Drug classification [scikit-learn]¶

Giskard is an open-source framework for testing all ML models, from LLMs to tabular models. Don’t hesitate to give the project a star on GitHub ⭐️ if you find it useful!

In this notebook, you’ll learn how to create comprehensive test suites for your model in a few lines of code, thanks to Giskard’s open-source Python library.

Use-case:

Multinomial classification of a drug’s type.
Reference notebook
Dataset

Outline:

Detect vulnerabilities automatically with Giskard’s scan
Automatically generate & curate a comprehensive test suite to test your model beyond accuracy-related metrics

Install dependencies¶

Make sure to install the giskard

[ ]:

%pip install giskard --upgrade

Import libraries¶

[1]:

from pathlib import Path
from urllib.request import urlretrieve

import numpy as np
import pandas as pd
from sklearn.svm import SVC
from imblearn.over_sampling import SMOTE
from sklearn.metrics import accuracy_score
from sklearn.preprocessing import OneHotEncoder
from sklearn.model_selection import train_test_split
from imblearn.pipeline import Pipeline as PipelineImb

from giskard import Dataset, Model, scan, testing

Define constants¶

[2]:

# Constants.
RANDOM_SEED = 0

TARGET_NAME = "Drug"

AGE_BINS = [0, 19, 29, 39, 49, 59, 69, 80]
AGE_CATEGORIES = ['<20s', '20s', '30s', '40s', '50s', '60s', '>60s']

NA_TO_K_BINS = [0, 9, 19, 29, 50]
NA_TO_K_CATEGORIES = ['<10', '10-20', '20-30', '>30']

# Paths.
DATA_URL = "ftp://sys.giskard.ai/pub/unit_test_resources/drug_classification_dataset/drug200.csv"
DATA_PATH = Path.home() / ".giskard" / "drug_classification_dataset" / "drug200.csv"

Dataset preparation¶

Load and preprocess data¶

[3]:

def fetch_from_ftp(url: str, file: Path) -> None:
    """Helper to fetch data from the FTP server."""
    if not file.parent.exists():
        file.parent.mkdir(parents=True, exist_ok=True)

    if not file.exists():
        print(f"Downloading data from {url}")
        urlretrieve(url, file)

    print(f"Data was loaded!")


def load_data() -> pd.DataFrame:
    """Load data."""
    fetch_from_ftp(DATA_URL, DATA_PATH)
    df = pd.read_csv(DATA_PATH)
    return df


def bin_numerical(df: pd.DataFrame) -> np.ndarray:
    """Perform numerical features binning."""

    def _bin_age(_df: pd.DataFrame) -> pd.DataFrame:
        """Bin age feature."""
        _df.Age = pd.cut(_df.Age, bins=AGE_BINS, labels=AGE_CATEGORIES)
        return _df

    def _bin_na_to_k(_df: pd.DataFrame) -> pd.DataFrame:
        """Bin Na_to_K feature."""
        _df.Na_to_K = pd.cut(_df.Na_to_K, bins=NA_TO_K_BINS, labels=NA_TO_K_CATEGORIES)
        return _df

    df = df.copy()
    df = _bin_age(df)
    df = _bin_na_to_k(df)

    return df

[ ]:

df_drug = load_data()
df_drug = bin_numerical(df_drug)

Train-test split¶

[5]:

X_train, X_test, y_train, y_test = train_test_split(df_drug.drop(TARGET_NAME, axis=1), df_drug[TARGET_NAME],
                                                    test_size=0.5, random_state=RANDOM_SEED)

Wrap dataset with Giskard¶

To prepare for the vulnerability scan, make sure to wrap your dataset using Giskard’s Dataset class. More details here.

[ ]:

raw_dataset = pd.concat([X_test, y_test], axis=1)
giskard_dataset = Dataset(
    df=raw_dataset,  # A pandas.DataFrame that contains the raw data (before all the pre-processing steps) and the actual ground truth variable (target).
    target=TARGET_NAME,  # Ground truth variable.
    name="Drug classification dataset",  # Optional.
    cat_columns=X_test.columns.tolist()  # List of categorical columns. Optional, but is a MUST if available. Inferred automatically if not.
)

Model building¶

Build estimator¶

[ ]:

pipeline = PipelineImb(steps=[
    ("one_hot_encoder", OneHotEncoder()),
    ("resampler", SMOTE(random_state=RANDOM_SEED)),
    ("classifier", SVC(kernel='linear', random_state=RANDOM_SEED, probability=True))
])

pipeline.fit(X_train, y_train)

y_train_pred = pipeline.classes_[pipeline.predict_proba(X_train).argmax(axis=1)]
y_test_pred = pipeline.classes_[pipeline.predict_proba(X_test).argmax(axis=1)]
train_metric = accuracy_score(y_train, y_train_pred)
test_metric = accuracy_score(y_test, y_test_pred)

print(f"Train accuracy score: {train_metric:.2f}\n"
      f"Test accuracy score: {test_metric:.2f}")

Wrap model with Giskard¶

To prepare for the vulnerability scan, make sure to wrap your model using Giskard’s Model class. You can choose to either wrap the prediction function (preferred option) or the model object. More details here.

[ ]:

def prediction_function(df: pd.DataFrame) -> np.ndarray:
    return pipeline.predict_proba(df)


wrapped_model = Model(
    model=prediction_function,  # A prediction function that encapsulates all the data pre-processing steps and that could be executed with the dataset used by the scan.
    model_type="classification",  # Either regression, classification or text_generation.
    name="Drug classifier",  # Optional.
    classification_labels=pipeline.classes_,  # Their order MUST be identical to the prediction_function's output order.
    feature_names=X_test.columns  # Default: all columns of your dataset.
)

# Validate wrapped model.
wrapped_y_test_pred = wrapped_model.predict(giskard_dataset).prediction
wrapped_test_metric = accuracy_score(y_test, wrapped_y_test_pred)
print(f"Wrapped Test accuracy score: {wrapped_test_metric:.2f}")

Detect vulnerabilities in your model¶

Scan your model for vulnerabilities with Giskard¶

Giskard’s scan allows you to detect vulnerabilities in your model automatically. These include performance biases, unrobustness, data leakage, stochasticity, underconfidence, ethical issues, and more. For detailed information about the scan feature, please refer to our scan documentation.

[ ]:

results = scan(wrapped_model, giskard_dataset)

[10]:

display(results)

Generate comprehensive test suites automatically for your model¶

Generate test suites from the scan¶

The objects produced by the scan can be used as fixtures to generate a test suite that integrate all detected vulnerabilities. Test suites allow you to evaluate and validate your model’s performance, ensuring that it behaves as expected on a set of predefined test cases, and to identify any regressions or issues that might arise during development or updates.

[11]:

test_suite = results.generate_test_suite("My first test suite")
test_suite.run()

2024-05-29 11:46:34,990 pid:52758 MainThread giskard.datasets.base INFO     Casting dataframe columns from {'Age': 'category', 'Sex': 'object', 'BP': 'object', 'Cholesterol': 'object', 'Na_to_K': 'category'} to {'Age': 'category', 'Sex': 'object', 'BP': 'object', 'Cholesterol': 'object', 'Na_to_K': 'category'}
2024-05-29 11:46:34,993 pid:52758 MainThread giskard.utils.logging_utils INFO     Predicted dataset with shape (22, 6) executed in 0:00:00.009771
Executed 'Precision on data slice “`Age` == "30s"”' with arguments {'model': <giskard.models.function.PredictionFunctionModel object at 0x178523b80>, 'dataset': <giskard.datasets.base.Dataset object at 0x1785cd450>, 'slicing_function': <giskard.slicing.slice.QueryBasedSliceFunction object at 0x1787d4c70>, 'threshold': 0.76}:
               Test failed
               Metric: 0.68


2024-05-29 11:46:35,010 pid:52758 MainThread giskard.datasets.base INFO     Casting dataframe columns from {'Age': 'category', 'Sex': 'object', 'BP': 'object', 'Cholesterol': 'object', 'Na_to_K': 'category'} to {'Age': 'category', 'Sex': 'object', 'BP': 'object', 'Cholesterol': 'object', 'Na_to_K': 'category'}
2024-05-29 11:46:35,011 pid:52758 MainThread giskard.utils.logging_utils INFO     Predicted dataset with shape (33, 6) executed in 0:00:00.007762
Executed 'Precision on data slice “`BP` == "NORMAL"”' with arguments {'model': <giskard.models.function.PredictionFunctionModel object at 0x178523b80>, 'dataset': <giskard.datasets.base.Dataset object at 0x1785cd450>, 'slicing_function': <giskard.slicing.slice.QueryBasedSliceFunction object at 0x1787b2b90>, 'threshold': 0.76}:
               Test failed
               Metric: 0.73


2024-05-29 11:46:35,023 pid:52758 MainThread giskard.datasets.base INFO     Casting dataframe columns from {'Age': 'category', 'Sex': 'object', 'BP': 'object', 'Cholesterol': 'object', 'Na_to_K': 'category'} to {'Age': 'category', 'Sex': 'object', 'BP': 'object', 'Cholesterol': 'object', 'Na_to_K': 'category'}
2024-05-29 11:46:35,025 pid:52758 MainThread giskard.utils.logging_utils INFO     Predicted dataset with shape (59, 6) executed in 0:00:00.006482
Executed 'Precision on data slice “`Na_to_K` == "10-20"”' with arguments {'model': <giskard.models.function.PredictionFunctionModel object at 0x178523b80>, 'dataset': <giskard.datasets.base.Dataset object at 0x1785cd450>, 'slicing_function': <giskard.slicing.slice.QueryBasedSliceFunction object at 0x1787b16c0>, 'threshold': 0.76}:
               Test failed
               Metric: 0.73


2024-05-29 11:46:35,034 pid:52758 MainThread giskard.datasets.base INFO     Casting dataframe columns from {'Age': 'category', 'Sex': 'object', 'BP': 'object', 'Cholesterol': 'object', 'Na_to_K': 'category'} to {'Age': 'category', 'Sex': 'object', 'BP': 'object', 'Cholesterol': 'object', 'Na_to_K': 'category'}
2024-05-29 11:46:35,036 pid:52758 MainThread giskard.utils.logging_utils INFO     Predicted dataset with shape (53, 6) executed in 0:00:00.006638
Executed 'Precision on data slice “`Cholesterol` == "HIGH"”' with arguments {'model': <giskard.models.function.PredictionFunctionModel object at 0x178523b80>, 'dataset': <giskard.datasets.base.Dataset object at 0x1785cd450>, 'slicing_function': <giskard.slicing.slice.QueryBasedSliceFunction object at 0x1787b0b80>, 'threshold': 0.76}:
               Test failed
               Metric: 0.75


2024-05-29 11:46:35,039 pid:52758 MainThread giskard.core.suite INFO     Executed test suite 'My first test suite'
2024-05-29 11:46:35,040 pid:52758 MainThread giskard.core.suite INFO     result: failed
2024-05-29 11:46:35,040 pid:52758 MainThread giskard.core.suite INFO     Precision on data slice “`Age` == "30s"” ({'model': <giskard.models.function.PredictionFunctionModel object at 0x178523b80>, 'dataset': <giskard.datasets.base.Dataset object at 0x1785cd450>, 'slicing_function': <giskard.slicing.slice.QueryBasedSliceFunction object at 0x1787d4c70>, 'threshold': 0.76}): {failed, metric=0.6818181818181818}
2024-05-29 11:46:35,040 pid:52758 MainThread giskard.core.suite INFO     Precision on data slice “`BP` == "NORMAL"” ({'model': <giskard.models.function.PredictionFunctionModel object at 0x178523b80>, 'dataset': <giskard.datasets.base.Dataset object at 0x1785cd450>, 'slicing_function': <giskard.slicing.slice.QueryBasedSliceFunction object at 0x1787b2b90>, 'threshold': 0.76}): {failed, metric=0.7272727272727273}
2024-05-29 11:46:35,041 pid:52758 MainThread giskard.core.suite INFO     Precision on data slice “`Na_to_K` == "10-20"” ({'model': <giskard.models.function.PredictionFunctionModel object at 0x178523b80>, 'dataset': <giskard.datasets.base.Dataset object at 0x1785cd450>, 'slicing_function': <giskard.slicing.slice.QueryBasedSliceFunction object at 0x1787b16c0>, 'threshold': 0.76}): {failed, metric=0.7288135593220338}
2024-05-29 11:46:35,041 pid:52758 MainThread giskard.core.suite INFO     Precision on data slice “`Cholesterol` == "HIGH"” ({'model': <giskard.models.function.PredictionFunctionModel object at 0x178523b80>, 'dataset': <giskard.datasets.base.Dataset object at 0x1785cd450>, 'slicing_function': <giskard.slicing.slice.QueryBasedSliceFunction object at 0x1787b0b80>, 'threshold': 0.76}): {failed, metric=0.7547169811320755}

[11]:

Test suite failed.

Test Precision on data slice “`Age` == "30s"”

Measured Metric = 0.68182 Failed

model Drug classifier

dataset Drug classification dataset

slicing_function `Age` == "30s"

threshold 0.76

Test Precision on data slice “`BP` == "NORMAL"”

Measured Metric = 0.72727 Failed

model Drug classifier

dataset Drug classification dataset

slicing_function `BP` == "NORMAL"

threshold 0.76

Test Precision on data slice “`Na_to_K` == "10-20"”

Measured Metric = 0.72881 Failed

model Drug classifier

dataset Drug classification dataset

slicing_function `Na_to_K` == "10-20"

threshold 0.76

Test Precision on data slice “`Cholesterol` == "HIGH"”

Measured Metric = 0.75472 Failed

model Drug classifier

dataset Drug classification dataset

slicing_function `Cholesterol` == "HIGH"

threshold 0.76

Customize your suite by loading objects from the Giskard catalog¶

The Giskard open source catalog will enable to load:

Tests such as metamorphic, performance, prediction & data drift, statistical tests, etc
Slicing functions such as detectors of toxicity, hate, emotion, etc
Transformation functions such as generators of typos, paraphrase, style tune, etc

To create custom tests, refer to this page.

For demo purposes, we will load a simple unit test (test_f1) that checks if the test F1 score is above the given threshold. For more examples of tests and functions, refer to the Giskard catalog.

[ ]:

test_suite.add_test(testing.test_f1(model=wrapped_model, dataset=giskard_dataset, threshold=0.7)).run()