Decision Support

evidencelib includes simple pignistic transforms for decision support.

m.pignistic()
m.pignistic_regions()
m.decision()

decision() returns the singleton with the largest value from pignistic(). It is a convenience method, not a replacement for application-specific utility or loss functions.

DST

In a DST frame, singleton hypotheses are disjoint. Therefore pignistic() returns a probability distribution over the singleton hypotheses.

frame = Frame.dst(["A", "B"])
A, B = frame.symbols()
m = frame.mass({A: 0.4, B: 0.2, A | B: 0.4})

assert sum(m.pignistic().values()) == 1.0

DSmT

In a free or hybrid DSmT frame, singleton hypotheses can overlap. Therefore pignistic() returns singleton event scores useful for ranking decisions, but the scores do not have to sum to one.

Use pignistic_regions() if you need a probability distribution over disjoint Venn regions.

frame = Frame.dsmt(["A", "B"])
A, B = frame.symbols()
m = frame.mass({A: 0.2, B: 0.3, A & B: 0.4, A | B: 0.1})

scores = m.pignistic()
regions = m.pignistic_regions()

regions is useful when the downstream calculation requires mutually exclusive states.