linkedin / greykite

A flexible, intuitive and fast forecasting library
BSD 2-Clause "Simplified" License
1.81k stars 106 forks source link

ChangePoint detection for not daily data #134

Open HiromasaYamanishi opened 9 months ago

HiromasaYamanishi commented 9 months ago

My data consists of records taken every 20 seconds, as shown below. However, when I performed change point detection with the following parameters, I was able to detect change points, but as shown in the attached figure, the trend line ended up being straight. I've read through the documentation but haven't been able to resolve this issue. I would appreciate it if you could help me.

data beh_time | count 2023-01-01 10:01:20 | 50 2023-01-01 10:01:40 | 47 2023-01-01 10:02:00 | 46 2023-01-01 10:02:20 | 48 2023-01-01 10:02:40 | 52 ....

parameter res = model.find_trend_changepoints( df=df, time_col="beh_time", # time column name value_col="count", actual_changepoint_min_distance=timedelta(seconds=300), resample_freq=timedelta(seconds=20), regularization_strength=0.1,)

result

greykite_event