Train

post

https://api.lensiq.app/train

Start a training job for one or more target columns.
Each target produces its own model. Returns task IDs to poll via GET /jobs/{task_id}.

Recent Requests

Time	Status	User Agent
Retrieving recent requests…

Loading…

Body Params

targets

array of objects

required

length ≥ 1

One entry per target column to predict. Pass a single item for standard single-target training. Pass multiple items to train one model per target, grouped under a shared group_id. Use extra_features on each item to declare which other selected targets will be available at prediction time.

Targets*

dataset_id

ID of the uploaded dataset to train on.

name

Optional human-readable label for this model (or group).

algorithms

array | null

Defaults to []

Subset of algorithms to train and compare. Leave empty (or omit) to run all supported algorithms and auto-select the best one.

Supported algorithms:

Classification: random_forest · xgboost · lightgbm · logistic_regression · decision_tree

Regression: random_forest · xgboost · lightgbm · linear_regression · decision_tree

Winner selection: AUC-ROC 60% + F1-Macro 40% (classification) · R² 50% + Normalized-RMSE 50% (regression).

preprocessing

object

Preprocessing options applied before training. All enabled by default.

enable_enhancement

boolean

Defaults to false

When true, automatically runs enhancement cycles after training. Diagnoses weaknesses (minority class, OOD fragility, miscalibration) and retrains with targeted fixes (SMOTE, tree-based retry, calibration). Best model wins. Dataset is deleted after all cycles complete. Check model enhancement_status for progress: validating → enhancing → stable.

Responses

200Successful Response

422Validation Error