Class SensitivityAnalysis

java.lang.Object

com.eqixiac.equinix.design.solve.SensitivityAnalysis

public final class SensitivityAnalysis
extends Object

Sensitivity of a recommendation, computed exactly from a SolveRun by re-ranking the portfolios the solver already evaluated with the solver's own Ranking rules; nothing is searched again.

1. Weight ranges. For each objective, its share t of the total weight is varied from 0 to 1 while the other weights keep their proportions. Normalised scores do not depend on weights, so each option's weighted score is linear in t and the range over which the recommendation stays best is found in closed form; the option that takes over just outside each end is named.
2. Nearest alternative. The other frontier option closest in normalised-score space, with the raw per-objective difference.
3. Binding constraints. For each constraint, the feasible set plus the portfolios that failed only that constraint is re-normalised and re-ranked: the constraint is binding when the recommendation changes.

Method Summary

Modifier and Type	Method	Description
static SensitivityReport	analyse(SolveRun run)

Methods inherited from class Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Method Details

analyse

public static SensitivityReport analyse(SolveRun run)

Parameters:

run - the solver's evidence

Returns:

the report, or null when nothing was feasible