Output Format

All outputs are written under {storages.artifacts.filesystem.base_path}/{run_id}/.

Output Files

Path	Description
sampling_day.json	Run metadata: resolved run date, sampling window
datasets/cases_{region}_sampled.csv	Case data trimmed to the modelling window
datasets/weather_{region}_sampled.csv	Weather features aligned to the modelling window
datasets/thresholds/{region}_all_thresholds.csv	Per-region threshold tables (both methods)
cutoffs.json	Case and weather cutoff dates, prediction calendar
predictions/Predictions_*.csv	District and state-level predictions
plots/{region}_{model}_{threshold}_{date}.png	Choropleth map images
results/AllMaps_{month}_{date}.zip	All map images bundled as a zip
reports/*.json	Structured report metadata
reports/*.tex / reports/*.pdf	LaTeX source and compiled PDF (if compile_pdf: true)

Predictions File

Two prediction files are generated per run:

• Predictions_{MonthRange}_District_{Date}.csv — district-level

• Predictions_{MonthRange}_{State}_{Date}.csv — state-level aggregate

Example filenames:

Predictions_Apr - May 2026_District_20260407.csv
Predictions_Apr - May 2026_Karnataka_20260407.csv

District-Level Columns

Column	Type	Description
district	String	Region identifier (e.g. district_374)
recordDate	Date	Date of the prediction record (YYYY-MM-DD)
ISOWeek	Integer	ISO week number
thresholdMethod	String	historical or previousNweeks
Mean	Float	Historical mean of cases used for threshold
StdDev	Float	Historical standard deviation
T0.00	Float	Threshold tier 0 (baseline)
T1.00	Float	Threshold tier 1 (elevated)
T2.00	Float	Threshold tier 2 (high)
startDatePredictedWeek	Date	Start of the predicted ISO week
dateOfComputingPrediction	Date	Date the prediction was computed
regionID	String	Region identifier (same as district)
prediction	Float	Predicted number of dengue cases
model	String	Model used (ensembleModel, NBR, or TSE)
predictionZone	Float	Zone classification based on threshold tiers

State-Level Columns

Column	Type	Description
dateOfComputingPrediction	Date	Date prediction was computed
startDatePredictedWeek	Date	Start of the predicted ISO week
regionID	String	State identifier (e.g. state_29)
prediction	Float	Predicted cases for the state
thresholdMethod	String	Threshold method used
predictionZone	Integer	Alert zone (0 = low, higher = elevated risk)
model	String	Model used for prediction

Threshold Methods

Two threshold methods are computed for each region and week:

• historical — based on the mean and standard deviation of historical cases (configurable lookback via thresholds.historical_n_years)

• previousNweeks — based on the mean of the most recent N weeks (configurable via thresholds.n_weeks)

Example Maps

The generate_maps step produces choropleth PNGs showing dengue risk levels per district. Below are two example outputs from the same prediction week (week of 2025-02-18, Andhra Pradesh), one for each threshold method:

Historical Thresholds

Previous N-Weeks Thresholds

Prediction Zones

predictionZone is a numerical classification of risk:

Zone	Meaning
0	Prediction below T0 (low / baseline)
1	Prediction between T0 and T1 (moderate)
2	Prediction between T1 and T2 (elevated)
3	Prediction above T2 (high)