factor_agent.evaluator.Evaluator#

class Evaluator(segment, prices, label, scoring, reference, warmup_bars, njobs, purge_bars=5, known=(), vocabulary=None, intraday=None, store=None, store_key=None)[source]#

Bases: object

Score graphs on one segment; see the module docstring.

Make one with for_segment; nothing changes its dates afterwards.

Methods#

`Evaluator.contributions`	Each named pool factor's `Contribution` within the group of `names`.
`Evaluator.evaluate`	Return one entry per graph, in order; see the module docstring.
`Evaluator.for_segment`	Return an evaluator of `segment`.
`Evaluator.join`	Add `(name, graph)` pairs to the reference pool and to the tried trials, in order, without scoring them.
`Evaluator.nearest_tried`	Each graph's most correlated tried trial and that correlation, `("", NaN)` without one; unscored.
`Evaluator.trace_coverage`	Where `graph` loses its values: the coverage of each of its nodes on the scored member days.
`Evaluator.tried`	The tried trials' names, in the order they joined, and the matrix of their correlations.

Attributes#

Evaluator.segment#: The segment this evaluator reads.