r - For循环交叉验证R

Question

我正在尝试创建一个循环（执行 10 次）以重复交叉验证以评估 4 个模型的预测性能，然后我必须计算我的性能的平均值。我必须承认我是 R 的新手，我正在为这个简单的任务而苦苦挣扎。我首先创建了我的公式来交叉验证我的模型。

loss.mse <- function(fit, df, y, transf){
  y_pred <- transf(predict(fit, df))
  out <- (y - y_pred)^2
  return(mean(out))
}
loss.mae <- function(fit, df, y, transf){
  y_pred <- transf(predict(fit, df))
  out <- abs(y - y_pred)
  return(mean(out))
}
validate.cv <- function(data, folds, model_fn, y_var, 
                        transf = identity, seed) 
{
  set.seed(seed)
  fold_id <- sample(rep(1:folds, length.out = nrow(data)))
  out.mse <- out.mae <- numeric(folds)
  for(test in 1:folds){
    data_test <- subset(data, fold_id == test)
    data_train <- subset(data, fold_id != test)
    fit <- model_fn(data_train)
    y_test <- y_var[fold_id == test]
    out.mse[test] <- loss.mse(fit, data_test, y_test, transf) 
    out.mae[test] <- loss.mae(fit, data_test, y_test, transf)  
  }
  return(list(MAE = mean(out.mae), MSE = out.mse, 
              RMSE = sqrt(mean(out.mse))))

然后我命名了我的模型并交叉验证了它们，但我不知道如何获得我的 10 次循环！

model_lm <- function(data) lm(StockPrice ~., data)

model_step <-  function(data) step(lm(StockPrice ~., data), trace = 0)

model_rpart <- function(data)
{
  set.seed(1234)
  mod.rpart <- rpart(StockPrice ~ ., data, cp = 0.0001, model = TRUE)
  cp.select <- function(big.tree) { 
  min.x <- which.min(big.tree$cptable[, 4])
  for(i in 1:nrow(big.tree$cptable)) {
    if(big.tree$cptable[i, 4] < 
       (big.tree$cptable[min.x, 4] + big.tree$cptable[min.x, 5])) 
      return(big.tree$cptable[i, 1]) 
       }
   }
  mod.rpart.prune <- prune(mod.rpart, cp = cp.select(mod.rpart))
  return(mod.rpart.prune)
}  

model_step_gam <- function(data)
{
  mod  <- model_step(data)
  predictors <-  all.vars(terms(mod))[-1]
  f <- as.formula(
  paste("StockPrice", 
        paste(paste("s(",predictors, ")"), collapse = " + "), 
        sep = " ~ "))
  mod_gam <- gam(f, data = data)

seed<-1234
m.log.full <-validate.cv(log.Finance, 10, model_lm, Finance$StockPrice,exp, seed)
m.log.step <-validate.cv(log.Finance, 10, model_step, Finance$StockPrice,exp, seed)
m.log.rpart <-validate.cv(log.Finance, 10, model_rpart, Finance$StockPrice,exp, seed)
m.log.gam <-validate.cv(log.Finance, 10, model_step_gam, Finance$StockPrice,exp, seed)
mat.test <-data.frame(Model =c("Full (log)","Step (log)","CART (log)", "Step GAM (log)"),
                      RMSE =c(m.log.full$RMSE, m.log.step$RMSE,m.log.rpart$RMSE, m.log.gam$RMSE),
                      MAE =c(m.log.full$MAE, m.log.step$MAE, m.log.rpart$MAE,m.log.gam$MAE))

print(mat.test)

如果您有任何想法，我很乐意尝试。预先感谢您的帮助 ：）

标签： rfor-loopcross-validation