chengqianghuang / exp-anomaly-detector-AIOps

Using DNN for univariate time series anomaly detection over AIOps Competition dataset
44 stars 29 forks source link

How to dicide the window size? #2

Open ab4738 opened 3 years ago

ab4738 commented 3 years ago

I saw the different window size set as W=[2, 5, 10, 25, 50, 100, 200, 300, 400, 500] in feature engineering. I'm curious about why you choose these numbers. Is there any reason or concern?
Thank for your patience and look forward to your reply.