error explain function

[JLB9791]

Hi Thomas,

great package! Thanks for that. I get a weird error that I can't seem to solve on the explain function. I followed all the steps from https://shiring.github.io/machine_learning/2017/04/23/lime on my data with the necessary changes of course to align with my work. I've build a classifier model with CARET and xgbtree

### --- try to explaining complex algorithms with lime function
explainer <- lime(df_train, mod_fit, bin_continuous = TRUE, n_bins = 5, n_permutations = 1000)

# select randomly 3 cases where the prediction was correct and 3 cases where the prediction was wrong
pred <- data.frame(sample_id = 1:nrow(df_test),
                   predict(mod_fit, df_test, type = "prob"),
                   actual = df_test[, y])
  pred$prediction <- colnames(pred)[2:4][apply(pred[, 2:4], 1, which.max)]
  pred$correct <- ifelse(pred$actual == pred$prediction, "correct", "wrong")

# Beware that we need to give our test-set data table row names with the sample names 
# or IDs to be displayed in the header of our explanatory plots below.
pred_cor <- filter(pred, correct == "correct")
pred_wrong <- filter(pred, correct == "wrong")

df_test_cor <- df_test %>%
  mutate(sample_id = 1:nrow(df_test)) %>%
  filter(sample_id %in% pred_cor$sample_id) %>%
  sample_n(size = 3) %>%
  remove_rownames() %>%
  tibble::column_to_rownames(var = "sample_id") %>%
  select(-bought_highbrow_wines)

df_test_wrong <- df_test %>%
  mutate(sample_id = 1:nrow(df_test)) %>%
  filter(sample_id %in% pred_wrong$sample_id) %>%
  sample_n(size = 3) %>%
  remove_rownames() %>%
  tibble::column_to_rownames(var = "sample_id") %>%
  select(-bought_highbrow_wines)

# to keep it readable select top x features for graphical display. Features are selected with highest weight, however other options are possible.
explanation_cor <- lime::explain(df_test_cor, explainer, n_labels=2, n_features = 5) 

the last line throws me the following error: Error in (function (..., row.names = NULL, check.rows = FALSE, check.names = TRUE, : arguments imply differing number of rows: 15000, 0

I any suggestions?

thx JL

[thomasp85]

Not really... if you provide a reproducible example I can look into it

[JLB9791]

Hi Thomas, the issue is solved. It was due to not finding the correct version of udunits2 on linux redhat server. Nothing to do with your package. Sorry for the inconvenience.