Holt's Exponential Smoothing
Holts’ Exponential Smoothing (HES) or Double Exponential Smoothing extends the concept of Simple Exponential Smoothing (SES) to capture trend along with level in the time series data to predict future values assuming there is no seasonality component in series
levelt = levelt-1 + trendt-1 + (alpha) * (observedt-1 - levelt-1 - trendt-1)
trendt = trendt-1 + beta *(levelt - levelt-1 - trendt-1 )
\
forecastt+k = levelt + trendt
The level is the forecast of the value in the series, which is controlled by the parameter alpha. While the trend is expected growth in the series, controlled by another parameter beta, the smoothing constant for trend. The generated optimal parameters for alpha, beta and step are derived through alpha_var, beta_var and step_var which is then substituted in the equation for forecasting using HES. In HES, the value of _**gamma **_is substituted as zero, since it assumes there is no seasonality.
python ponyge.py --parameters gets/des.txt
By default it runs on Daily Waste Generated dataset. Command to run on the custom dataset:
python ponyge.py --parameters gets/des.txt --dataset_train path_to_traindataset --dataset_test path_to_testdataset.