PETA-AI

analyst workspace · RPJMN 2025–2029 (demo corpus)

Coherence workspace

last full pass · pass 4 of 4 · 9,412 pages

demo

Provisional / pre-validation. Backtesting against RPJMN 2015–2019 → 2020–2024 has not been run. Every number on this screen is illustrative demo data.

draft under simulation

KRISNA irrigation / paddy-land retrodiction

A historical planning decision already known to have caused downstream problems, fed back into the system as if it were still an unsigned draft.

Irrigated-hectarage target raised without reconciling the paddy-land data base+18% target, unchanged land-conversion rules

forecast cascade · probability of missing target

interval shown, never a bare point estimate

  • Irrigation Network Rehabilitation

    Public Works

    81% · CI 66%–91%

    12-month horizonGP forecast + survival analogue

  • Sustainable Paddy Field Protection

    Agriculture

    74% · CI 55%–88%

    18-month horizonCausal model over coherence graph

  • Rural Water Access

    Public Works

    49% · CI 30%–68%

    24-month horizonABM cascade, 2,000 runs

  • Food Price Stabilisation

    Trade

    38% · CI 19%–60%

    24-month horizonABM cascade, 2,000 runs

Cascade output is validated at pattern level — which programs are affected and roughly what magnitude — not exact numeric agreement.

retrodiction check

This scenario has a recorded historical outcome. Run the forecast first, then reveal it — that ordering is what makes the match checkable rather than asserted.

Forecast backtesting (train on RPJMN 2015–2019, test against 2020–2024) has not been run. Every hazard and interval above is illustrative demo data labelled provisional.