Simple pipeline | Basic Feature Engineering #990

Paulitos · 2024-01-07T00:04:14Z

Paulitos
Jan 7, 2024

Hi, I'm trying to do this code in R (scikit-learn code provided):

from sklearn.pipeline import Pipeline
from sklearn.impute import SimpleImputer
from sklearn.preprocessing import StandardScaler
from sklearn.compose import ColumnTransformer
from sklearn.preprocessing import OrdinalEncoder
from sklearn.preprocessing import OneHotEncoder

num_pipeline = Pipeline([
    ('imputer', SimpleImputer(strategy='median')),
    ('std_scaler', StandardScaler())
])

cat_pipeline = Pipeline([
    ('oh_encoder', OneHotEncoder())
])

full_pipeline = ColumnTransformer([
    ('num', num_pipeline, num_attribs),
    ('cat', cat_pipeline, cat_attribs)
])

# non-selected columns such as 'PassengerId' and 'Cabin' will be
# dropped after passing through the column transformer
X_train_copy_transformed = df_transform(X_train_copy)
print(X_train_copy_transformed.columns)
X_train_copy_prepared = full_pipeline.fit_transform(X_train_copy_transformed)
print(X_train_copy_prepared[0])

What I have so far is this:

library(mlr3)
library(mlr3pipelines)

num_pipeline <- po("imputemedian") %>>% po("scale", center = TRUE, scale = TRUE)
cat_pipeline <- po("encode", method = "one-hot")

full_pipeline <- po("pipe", list(
  po("select", num_attribs) %>>% num_pipeline,
  po("select", cat_attribs) %>>% cat_pipeline
))

X_train_copy_transformed <- df_transform(X_train_copy)
print(colnames(X_train_copy_transformed))
X_train_copy_prepared <- full_pipeline$train(list(data = list(x = X_train_copy_transformed)))$predict(list(x = X_train_copy_transformed))
print(X_train_copy_prepared[1, ])

The error has to do with Error in .__PipeOp__initialize(self = self, private = private, super = super, : Assertion on 'id' failed: Must have length 1. but I don't really understand the select function, I think is because I still have problems with oop. Anyways, if you want to try the notebook I'm working with here the notebook for you to try with the data and deal with the errors. Any help would be appreciated.

https://www.kaggle.com/code/paulitos/spaceship-titanic-in-r-paulos

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Simple pipeline | Basic Feature Engineering #990

{{title}}

Replies: 0 comments

Select a reply

Simple pipeline | Basic Feature Engineering #990

Paulitos Jan 7, 2024

Replies: 0 comments

Paulitos
Jan 7, 2024