Non-personalised
| Method | Notes |
|---|---|
| Global popularity | a strong baseline; always report it |
| Recent / trending popularity | time-decayed counts; strong for news and fashion |
| Segment popularity | per country, category, device, new vs returning |
| Association rules / co-occurrence ("bought together") | item-to-item, no user history needed |
Neighbourhood (memory-based) CF
- Item-item kNN: score unseen items by their similarity to items the user interacted with. It is stable, explainable ("because you watched X") and strong on small and medium data. Use cosine with shrinkage, or adjusted cosine for ratings.
- User-user kNN: weaker and less stable when users have few interactions.
- Session kNN: neighbours by session rather than user.
Latent factor models
| Model | Feedback | Objective |
|---|---|---|
| Funk SVD / biased MF | explicit ratings | squared error on observed ratings |
| ALS for implicit feedback (Hu, Koren, Volinsky 2008) | counts, plays, clicks | weighted squared error with confidence 1 + alpha x count on all user-item pairs |
| BPR (Rendle 2009) | implicit | pairwise ranking: seen items above unseen |
| EASE (Steck 2019) | implicit | closed-form item-item linear model; very strong baseline |
| SLIM | implicit | sparse item-item regression |
Libraries: implicit (ALS, BPR, GPU), LightFM (hybrid MF with features), surprise (explicit), RecBole and
Microsoft Recommenders (benchmarks), TensorFlow Recommenders / TorchRec (two-tower).
Content-based and hybrid
- Represent items by attributes, text embeddings or image embeddings, and users by aggregates of the items they liked.
- Content handles new items immediately. CF captures taste that content misses.
- Hybrids: LightFM (MF with side features), weighted blending, switching (content when an item has few interactions), or features in a re-ranker.
Sequential and session-based
Next-item prediction from the order of interactions: Markov chains, GRU4Rec, SASRec and BERT4Rec. Use them when order matters (playlists, e-commerce sessions) and data is large.
Two-tower retrieval + ranking
A user tower and an item tower produce embeddings, and dot-product retrieval (ANN index: FAISS, ScaNN) yields candidates. A ranking model (GBDT LambdaMART or a neural network) then scores the candidates with rich features. This is the standard at large scale.
Cold start
| Case | Strategy |
|---|---|
| New user, no history | popularity by context (country, device, referrer), onboarding questions, trending items |
| New user, a few interactions | item-kNN from those items, session-based models |
| New item | content similarity, an exploration slot, boosting in the ranker with uncertainty |
| New system | editorial lists, popularity, content-based; collect interactions first |
Evaluate cold users and items separately. Overall metrics are dominated by warm users.
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.