Scenario Discovery

Scenario discovery identifies which scenarios meet a target outcome criterion, helping to understand what conditions or interventions lead to desired results.

Setup

using ADRIA, ADRIAanalysis, Statistics

ResultSet

The examples below assume an ADRIAResultSet rs:

dom = ADRIA.load_domain("path to domain data", "<RCP>")
scens = ADRIA.sample(dom, 4096)
rs = ADRIA.run_scenarios(dom, scens, "45")

See Loading a Domain, Generating scenarios and Running scenarios for more detail.

Screening scenarios

screen_scenarios identifies scenarios where all outcomes meet a given rule after column-wise normalization. A scenario is selected only when the rule holds for every outcome column.

tac = ADRIA.metrics.scenario_total_cover(rs)
mean_tac = vec(mean(tac, dims=1))

rsv = ADRIA.metrics.scenario_rsv(rs)
mean_sv = vec(mean(rsv, dims=1))

r_juves = ADRIA.metrics.scenario_relative_juveniles(rs)
mean_juves = vec(mean(r_juves, dims=1))

# Stack outcomes column-wise (one column per metric)
y = hcat(mean_tac, mean_sv, mean_juves)

# Select scenarios where all normalized outcomes are at or above the 30th percentile
robust = screen_scenarios(y, x -> x >= 0.3)

The returned robust is a vector of scenario indices. Use it to inspect the inputs that produced those outcomes:

rs.inputs[robust, :]

Or create a binary classification vector for further analysis (e.g., random forests):

behave = zeros(size(y, 1))
behave[robust] .= 1.0

---

This page was generated using Literate.jl.