Skip to content

Explorer

explorer

Self-contained interactive HTML rules explorer for fitted FlagGAM estimators.

This module is an ORIGINAL ADDITION (not part of Zhao & Welsch, arXiv:2605.31189): the paper specifies no explorer/export API. export_rules_html renders a single HTML document with inlined CSS and vanilla JS and NO external resources (no CDN, fonts, or images), so the result is fully self-contained: it works offline and can be embedded in an iframe (e.g. the docs site) with no network access.

export_rules_html 

export_rules_html(estimator: Any, path: str | Path | None = None, title: str = 'FlagGAM rules explorer') -> str

Render a self-contained interactive HTML explorer of the fitted rules.

Requires an estimator fitted with representation='full' and the additive head (binary classification or regression only; the compact head's coefficients are per-class scores, not per-rule weights). Returns the complete HTML document as a string; if path is given, also writes it (UTF-8) and still returns the string.

Source code in src/flaggam/explorer.py 
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
def export_rules_html(
    estimator: Any, path: "str | Path | None" = None, title: str = "FlagGAM rules explorer"
) -> str:
    """Render a self-contained interactive HTML explorer of the fitted rules.

    Requires an estimator fitted with `representation='full'` and the additive
    head (binary classification or regression only; the compact head's
    coefficients are per-class scores, not per-rule weights). Returns the
    complete HTML document as a string; if `path` is given, also writes it
    (UTF-8) and still returns the string.
    """
    check_is_fitted(estimator, "core_")
    if getattr(estimator, "representation", "full") == "compact":
        raise ValueError(
            "export_rules_html requires representation='full'; under 'compact' the head "
            "coefficients are per-class scores, not per-rule weights"
        )
    if hasattr(estimator, "classes_") and len(estimator.classes_) > 2:
        raise ValueError("export_rules_html supports binary classification and regression only")
    if not isinstance(estimator.head_, _ADDITIVE_HEADS):
        raise ValueError("export_rules_html requires the additive head")

    payload = _build_payload(estimator, title)
    data = json.dumps(payload, allow_nan=False).replace("</", "<\\/")
    doc = _TEMPLATE.replace("%%DATA%%", data).replace("%%TITLE%%", html.escape(title, quote=True))

    if path is not None:
        Path(path).write_text(doc, encoding="utf-8")
    return doc