cross_validate
from sklearn.model_selection import cross_validate, StratifiedKFold
res = cross_validate(pipe, X, y, cv=StratifiedKFold(5, shuffle=True, random_state=0),
scoring={"auc": "roc_auc", "ap": "average_precision", "ll": "neg_log_loss"},
return_train_score=True, return_estimator=False, n_jobs=-1)
for k in ("auc", "ap", "ll"):
print(k, res[f"test_{k}"].mean().round(4), "+/-", res[f"test_{k}"].std().round(4),
"train", res[f"train_{k}"].mean().round(4))cross_val_predict gives out-of-fold predictions for every row (for error analysis and calibration plots);
do not compute a single "CV score" from them for non-additive metrics without care.
Parameter names
<step>__<param> recursively: prep__num__simpleimputer__strategy, clf__C,
prep__cat__onehotencoder__min_frequency. Inspect with pipe.get_params().keys().
Search over whole steps:
param_grid = [
{"clf": [LogisticRegression(max_iter=5000)], "clf__C": [0.01, 0.1, 1, 10]},
{"clf": [HistGradientBoostingClassifier(random_state=0)], "clf__learning_rate": [0.05, 0.1]},
]Searches
| Search | Use |
|---|---|
GridSearchCV | <= ~50 combinations, discrete choices |
RandomizedSearchCV | continuous ranges (scipy.stats.loguniform), 20-100 iterations |
HalvingRandomSearchCV / HalvingGridSearchCV | many candidates; successive halving on samples (from sklearn.experimental import enable_halving_search_cv first) |
| Optuna / other Bayesian | expensive models, conditional spaces (see optuna-study-writer) |
refit= chooses the metric for the final model when scoring is a dict; refit=callable implements a
one-standard-error rule.
Nested CV
inner = StratifiedKFold(5, shuffle=True, random_state=1)
outer = StratifiedKFold(5, shuffle=True, random_state=2)
search = RandomizedSearchCV(pipe, dist, n_iter=30, cv=inner, scoring="average_precision", random_state=0)
outer_scores = cross_validate(search, X, y, cv=outer, scoring="average_precision")["test_score"]The outer score estimates the performance of "run this search, take the best"; search.best_score_ alone
is biased upward.
Groups and metadata routing
from sklearn.model_selection import StratifiedGroupKFold
cv = StratifiedGroupKFold(5, shuffle=True, random_state=0)
cross_validate(pipe, X, y, cv=cv, groups=customer_ids)
search.fit(X, y, groups=customer_ids) # routed to the search's cvSample weights and other per-row metadata use metadata routing (scikit-learn >= 1.4,
sklearn.set_config(enable_metadata_routing=True) and .set_fit_request(sample_weight=True) on consumers).
Time series
from sklearn.model_selection import TimeSeriesSplit
cv = TimeSeriesSplit(n_splits=5, gap=30, test_size=None) # rows must be sorted by timegap drops rows between train and test (use the label window length when labels look ahead).
Threshold tuning
from sklearn.model_selection import TunedThresholdClassifierCV # sklearn >= 1.5
tuned = TunedThresholdClassifierCV(pipe, scoring="f1", cv=5).fit(X, y)
print(tuned.best_threshold_)For cost-based thresholds pass a scorer built with make_scorer on a cost function.
Performance tips
n_jobs=-1on the search or CV, not on both it and the estimator.memory=caching when only model parameters vary.HistGradientBoosting*instead ofGradientBoosting*for > 10k rows.error_score="raise"while developing, so failures surface instead of NaN scores.
anomaly-detection-model-builder
Build and evaluate anomaly detection for tabular, sensor and transaction data with few or no labels: define the anomaly type (point, contextual, collective), choose detectors (robust z-score/MAD and IQR rules, Mahalanobis with robust covariance, k-NN distance, LOF, Isolation Forest, one-class SVM, PCA reconstruction / autoencoders, forecasting residuals for time series), combine them by rank, explain each alert by its deviating features, set thresholds from an alert budget or the score distribution, and evaluate with precision@k, PR-AUC and recall on known incidents, or detector agreement when no labels exist. Includes pure-Python detectors (robust z, Mahalanobis, kNN, Isolation Forest, ensemble) and an evaluation script.
catboost-model-builder
Build CatBoost models that use its strengths correctly: raw categorical columns as cat_features (ordered target statistics, no manual encoding), text_features, Pool construction with NaN-safe categories, ordered boosting and has_time for temporal data, symmetric (oblivious) trees, overfitting detector with eval_set and use_best_model, losses for classification, regression (Quantile, Poisson, Tweedie) and ranking (YetiRank), class weights, monotone constraints, feature importance (PredictionValuesChange, LossFunctionChange) and SHAP values, GPU training and model export. Plans cat/text/numeric columns from a CSV and generates a runnable training script, and checks parameter sets for the conflicts CatBoost rejects (iterations vs n_estimators, Bayesian bootstrap with subsample, multiple class-weight options, depth > 16, use_best_model without eval_set).
categorical-encoding-selector
Choose and implement the right encoding for every categorical column given the model family and the data: one-hot (with infrequent-level grouping and handle_unknown), ordinal with an explicit order, cross-fitted target encoding (scikit-learn TargetEncoder, CatBoost ordered statistics, category_encoders inside CV), frequency/count encoding, feature hashing, native categorical support in LightGBM, HistGradientBoosting, XGBoost and CatBoost, prefix levels for codes (zip3, NAICS-2), and embeddings. Measures cardinality, rare-level share, unseen-level rate on a time-ordered holdout and target signal per column, emits the ColumnTransformer, and quantifies target-encoding leakage (in-sample vs out-of-fold vs holdout AUC).
classification-model-selector
Choose the right classification algorithm for tabular, text or mixed data and prove the choice with a fair bake-off: logistic regression (L2, L1, elastic net), linear SVM, naive Bayes, k-NN, RBF SVM, decision trees, random forest, extra trees, HistGradientBoosting, LightGBM, XGBoost, CatBoost, explainable boosting machines (EBM/GAM), points scorecards and MLPs, matched to data size, width, categorical and high-cardinality share, missingness, class imbalance, text, and constraints such as interpretability, monotonic constraints, calibrated probabilities, latency budgets and tiny deployable models. Ranks candidates with starting hyperparameters and cautions, then generates a bake-off script running every candidate on identical CV folds with per-fold metrics, fit time and prediction latency.
clustering-algorithm-selector
Choose the right clustering algorithm and check that clustering makes sense at all: cluster tendency (Hopkins statistic with a correlated-features caveat, nearest-neighbour isolation), then k-means / MiniBatchKMeans, Gaussian mixtures with BIC, agglomerative (Ward, average, complete, single) with dendrograms, DBSCAN, HDBSCAN, OPTICS, spectral clustering, BIRCH, mean shift, k-prototypes and k-modes for mixed or categorical data, and Gower distance, matched to data size, expected shapes, noise, varying density, known k, soft memberships, hierarchy and streaming. Covers preprocessing (scaling, encoding, dimensionality, distance metrics), validation (internal indices, bootstrap stability, external labels, business profiling) and assigning new points.
collaborative-filtering-implementer
Implement collaborative filtering recommenders that beat popularity: item-based kNN with cosine, Jaccard or asymmetric similarity, shrinkage, neighbourhood size, positive-only or mean-centred rating signals and 'because you liked X' explanations; implicit ALS with confidence weighting, BPR pairwise ranking and the closed-form EASE model in plain NumPy, tuned on a temporal validation split by recall@k and NDCG@k and refit on all data; plus user-kNN, explicit matrix factorisation, the implicit/LightFM libraries, hyperparameter ranges, cold-start fallbacks, scaling with sparse matrices and approximate nearest neighbours. Produces top-k lists in the format of the recommendation evaluator.