ts-hyperparam-opt 0.1.3

Convenience package for parallelized hyperparameter optimization (e.g. in Jupyter Notebooks) using grid search and CV

Time Series Hyperparameter Optimization (CV + Parallel)

Convenience package for optimizing hyperparameters for Time Series forecasting using methods like ExponentialSmoothing or SARIMAX. Especially useful for Jupyter Notebooks where parallelization (with e.g. ProcessPoolExecutor) only works when importing the function used in parallel.

Install it from PyPI

pip install ts-hyperparam-opt

Usage

from ts_hyperparam_opt import parallel_hyperparameter_optimization as pho

params_sarima = [
    [(1,1,1), (1,1,1,7)],
    [(1,1,0), (1,1,1,7)]
    ]

if __name__ == '__main__':
    freeze_support()
    results = process_map(functools.partial(pho.optimize_hyperparams,
                            data=df_data, func="sarima", 
                            n_steps=15), params_sarima)
    results_sorted = pho.sort_results(results)

Development

Alpha Version

Currently supported methods:

• (Triple) Exponential Smoothing (Holt-Winters)
• SARIMA(X)