pywrdrb.pre.PredictedDiversionPreprocessor

class pywrdrb.pre.PredictedDiversionPreprocessor(start_date=None, end_date=None, modes=('regression_disagg',), use_log=True, remove_zeros=False, use_const=False)

Predicts catchment inflows at Montague and Trenton using specified modes (e.g., regression, perfect foresight, moving average).

Example usage: ```python from pywrdrb.pre import PredictedDiversionPreprocessor diversion_predictor = PredictedDiversionPreprocessor(

start_date=”1983-10-01”, end_date=”2016-12-31”, modes=(“regression_disagg”,),

) diversion_predictor.process() diversion_predictor.save() ```

__init__(start_date=None, end_date=None, modes=('regression_disagg',), use_log=True, remove_zeros=False, use_const=False)

Initialize the PredictedDiversionPreprocessor.

Parameters:

• start_date (str, None) – Start date for the time series. If None, match the input data.

• end_date (str, None) – End date for the time series. If None, match the input data.

• modes (tuple) – Modes to use for prediction. Default is (‘regression_disagg’,). Options include: “regression_disagg”, “perfect_foresight”, “moving_average”, “same_day”.

• use_log (bool) – Whether to use log transformation for model vars. Default is True.

• remove_zeros (bool) – Whether to remove zero values. Default is False.

• use_const (bool) – Whether to use a constant/intercept in regression. Default is False.

Return type:

None

Methods

__init__([start_date, end_date, modes, ...])	Initialize the PredictedDiversionPreprocessor.
get_prediction_node_lag_combinations()	Return dict of predicted diversion column names formatted as (node, lag, mode) tuples.
load()	Load NJ diversions and catchment WC data (used for structural compatibility).
make_predictions(regressions)	Generate lead-time predictions using the timeseries data and trained models.
process()	Run full prediction workflow.
save()	Save predicted diversion time series to CSV.
train_regressions()	Train the AR models for different node, lag combinations.