How To Run an Automatic Baseline Workflow¶

Automatic workflows fit baselines, benchmarks, and batch experiments. Use GLMStudy for models you expect to inspect and keep.

Automatic Study Design¶

from glm_factor_optimizer import GLMStudy

study = GLMStudy(
    df,
    target="events",
    exposure="hours",
    prediction="predicted_count",
    factor_kinds={"segment": "categorical", "region": "categorical"},
)

study.split(seed=42)
study.auto_design(
    ["score", "segment", "region", "age"],
    top_n=5,
    trials=50,
    accept_interactions=False,
)

study.validation_report()["summary"]

This uses the same ranking, factor optimization, and acceptance methods as the manual notebook flow.

Sequential GLMWorkflow¶

For a smaller automatic wrapper:

from glm_factor_optimizer import GLMWorkflow

workflow = GLMWorkflow(
    target="events",
    family="poisson",
    exposure="hours",
    prediction="predicted_count",
    factor_kinds={"segment": "categorical"},
    rank_candidates=True,
    top_n=5,
    trials=50,
)

result = workflow.fit(df, ["score", "segment", "region", "age"])

Use GLMWorkflow for a compact, stateless result. Use GLMStudy for notebook iteration, audit history, refinements, and manual accept/reject decisions.

Saving Automatic Runs¶

study.finalize()
study.save("runs")

or:

workflow = GLMWorkflow(..., output_dir="runs")
result = workflow.fit(df, factors)