cat_imputer = SimpleImputer(missing_values=np.nan, strategy="constant", fill_value="None")

cat_transforms = Pipeline(steps=[("impute", cat_imputer), ("encode", enc)])

ct = ColumnTransformer([("cat", cat_transforms, cat_cols), ("num", num_imputer, num_cols)])

"# Separating data into categorical and non-categorical (numeric for this example) columns\ncat_cols = list_categorical_attributes(flow_type=flow_type)\nnum_cols = list(set(X.columns) - set(cat_cols))\n\n# Missing value imputers\ncat_imputer = SimpleImputer(missing_values=np.nan, strategy=\"constant\", fill_value=\"None\")\nnum_imputer = SimpleImputer(missing_values=np.nan, strategy=\"constant\", fill_value=-1)\n\n# Creating the one-hot encoder\nenc = OneHotEncoder(handle_unknown=\"ignore\")\n\n# Pipeline to handle categorical column transformations\ncat_transforms = Pipeline(steps=[(\"impute\", cat_imputer), (\"encode\", enc)])\n\n# Combining column transformers\nct = ColumnTransformer([(\"cat\", cat_transforms, cat_cols), (\"num\", num_imputer, num_cols)])\n\n# Creating the full pipeline with the surrogate model\nclf = RandomForestRegressor(n_estimators=50)\nmodel = Pipeline(steps=[(\"preprocess\", ct), (\"surrogate\", clf)])"

cat_imp = make_pipeline(

SimpleImputer(strategy="most_frequent"),

OneHotEncoder(handle_unknown="ignore", sparse=False),

TruncatedSVD(),

)

ct = ColumnTransformer([("cat", cat_imp, cat), ("cont", "passthrough", cont)])

import numpy as np

clf = tree.ExtraTreeClassifier()

task = openml.tasks.get_task(1)

pipeline.Pipeline(

[

("Imputer", impute.SimpleImputer(strategy="most_frequent")),

(

"Encoder",

preprocessing.OneHotEncoder(

sparse=False, handle_unknown="ignore"

),

),

]

),

("continuous", "passthrough", cont), # returns the numeric feature indices

pipeline.Pipeline(

[

("Imputer", impute.SimpleImputer(strategy="most_frequent")),

(

"Encoder",

preprocessing.OneHotEncoder(

sparse=False, handle_unknown="ignore"

),

),

]

),

("continuous", "passthrough", numeric_feature_indices),

run = openml.runs.run_model_on_task(pipe, task, avoid_duplicate_runs=False, upload_flow=False)

"# License: BSD 3-Clause\n\nimport openml\nimport numpy as np\nfrom sklearn import compose, ensemble, impute, neighbors, preprocessing, pipeline, tree"

"# Get a task\ntask = openml.tasks.get_task(403)\n\n# Build any classifier or pipeline\nclf = tree.ExtraTreeClassifier()\n\n# Run the flow\nrun = openml.runs.run_model_on_task(clf, task)\n\nprint(run)"

"### It also works with pipelines\n\nWhen you need to handle 'dirty' data, build pipelines to model then automatically.\n\n"

"task = openml.tasks.get_task(1)\n\n# OpenML helper functions for sklearn can be plugged in directly for complicated pipelines\nfrom openml.extensions.sklearn import cat, cont\n\npipe = pipeline.Pipeline(\n steps=[\n (\n \"Preprocessing\",\n compose.ColumnTransformer(\n [\n (\n \"categorical\",\n pipeline.Pipeline(\n [\n (\"Imputer\", impute.SimpleImputer(strategy=\"most_frequent\")),\n (\n \"Encoder\",\n preprocessing.OneHotEncoder(\n sparse=False, handle_unknown=\"ignore\"\n ),\n ),\n ]\n ),\n cat, # returns the categorical feature indices\n ),\n (\"continuous\", \"passthrough\", cont), # returns the numeric feature indices\n ]\n ),\n ),\n (\"Classifier\", ensemble.RandomForestClassifier(n_estimators=10)),\n ]\n)\n\nrun = openml.runs.run_model_on_task(pipe, task, avoid_duplicate_runs=False)\nmyrun = run.publish()\nprint(\"Uploaded to http://test.openml.org/r/\" + str(myrun.run_id))\n\n\n# The above pipeline works with the helper functions that internally deal with pandas DataFrame.\n# In the case, pandas is not available, or a NumPy based data processing is the requirement, the\n# above pipeline is presented below to work with NumPy.\n\n# Extracting the indices of the categorical columns\nfeatures = task.get_dataset().features\ncategorical_feature_indices = []\nnumeric_feature_indices = []\nfor i in range(len(features)):\n if features[i].name == task.target_name:\n continue\n if features[i].data_type == \"nominal\":\n categorical_feature_indices.append(i)\n else:\n numeric_feature_indices.append(i)\n\npipe = pipeline.Pipeline(\n steps=[\n (\n \"Preprocessing\",\n compose.ColumnTransformer(\n [\n (\n \"categorical\",\n pipeline.Pipeline(\n [\n (\"Imputer\", impute.SimpleImputer(strategy=\"most_frequent\")),\n (\n \"Encoder\",\n preprocessing.OneHotEncoder(\n sparse=False, handle_unknown=\"ignore\"\n ),\n ),\n ]\n ),\n categorical_feature_indices,\n ),\n (\"continuous\", \"passthrough\", numeric_feature_indices),\n ]\n ),\n ),\n (\"Classifier\", ensemble.RandomForestClassifier(n_estimators=10)),\n ]\n)\n\nrun = openml.runs.run_model_on_task(pipe, task, avoid_duplicate_runs=False, dataset_format=\"array\")\nmyrun = run.publish()\nprint(\"Uploaded to http://test.openml.org/r/\" + str(myrun.run_id))"

"# To perform the following line offline, it is required to have been called before\n# such that the task is cached on the local openml cache directory:\ntask = openml.tasks.get_task(6)\n\n# The following lines can then be executed offline:\nrun = openml.runs.run_model_on_task(pipe, task, avoid_duplicate_runs=False, upload_flow=False)\n\n# The run may be stored offline, and the flow will be stored along with it:\nrun.to_filesystem(directory=\"myrun\")\n\n# They may be loaded and uploaded at a later time\nrun = openml.runs.OpenMLRun.from_filesystem(directory=\"myrun\")\nrun.publish()\n\n# Publishing the run will automatically upload the related flow if\n# it does not yet exist on the server."

"# first, let's download the task that we are interested in\ntask = openml.tasks.get_task(6)\n\n\n# we will create a fairly complex model, with many preprocessing components and\n# many potential hyperparameters. Of course, the model can be as complex and as\n# easy as you want it to be\n\n\ncat_imp = make_pipeline(\n SimpleImputer(strategy=\"most_frequent\"),\n OneHotEncoder(handle_unknown=\"ignore\", sparse=False),\n TruncatedSVD(),\n)\nct = ColumnTransformer([(\"cat\", cat_imp, cat), (\"cont\", \"passthrough\", cont)])\nmodel_original = Pipeline(steps=[(\"transform\", ct), (\"estimator\", RandomForestClassifier()),])\n\n# Let's change some hyperparameters. Of course, in any good application we\n# would tune them using, e.g., Random Search or Bayesian Optimization, but for\n# the purpose of this tutorial we set them to some specific values that might\n# or might not be optimal\nhyperparameters_original = {\n \"estimator__criterion\": \"gini\",\n \"estimator__n_estimators\": 50,\n \"estimator__max_depth\": 10,\n \"estimator__min_samples_leaf\": 1,\n}\nmodel_original.set_params(**hyperparameters_original)\n\n# solve the task and upload the result (this implicitly creates the flow)\nrun = openml.runs.run_model_on_task(model_original, task, avoid_duplicate_runs=False)\nrun_original = run.publish() # this implicitly uploads the flow"