# Model operations

Operations that analyse the data or learn from it. Among them are statistical tests and regressions, clustering and outlier detection, training, scoring and evaluating models, forecasting a time series, and working with text, networks, geography and images. Some return the table with columns added, some a table describing the input, and some a trained model or a file.

Page: <https://meridian.online/docs/arcform/operations/model>

## One row out per row in, with columns added or changed

- [Find outliers](https://meridian.online/docs/arcform/operations/find-outliers): Outlier or anomaly detection (Runs as a SQL step)
- [Cluster rows](https://meridian.online/docs/arcform/operations/cluster-rows): K-means, hierarchical, DBSCAN, other clustering of rows (Not built yet)
- [Build features for a model](https://meridian.online/docs/arcform/operations/build-features-for-a-model): Generate features or row weights for a model (Runs as a SQL step)
- [Work with geography](https://meridian.online/docs/arcform/operations/work-with-geography): Geocode, spatial operations on geometry, grid cells, routing (Runs as a SQL step)
- [Reduce dimensions](https://meridian.online/docs/arcform/operations/reduce-dimensions): PCA, SVD, UMAP, t-SNE, other dimensionality reduction (Not built yet)
- [Classify text](https://meridian.online/docs/arcform/operations/classify-text): Classify text into categories (Not built yet)
- [Embed text as vectors](https://meridian.online/docs/arcform/operations/embed-text-as-vectors): Text or image to vector (Runs as a SQL step)
- [Extract meaning from text](https://meridian.online/docs/arcform/operations/extract-meaning-from-text): Sentiment, named entities, topics, language detection (Not built yet)
- [Ask a language model](https://meridian.online/docs/arcform/operations/ask-a-language-model): Call a large language model (Not built yet)
- [Read text and objects in images](https://meridian.online/docs/arcform/operations/read-text-and-objects-in-images): Image transformation, recognition, OCR, barcodes (Not built yet)

## Fewer rows, same columns

- [Search text by relevance](https://meridian.online/docs/arcform/operations/search-text-by-relevance): Full-text search ranked by relevance (Runs as a SQL step)

## More rows, each traceable to one input row

- [Split text into words](https://meridian.online/docs/arcform/operations/split-text-into-words): Tokenise, stem, n-grams, stop words, TF-IDF (Runs as a SQL step)
- [Forecast a time series](https://meridian.online/docs/arcform/operations/forecast-a-time-series): ARIMA, ETS, other forecasting (Not built yet)

## A new grain: each row stands for a group of input rows

- [Find items that occur together](https://meridian.online/docs/arcform/operations/find-items-that-occur-together): Association rules, frequent itemsets, market basket (Runs as a SQL step)
- [Analyse a network](https://meridian.online/docs/arcform/operations/analyse-a-network): Network or graph analysis (Runs as a SQL step)
- [Resample a time series](https://meridian.online/docs/arcform/operations/resample-a-time-series): Resample, decompose, difference, smooth, fill time gaps (Runs as a SQL step)

## One table from two or more

- [Find similar rows](https://meridian.online/docs/arcform/operations/find-similar-rows): Row-to-row distance, nearest-neighbour search, vector search (Runs as a SQL step)

## A table describing the input rather than holding its rows

- [Evaluate a model](https://meridian.online/docs/arcform/operations/evaluate-a-model): Performance metrics, confusion matrix, ROC, lift, cross-validation (Runs as a SQL step)
- [Explain a model](https://meridian.online/docs/arcform/operations/explain-a-model): Variable importance, partial dependence, SHAP, model inspection (Not built yet)
- [Optimise a plan](https://meridian.online/docs/arcform/operations/optimise-a-plan): Linear programming, simulation, prescriptive optimisation (Not built yet)
- [Use a probability distribution](https://meridian.online/docs/arcform/operations/use-a-probability-distribution): Distribution functions, density estimation (Not built yet)
- [Fit a regression](https://meridian.online/docs/arcform/operations/fit-a-regression): Linear, logistic or generalised linear regression as statistics (Runs as a SQL step)
- [Run a hypothesis test](https://meridian.online/docs/arcform/operations/run-a-hypothesis-test): Hypothesis test, ANOVA, chi-square, t-test (Not built yet)

## A model or a file, not a table

- [Calibrate a model's confidence](https://meridian.online/docs/arcform/operations/calibrate-a-models-confidence): Rescale a model's confidences (Not built yet)
- [Save or load a model](https://meridian.online/docs/arcform/operations/save-or-load-a-model): Store, load, export or import a trained model (Not built yet)
- [Recommend items](https://meridian.online/docs/arcform/operations/recommend-items): Collaborative filtering, matrix completion (Not built yet)
- [Model a sequence](https://meridian.online/docs/arcform/operations/model-a-sequence): Hidden Markov and other sequence models, reinforcement learning (Not built yet)
- [Model time to an event](https://meridian.online/docs/arcform/operations/model-time-to-an-event): Time-to-event models (Not built yet)
- [Combine models](https://meridian.online/docs/arcform/operations/combine-models): Bagging, boosting, stacking, voting, a choice of learners, AutoML (Not built yet)
- [Train a linear model](https://meridian.online/docs/arcform/operations/train-a-linear-model): Linear models, SVM, naive Bayes trained as a predictor (Not built yet)
- [Train a neural network](https://meridian.online/docs/arcform/operations/train-a-neural-network): Neural network, deep learning (Not built yet)
- [Learn rules from data](https://meridian.online/docs/arcform/operations/learn-rules-from-data): Rule induction, subgroup discovery (Not built yet)
- [Train a tree model](https://meridian.online/docs/arcform/operations/train-a-tree-model): Decision tree, random forest, gradient boosted trees (Not built yet)
- [Tune a model](https://meridian.online/docs/arcform/operations/tune-a-model): Hyperparameter search, feature selection, feature weighting (Not built yet)

## A model or file and a table in, the table out with columns added

- [Score with a model](https://meridian.online/docs/arcform/operations/score-with-a-model): Score or predict with a trained model (Not built yet)
- [Classify by nearest neighbours](https://meridian.online/docs/arcform/operations/classify-by-nearest-neighbours): K-nearest neighbours (Runs as a SQL step)

## Not a data operation

- [Assemble a modelling pipeline](https://meridian.online/docs/arcform/operations/assemble-a-modelling-pipeline): Group models, assemble a modelling pipeline (A feature, not a step)