All shap values were NA.

[hyr13579]

library(fastshap)
train_data <- as.data.frame(train_data)
shap <- explain(
rf,
X = as.data.frame(train_data),
nsim = 1,
adjust = FALSE,
shap_only = TRUE,
pred_wrapper = function(model,newdata){
as.data.frame(predict(model,newdata) ) %>% pull(1)
}
)

When I run the above code, all the shap values I get are NA.There are a lot of warniings.
Warnings ( )
1: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
2: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
3: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
4: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
5: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
6: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
7: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
8: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
9: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
10: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
11: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
12: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
13: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
14: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
15: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
16: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
17: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
18: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
19: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
20: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
21: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
22: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
23: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
24: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
25: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
26: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
27: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
28: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
29: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
30: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
31: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
32: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
33: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
34: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
35: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
36: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
37: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
38: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
39: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
40: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
41: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
42: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
43: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
44: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
45: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
46: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
47: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
48: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
49: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors
50: In Ops.factor(pred_wrapper(object, newdata = B[[1L]]), ... :
‘-’ not meaningful for factors

[hyr13579]

rf is a Multiclass classification model，Can you tell me how to calculate the shap value? Thank you vary much！

[bgreenwell]

Hi @hyr13579 thank you for posting an issue. However, without a reproducible example to run on my end, I can’t do much to help diagnose your issue. Please post a fully reproducible example and consider using the reprex package.

[hyr13579]

Thank you. I changed the code. It worked.
shap<- explain(rf,
X = as.data.frame(train_data),
nsim = 100,
pred_wrapper = function(model, newdata){
as.data.frame(predict(model, newdata, type = "prob")) %>% pull(1)})
When I run the above code, all the shap values I get are number.

[bgreenwell]

Glad you got it to work!