High variance (large gap)
| Model | Levers, strongest first |
|---|---|
| Gradient boosting (LightGBM/XGBoost/HGB) | larger min_child_samples / min_samples_leaf, fewer leaves / lower depth, subsample + colsample < 1, L2 (reg_lambda), lower learning rate with early stopping |
| CatBoost | larger l2_leaf_reg, lower depth, early stopping (od_wait), more random_strength |
| Random forest | larger min_samples_leaf, lower max_features, max_depth limit |
| Linear / logistic | stronger penalty (lower C), L1 to drop features, fewer high-cardinality one-hot columns |
| k-NN | larger k |
| SVM (RBF) | lower C, lower gamma |
| Any | remove noisy / ID-like / high-cardinality features, target-encode with cross-fitting, bag several models |
Confirm: the gap at the largest size shrinks and the validation score does not fall.
Data-limited (large gap and rising validation)
Do the variance fixes now; plan data collection from the 2x prediction (see data-size-extrapolation.md). Confirm after collecting: the new point lies near the prediction. If it lies below, the new data differs (drift, labelling change).
High bias (gap closed, score low)
| Model | Levers |
|---|---|
| Linear / logistic | interactions, splines, log/ratio transforms, weaker penalty, or switch to boosting |
| Gradient boosting | more leaves / depth, smaller min_child_samples, more rounds (higher early-stopping patience), less regularisation |
| Random forest | min_samples_leaf=1, max_features up |
| Any | new features with signal: aggregates over history, entity-level statistics, domain ratios, external data |
| Any | check the target: label noise caps every model; measure label agreement on a sample |
Confirm: the training score rises; the validation score follows; the gap may reopen slightly (expected).
Converged
The model is as good as this data and these features allow. Gains now come from new information (features, sources) or a different problem framing, not from tuning or rows. Compare with the irreducible-error estimate from ml-feasibility-assessor if one exists.
Data problems the curve reveals
| Symptom | Check |
|---|---|
| Validation falls at larger sizes | shuffle=True? sorted file? drift between early and late rows (adversarial validation) |
| Wild fold-to-fold swings | groups split across folds, very small classes, outliers in the target |
| Validation above training | tiny validation folds, regularisation stronger at train time, duplicated rows across folds |
| Curve far better than production | leakage: run target-leakage-detector |
One change at a time
Re-run the same curve (same splitter, sizes and seeds) after each change and compare the two CSVs. Changing regularisation and features together hides which one helped.
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.