Final Project: Build a Complete Classifier Pipeline

Every piece from this course now exists on its own: loading and splitting data, training classifiers, evaluating with more than just accuracy, scaling features safely with Pipeline, and validating robustly with cross-validation and GridSearchCV. This final project assembles all of it into one complete, defensible ML workflow.

Requirements

Your finished ml_project.py should satisfy every one of these:

1. Load a classification dataset (load_iris, load_wine, load_breast_cancer, or your own via make_classification) and split it with train_test_split.
2. Build a Pipeline combining StandardScaler with a classifier of your choice.
3. Fit the pipeline and report accuracy, a confusion matrix, precision, and recall on the held-out test set.
4. Run cross_val_score on the same pipeline and print the mean cross-validated accuracy, does it roughly agree with your single-split test accuracy?
5. Use GridSearchCV to tune at least one hyperparameter of your classifier, and report the best parameters found.

from sklearn.datasets import load_wine
from sklearn.model_selection import train_test_split, cross_val_score, GridSearchCV
from sklearn.preprocessing import StandardScaler
from sklearn.neighbors import KNeighborsClassifier
from sklearn.pipeline import Pipeline
from sklearn.metrics import accuracy_score, confusion_matrix, precision_score, recall_score

# 1. Load and split
data = load_wine()
X_train, X_test, y_train, y_test = train_test_split(
    data.data, data.target, test_size=0.2, random_state=42
)

# 2. Build the pipeline
pipeline = Pipeline([
    ('scaler', StandardScaler()),
    ('knn', KNeighborsClassifier()),
])

# 3. Fit and evaluate
pipeline.fit(X_train, y_train)
preds = pipeline.predict(X_test)
print('Accuracy:', accuracy_score(y_test, preds))
print('Confusion matrix:\n', confusion_matrix(y_test, preds))
print('Precision (macro):', precision_score(y_test, preds, average='macro'))
print('Recall (macro):', recall_score(y_test, preds, average='macro'))

# 4. Cross-validate
cv_scores = cross_val_score(pipeline, data.data, data.target, cv=5)
print('Mean CV accuracy:', cv_scores.mean())

# 5. Tune a hyperparameter
param_grid = {'knn__n_neighbors': [1, 3, 5, 7, 9, 11]}
grid = GridSearchCV(pipeline, param_grid, cv=5)
grid.fit(data.data, data.target)
print('Best params:', grid.best_params_)
print('Best CV score:', grid.best_score_)

Notice param_grid uses the key 'knn__n_neighbors', the double-underscore syntax lets GridSearchCV reach inside a Pipeline and tune a specific step's parameter, stepname__paramname.

Stretch goals

• Try a different classifier in the pipeline (RandomForestClassifier, SVC, LogisticRegression) and compare cross-validated scores.
• Try a different dataset (load_breast_cancer is a good binary-classification option) and see how precision/recall behave differently from a balanced multi-class problem like wine or iris.
• Expand param_grid to tune more than one hyperparameter at once, GridSearchCV will try every combination.
• Plot the confusion matrix visually instead of printing raw numbers (matplotlib is available in this sandbox too).

Submit a link to your finished project (a repo or gist) below, an instructor will review it before you can mark this lesson complete.