR\u5305\u81ea\u5e26\u7684<\/p>\n

\u4e0b\u9762\u6211\u4eec\u7528\u8fd9\u51e0\u4e2a\u6570\u636e\u8fdb\u884c\u5b66\u4e60\u548c\u5c55\u793a<\/p>\n

data(veteran, package = “randomForestSRC”)<\/p>\n

head(veteran)<\/p>\n

<\/p>\n

data(pbc, package = “randomForestSRC”)<\/p>\n

head(pbc)<\/p>\n

<\/p>\n

data(wihs, package = “randomForestSRC”)<\/p>\n

head(wihs)<\/p>\n

<\/p>\n

\u6211\u4eec\u4f7f\u7528\u80ba\u764c\u7684\u6570\u636e\u96c6\u8fdb\u884c\u5206\u6790<\/p>\n

\u80ba\u764c\u7684\u4e24\u79cd\u6cbb\u7597\u65b9\u6848\u7684\u662f\u968f\u673a\u8bd5\u9a8c<\/p>\n

\u4f8b\u5b50\u4e00<\/p>\n

v.obj <- rfsrc(Surv(time, status) ~ ., data = veteran, ntree = 100, nsplit = 10,<\/p>\n

na.action = “na.impute”, tree.err = TRUE, importance = TRUE, block.size = 1)<\/p>\n

###\u8bbe\u7f6e\u6811\u7684\u4e2a\u6570\u4e3a3\uff1a<\/p>\n

## plot tree number 3<\/p>\n

plot(get.tree(v.obj, 3))<\/p>\n

<\/p>\n

###\u7ed8\u5236\u8bad\u7ec3\u68ee\u6797\u7684\u7ed3\u679c<\/p>\n

## plot results of trained forest<\/p>\n

plot(v.obj)<\/p>\n

<\/p>\n

\u6211\u4eec\u76f4\u63a5\u7ed8\u5236\u524d10\u4e2a\u4e2a\u4f53\u7684\u751f\u5b58\u66f2\u7ebf<\/p>\n

#\u7ed8\u5236\u524d\u5341\u4e2a\u4e2a\u4f53\u751f\u5b58\u66f2\u7ebf<\/p>\n

matplot(v.obj$time.interest, 100 * t(v.obj$survival.oob[1:10, ]), xlab = “Time”,<\/p>\n

ylab = “Survival”, type = “l”, lty = 1)<\/p>\n

<\/p>\n

##\u4f7f\u7528\u51fd\u6570plot.survival\u7ed8\u5236\u524d10\u4e2a\u4e2a\u4f53\u7684\u751f\u5b58\u66f2\u7ebf<\/p>\n

plot.survival(v.obj, subset = 1:10)<\/p>\n

<\/p>\n

\u4e0b\u9762\u6211\u4eec###\u5feb\u901f\u4f18\u5316\u8001\u6570\u636e\u7684\u8282\u70b9\u5927\u5c0f##\u6700\u4f18\u7684\u751f\u5b58\u8282\u70b9\u5927\u5c0f\u6bd4\u5176\u4ed6\u5bb6\u65cf#<\/p>\n

tune.nodesize(Surv(time, status) ~ ., veteran)<\/p>\n

<\/p>\n

<\/p>\n

\u4f8b\u5b502<\/p>\n

vd <- veteran<\/p>\n

vd$celltype = factor(vd$celltype)<\/p>\n

vd$diagtime = factor(vd$diagtime)<\/p>\n

vd.obj <- rfsrc(Surv(time, status) ~ ., vd, ntree = 100, nodesize = 5)<\/p>\n

plot(get.tree(vd.obj, 3))<\/p>\n

<\/p>\n

\u4e0b\u9762\u6211\u4eec\u4f7f\u7528\u75be\u75c5\u7684\u4f8b\u5b50\u6f14\u793a<\/p>\n

\u539f\u53d1\u6027\u80c6\u6c41\u6027\u809d\u786c\u5316PBC<\/p>\n

\u6765\u770b\u4e00\u4e0b<\/p>\n

pbc.obj <- rfsrc(Surv(days, status) ~ ., pbc)<\/p>\n

print(pbc.obj)<\/p>\n

<\/p>\n

#<\/p>\n

pbc.obj2 <- rfsrc(Surv(days, status) ~ ., pbc, nsplit = 10, na.action = “na.impute”)<\/p>\n

##\u4e0e\u4e0a\u9762\u76f8\u540c\uff0c\u4f46\u8fed\u4ee3\u4e22\u5931\u7684\u6570\u636e\u7b97\u6cd5<\/p>\n

pbc.obj3 <- rfsrc(Surv(days, status) ~ ., pbc, na.action = “na.impute”, nimpute = 3)<\/p>\n

## \u4f30\u7b97\u6570\u636e\u7684\u5feb\u901f\u65b9\u6cd5\uff08\u4e0d\u8fdb\u884c\u63a8\u65ad\uff09<\/p>\n

pbc.imp <- impute(Surv(days, status) ~ ., pbc, splitrule = “random”)<\/p>\n

\u6211\u4eec\u5bf9\u4e24\u79cd\u65b9\u6cd5\u6bd4\u8f83\u4e00\u4e0b<\/p>\n

\u6bd4\u8f83RF-SRC\u548cCox\u56de\u5f52\uff0c\u8bf4\u660e\u6027\u80fd\u7684c\u6307\u6570\u548cBrier\u8bc4\u5206\u63aa\u65bd\uff0c\u5047\u8bbe\u52a0\u8f7d\u4e86\u201cpec\u201d\u548c\u201csurvival\u201d<\/p>\n

##\u6bd4\u8f83RF-SRC\u548cCOX<\/p>\n

require(“survival”)<\/p>\n

require(“pec”)<\/p>\n

require(“prodlim”)<\/p>\n

## pec\u6240\u9700\u7684\u9884\u6d4b\u51fd\u6570<\/p>\n

predictSurvProb.rfsrc <- function(object, newdata, times, …) {<\/p>\n

ptemp <- predict(object, newdata = newdata, …)$survival<\/p>\n

pos <- sindex(jump.times = object$time.interest, eval.times = times)<\/p>\n

p <- cbind(1, ptemp)[, pos + 1]<\/p>\n

if (NROW(p) != NROW(newdata) || NCOL(p) != length(times))<\/p>\n

stop(“Prediction failed”)<\/p>\n

p<\/p>\n

}<\/p>\n

## \u6570\u636e\uff0c\u516c\u5f0f\u89c4\u8303<\/p>\n

data(pbc, package = “randomForestSRC”)<\/p>\n

pbc.na <- na.omit(pbc) ##remove NA’s<\/p>\n

surv.f <- as.formula(Surv(days, status) ~ .)<\/p>\n

pec.f <- as.formula(Hist(days, status) ~ 1)<\/p>\n

## \u8fd0\u884ccox\/rfsrc\u6a21\u578b\u8fdb\u884c\u8bf4\u660e\u6211\u4eec\u4f7f\u7528\u4e86\u5c11\u91cf\u7684\u6811<\/p>\n

cox.obj <- coxph(surv.f, data = pbc.na, x = TRUE)<\/p>\n

rfsrc.obj <- rfsrc(surv.f, pbc.na, ntree = 150)<\/p>\n

## \u8ba1\u7b97\u671f\u671bBrier\u5206\u6570\u7684bootstrap\u4ea4\u53c9\u9a8c\u8bc1\u4f30\u8ba1<\/p>\n

set.seed(17743)<\/p>\n

prederror.pbc <- pec(list(cox.obj, rfsrc.obj), data = pbc.na, formula = pec.f, splitMethod = “bootcv”,<\/p>\n

B = 50)<\/p>\n

print(prederror.pbc)<\/p>\n

##<\/p>\n

plot(prederror.pbc)<\/p>\n

<\/p>\n

\u4e0b\u9762\u6211\u4eec\u8ba1\u7b97Cox\u56de\u5f52<\/p>\n

#Cox<\/p>\n

## \u8ba1\u7b97cox\u56de\u5f52\u7684\u888b\u5916C\u6307\u6570\u5e76\u4e0erfsrc\u8fdb\u884c\u6bd4\u8f83<\/p>\n

rfsrc.obj <- rfsrc(surv.f, pbc.na)<\/p>\n

cat(“out-of-bag Cox Analysis …”, “\\n”)<\/p>\n

## out-of-bag Cox Analysis …<\/p>\n

cox.err <- sapply(1:100, function(b) {<\/p>\n

if (b%%10 == 0)<\/p>\n

cat(“cox bootstrap:”, b, “\\n”)<\/p>\n

train <- sample(1:nrow(pbc.na), nrow(pbc.na), replace = TRUE)<\/p>\n

cox.obj <- tryCatch({<\/p>\n

coxph(surv.f, pbc.na[train, ])<\/p>\n

}, error = function(ex) {<\/p>\n

NULL<\/p>\n

})<\/p>\n

if (!is.null(cox.obj)) {<\/p>\n

get.cindex(pbc.na$days[-train], pbc.na$status[-train], predict(cox.obj, pbc.na[-train,<\/p>\n

]))<\/p>\n

} else NA<\/p>\n

})<\/p>\n

cat(“E:\\\\SXG”)<\/p>\n

##<\/p>\n

## OOB error rates<\/p>\n

cat(“\\tRSF : “, rfsrc.obj$err.rate[rfsrc.obj$ntree], “\\n”)<\/p>\n

## RSF : 0.1714494<\/p>\n

cat(“\\tCox regression : “, mean(cox.err, na.rm = TRUE), “\\n”)<\/p>\n

<\/p>\n

\u4e0b\u9762\u6211\u4eec\u5bf9\u751f\u5b58\u7ade\u4e89\u98ce\u9669\u6bd4\u4f8b\u6a21\u578b\u8fdb\u884c\u6784\u5efa<\/p>\n

\u6765\u770b\u4e00\u4e0b<\/p>\n

#\u751f\u5b58\u7ade\u4e89\u98ce\u9669\u6bd4\u4f8b<\/p>\n

wihs.obj <- rfsrc(Surv(time, status) ~ ., wihs, nsplit = 3, ntree = 100)<\/p>\n

plot.competing.risk(wihs.obj)<\/p>\n

<\/p>\n

cif <- wihs.obj$cif.oob<\/p>\n

Time <- wihs.obj$time.interest<\/p>\n

idu <- wihs$idu<\/p>\n

cif.haart <- cbind(apply(cif[, , 1][idu == 0, ], 2, mean), apply(cif[, , 1][idu ==<\/p>\n

1, ], 2, mean))<\/p>\n

cif.aids <- cbind(apply(cif[, , 2][idu == 0, ], 2, mean), apply(cif[, , 2][idu ==<\/p>\n

1, ], 2, mean))<\/p>\n

matplot(Time, cbind(cif.haart, cif.aids), type = “l”, lty = c(1, 2, 1, 2), col = c(4,<\/p>\n

4, 2, 2), lwd = 3, ylab = “Cumulative Incidence”)<\/p>\n

legend(“bottomright”, legend = c(“HAART (Non-IDU)”, “HAART (IDU)”, “AIDS (Non-IDU)”,<\/p>\n

“AIDS (IDU)”), lty = c(1, 2, 1, 2), col = c(4, 4, 2, 2), lwd = 3, cex = 0.6)#\u56fe\u6ce8<\/p>\n

<\/p>\n

\u5c0f\u4f19\u4f34\u53ef\u80fd\u4e0d\u592a\u61c2\u7ed3\u679c\u7684\u542b\u4e49\uff0c\u5c0f\u679c\u8fd9\u91cc\u7b80\u5355\u89e3\u8bfb\u4e00\u4e9b<\/p>\n

\u968f\u673a\u751f\u5b58\u68ee\u6797\u53ef\u4ee5\u5bf9\u53d8\u91cf\u91cd\u8981\u6027\u8fdb\u884c\u6392\u540d\uff0cVIMP\u6cd5\u548c\u6700\u5c0f\u6df1\u5ea6\u6cd5\u662f\u6700\u5e38\u7528\u7684\u65b9\u6cd5\uff1a\u53d8\u91cfVIMP\u503c\u5c0f\u4e8e0\u8bf4\u660e\u8be5\u53d8\u91cf\u964d\u4f4e\u4e86\u9884\u6d4b\u7684\u51c6\u786e\u6027\uff0c\u800c\u5f53VIMP\u503c\u5927\u4e8e0\u5219\u8bf4\u660e\u8be5\u53d8\u91cf\u63d0\u9ad8\u4e86\u9884\u6d4b\u7684\u51c6\u786e\u6027\uff1b\u6700\u5c0f\u6df1\u5ea6\u6cd5\u901a\u8fc7\u8ba1\u7b97\u8fd0\u884c\u5230\u6700\u7ec8\u8282\u70b9\u65f6\u7684\u6700\u5c0f\u6df1\u5ea6\u6765\u7ed9\u51fa\u5404\u53d8\u91cf\u5bf9\u4e8e\u7ed3\u5c40\u4e8b\u4ef6\u7684\u91cd\u8981\u6027\u3002<\/p>\n

\u4e0a\u8ff0\u4e3a\u7efc\u5408\u4e24\u79cd\u65b9\u6cd5\u7684\u6563\u70b9\u56fe\uff0c\u5176\u4e2d\uff0c\u84dd\u8272\u70b9\u4ee3\u8868VIMP\u503c\u5927\u4e8e0\uff0c\u7ea2\u8272\u5219\u4ee3\u8868VIMP\u503c\u5c0f\u4e8e0\uff1b\u5728\u7ea2\u8272\u5bf9\u89d2\u865a\u7ebf\u4e0a\u7684\u70b9\u4ee3\u8868\u4e24\u79cd\u65b9\u6cd5\u5bf9\u8be5\u53d8\u91cf\u7684\u6392\u540d\u76f8\u540c\uff0c\u9ad8\u4e8e\u5bf9\u89d2\u865a\u7ebf\u7684\u70b9\u4ee3\u8868\u5176VIMP\u6392\u540d\u66f4\u9ad8\uff0c\u4f4e\u4e8e\u5bf9\u89d2\u865a\u7ebf\u7684\u70b9\u5219\u4ee3\u8868\u5176\u6700\u5c0f\u6df1\u5ea6\u6392\u540d\u66f4\u9ad8\u3002\u76f8\u8f83\u4e8eCox\u6bd4\u4f8b\u98ce\u9669\u56de\u5f52\u6a21\u578b\u7b49\u4f20\u7edf\u751f\u5b58\u5206\u6790\u65b9\u6cd5\uff0c\u968f\u673a\u751f\u5b58\u68ee\u6797\u6a21\u578b\u7684\u9884\u6d4b\u51c6\u786e\u5ea6\u81f3\u5c11\u7b49\u540c\u6216\u4f18\u4e8e\u4f20\u7edf\u751f\u5b58\u5206\u6790\u65b9\u6cd5\u3002<\/p>\n

\u968f\u673a\u68ee\u6797\u5462\u5176\u5b9e\u6709\u5f88\u591a\u7684\u4f18\u52bf\uff0c\u5c0f\u679c\u8fd9\u91cc\u5c31\u4e0d\u4e00\u4e00\u4ecb\u7ecd\u4e86\uff0c\u5c0f\u4f19\u4f34\u5b66\u4f1a\u4e86\u5417\uff0c\u5c06\u81ea\u5df1\u7684\u6570\u636e\u53bb\u505a\u4e2a\u5b9e\u9a8c\uff0c\u52a8\u624b\u8bd5\u8bd5\u8fd9\u4e2a\u65b0\u65b9\u6cd5\u5427\uff01<\/p>\n

\u5c0f\u4f19\u4f34\u591a\u591a\u7406\u89e3\u4ee3\u7801\u7684\u610f\u4e49\u54e6~<\/p>\n","protected":false},"excerpt":{"rendered":"

\u5c0f\u4f19\u4f34\u505a\u751f\u5b58\u5206\u6790\u7684\u65f6\u5019\u907f\u514d\u4e0d\u4e86\u6784\u5efaCox\u56de\u5f52\u98ce\u9669\u6a21\u578b\uff0c\u5c0f\u679c\u544a\u8bc9\u5927\u5bb6\u4e00\u4e2a\u65b0\u7684\u65b9\u5f0f\uff0c\u5c0f\u4f19\u4f34\u53ef\u4ee5\u4e0d\u518d\u7528Cox\u56de\u5f52\u4e86\uff0c […]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"jetpack_post_was_ever_published":false,"_jetpack_newsletter_access":""},"categories":[1],"tags":[],"jetpack_featured_media_url":"","_links":{"self":[{"href":"http:\/\/www.biocloudservice.com\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/27639"}],"collection":[{"href":"http:\/\/www.biocloudservice.com\/wordpress\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/www.biocloudservice.com\/wordpress\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/www.biocloudservice.com\/wordpress\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"http:\/\/www.biocloudservice.com\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=27639"}],"version-history":[{"count":1,"href":"http:\/\/www.biocloudservice.com\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/27639\/revisions"}],"predecessor-version":[{"id":27657,"href":"http:\/\/www.biocloudservice.com\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/27639\/revisions\/27657"}],"wp:attachment":[{"href":"http:\/\/www.biocloudservice.com\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=27639"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/www.biocloudservice.com\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=27639"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/www.biocloudservice.com\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=27639"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}