Monotonic constraints
When domain knowledge says a response must be monotonic (more income never raises credit risk, a higher price never raises demand), enforce it in the model instead of hoping:
| Library | Setting |
|---|---|
| LightGBM | monotone_constraints=[1, -1, 0, ...] (1 increasing, -1 decreasing), monotone_constraints_method="advanced" |
| XGBoost | monotone_constraints="(1,-1,0)" or a dict by feature name |
| CatBoost | monotone_constraints={"tenure": -1} (dict by name, or a list of ints by position) |
| HistGradientBoosting | monotonic_cst={"tenure_months": -1} |
Verify afterwards: the share of ICE curves moving the wrong way should be 0 (ice_analysis.py: up/down counts).
Compare the held-out score with and without constraints. A small loss is often worth the plausibility.
Plausibility review
For each top feature, ask a domain expert to state the expected shape before showing the plot. Then compare:
| Outcome | Action |
|---|---|
| Shape matches | document it |
| Shape differs but data supports it | investigate the data. It may be a real finding, a proxy effect or leakage |
| Shape differs in low-support regions only | cap the feature or constrain it |
| Saw-tooth or implausible thresholds | regularise, increase min_child_samples, or constrain |
Regulated models (credit, insurance, healthcare)
- Regulators and model-risk teams expect documented, plausible, usually monotonic responses for key drivers.
- Keep PDP/ALE plots with the model documentation (model-card-writer).
- Check fairness proxies: plot predictions against features correlated with protected attributes, such as zip code against race.
- Re-run the plots at each retrain and diff the shapes.
Stakeholder charts
- One feature per chart. Use ALE or PDP with a band (the 10th-90th percentile of centred ICE) rather than hundreds of lines.
- Label the y-axis in business units ("predicted churn probability, percentage points").
- Shade or cut regions without data.
- Use one sentence: "Customers with longer tenure are scored as much less likely to churn, mostly in the first two years."
band_lo, band_hi = np.percentile(centred, [10, 90], axis=0)
ax.fill_between(grid, band_lo, band_hi, alpha=0.25, label="10-90% of customers")
ax.plot(grid, centred.mean(axis=0), lw=2.5, label="average")Monitoring
Partial dependence shapes are a model fingerprint. After a retrain, compare the PDP/ALE curves of the top features with the previous version. A changed shape without a known data change is a red flag.
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.