robil/siagaai-flood-risk-model

SiagaAI Flood Risk Model

Model prediksi risiko banjir untuk 47 kota utama di Indonesia, dikembangkan sebagai bagian dari sistem SiagaAI.

Model Info

Attribute	Value
Model type	Random Forest Classifier
Version	20260521_0852
Training samples	22,980
Features	17
Classes	aman, awas, siaga, waspada

Metrics (Test Set)

Metric	Value
Accuracy	0.9915
F1 Weighted	0.9915
F1 Macro	0.9886
ROC-AUC (OvR)	0.9999

Per-Class Metrics

Class	Precision	Recall	F1-Score	Support
aman	0.9735	0.9946	0.9840	185
awas	0.9959	0.9951	0.9955	1231
siaga	0.9784	0.9861	0.9822	505
waspada	0.9966	0.9889	0.9927	899

Top 5 Features

• drainage_score
• rainfall_3h
• humidity
• elevation_m
• saturation_index

Risk Classes

Level	Description
aman	Kondisi normal, tidak ada risiko signifikan
waspada	Risiko mulai meningkat, pantau kondisi cuaca
siaga	Risiko cukup tinggi, bersiap evakuasi
awas	BAHAYA — risiko sangat tinggi, segera evakuasi

Usage

from huggingface_hub import hf_hub_download
import joblib

# Download model
model = joblib.load(hf_hub_download(
    repo_id="robil/siagaai-flood-risk-model",
    filename="flood_risk_model.pkl"
))

Files

File	Description
flood_risk_model.pkl	Trained Random Forest model
feature_scaler.pkl	StandardScaler untuk fitur
label_encoder.pkl	LabelEncoder untuk kelas risiko
model_metadata.json	Metrics, hyperparameters, metadata