Giters

facebookresearch / Kats

Kats, a kit to analyze time series data, a lightweight, easy-to-use, generalizable, and extendable framework to perform time series analysis, from understanding the key statistics and characteristics, detecting change points and anomalies, to forecasting future trends.

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Frequency error when using holtwinters with alpha.

LumingSun opened this issue · comments

commented

I'm using holtwinters model for hourly data. Everything is OK with fcst = m.predict(steps=24,freq="h",include_history=False) and the result is hourly-distributed.

But if I add the alpha parameter to get forecasting bound fcst = m.predict(steps=24,freq="h",alpha=alpha,include_history=False), the forecasting results become 24 daily data with the same hour value as the last timestamp in history data.

commented

I could see default argument for frequency is "D"

Kats/kats/utils/emp_confidence_int.py

Line 171 in 622dca5

def get_eci(self, steps: int, freq: str = "D") -> pd.DataFrame:

In the case when alpha is passed as an argument forecast is obtained from get_eci function where only steps is passed. Additionally freq also need to be passed.

Kats/kats/models/holtwinters.py

Line 203 in 622dca5

fcst = eci.get_eci(steps=steps)