xgboost 错误预测 R

Question

I am using xgboost to predict airbnb destinations (similar to the Kaggle competition but for a class project). However when running the prediction command I receive this error message:

Error in predict.xgb.Booster(bst, dval) : Feature names stored in object and newdata are different!

How can I fix this problem?

Here is my code:

    setwd("~/Documents/Big Data/Datasets-20180304")
airbnb <- read.csv("airbnb_train.csv", header = T, stringsAsFactors = F)
airbnb_test <- read.csv("airbnb_test.csv", header = T, stringsAsFactors = F)
airbnb <- na.omit(airbnb)
airbnb_test <- na.omit(airbnb_test)
airbnb$country_destination <- as.factor(airbnb$country_destination)

airbnb$country_destination[airbnb$country_destination==0] <- NA
airbnb$country_destination <- recode(airbnb$country_destination, "c('1') = '0'; c('2') = '1'")
airbnb <- na.omit(airbnb)
airbnb_test <- na.omit(airbnb_test)

set.seed(1234)
train_index <- sample(1:nrow(airbnb),size = 0.7*nrow(airbnb),replace = F)
train <- airbnb[train_index,]
validation <- airbnb[-train_index,]

options(na.action='na.pass')
new_tr = sparse.model.matrix(country_destination~.-1,data = train, with = F)
train_label <- train$country_destination
train_label <- as.numeric(train_label)-1
dtrain <- xgb.DMatrix(data = new_tr, label=train_label)


new_val = sparse.model.matrix(country_destination~.-1,data = validation, with = F)
val_label <- validation$country_destination
val_label <- as.numeric(val_label)-1
dval <- xgb.DMatrix(data = new_val, label=val_label)

#default parameters
params <- list(
  booster = "gbtree",
  objective = "binary:logistic",
  eta=0.3,
  gamma=0,
  max_depth=6,
  min_child_weight=1,
  subsample=1,
  colsample_bytree=1
)

bst <- xgboost(data = dtrain, label = train_label, max_depth = 2, eta = 1, nthread = 2, nrounds = 8, objective = "binary:logistic")

xgbpred <- predict(bst,dval)

What am I doing wrong? How can I ensure that both 'bst' and 'dval' have the same feature_names?

Answer 1

Can you share what are your names(bst) and names(dval) ? After you applying boosting model:

bst <- xgboost(data = dtrain, label = train_label, max_depth = 2, eta = 1, nthread = 2, nrounds = 8, objective = "binary:logistic")

As a workaround you could simply do:

names(bst) <- names(dval)

and then try your prediction:

xgbpred <- predict(bst,dval)

Answer 2

I was stuck with similar problem and this worked for me.

Try removing predicting variable, ie train$country_destination in your case, from 'dtrain' and 'dtest' (even if you have blank values filled in there). Try training the model again after making that change.

Answer 3

If you look at this page ( https://rdrr.io/cran/xgboost/src/R/xgb.Booster.R ), you will see that some R users are likely to get the following error message: "Feature names stored in object and newdata are different!".

Here is the code from this page related to the error message:

predict.xgb.Booster <- function(object, newdata, missing = NA, outputmargin = FALSE, ntreelimit = NULL,predleaf = FALSE, predcontrib = FALSE, approxcontrib = FALSE, predinteraction = FALSE,reshape = FALSE, ...)

object <- xgb.Booster.complete(object, saveraw = FALSE)
      if (!inherits(newdata, "xgb.DMatrix"))
        newdata <- xgb.DMatrix(newdata, missing = missing)
      if (!is.null(object[["feature_names"]]) &&
          !is.null(colnames(newdata)) &&
          !identical(object[["feature_names"]], colnames(newdata)))
        stop("Feature names stored in `object` and `newdata` are different!")

identical(object[["feature_names"]], colnames(newdata)) => If the column names of object (ie your model based on your training set) are not identical to the column names of newdata (ie your test set), you will get the error message.

For more details:

train_matrix <- xgb.DMatrix(as.matrix(training %>% select(-target)), label = training$target, missing = NaN)
object <- xgb.train(data=train_matrix, params=..., nthread=2, nrounds=..., prediction = T)
newdata <- xgb.DMatrix(as.matrix(test %>% select(-target)), missing = NaN)

While setting by yourself object and newdata with your data thanks to the code above, you can probably fix this issue by looking at the differences between object[["feature_names"]] and colnames(newdata) . Probably some columns that don't appear in the same order or something.

Answer 4

Following guiotan answers, using

library(dplyr)

You should be able to write:

xgbpred <- predict(bst, dval %>% select(bst$feature_names))

If you trained xgboost using caret , a solution would be to write the following.

xgbpred <- predict(bst, dval %>% select(bst$finalModel$feature_names))

At least this worked for me.