search

facebook / prophet

Tool for producing high quality forecasts for time series data that has multiple seasonality with linear or non-linear growth.
https://facebook.github.io/prophet
MIT License
18.09k stars 4.49k forks source link

fbprophet for anomaly detection giving different results on same data #1955

Open Sreedhanya-K opened 3 years ago

Sreedhanya-K commented 3 years ago

I am using fbprophet for anomaly detection, and for the same data sets, the results are different for different runs in my laptop. The number of anomalies and the anomalies which the code pick up are different for 1st, 2nd third.. runs. I am confused, how can I use this , though it is giving good results many a times. The false positives are too many in a few runs. Can someone please help me to fix this. I have tried algorithm = 'Newton'. Still the result is not constant. Sree

tcuongd commented 3 years ago

Hmm that's interesting, I didn't think there would be that much variability in the parameter estimates. We could probably help you more if you posted some sample data and the algorithm you're using to detect anomalies?

wangzhongqing commented 3 years ago

I also encountered this problem and I am very confused now. Have you solved it?

wangzhongqing commented 3 years ago

I use the official example, and the result of each run is different.

wangzhongqing commented 3 years ago

Hmm that's interesting, I didn't think there would be that much variability in the parameter estimates. We could probably help you more if you posted some sample data and the algorithm you're using to detect anomalies?

Hi：I use the official example, and the result of each run is different.Can u help me?The [upper & lower ] of results are different each time. yhat is all the same.

soojek48 commented 3 years ago

Hmm that's interesting, I didn't think there would be that much variability in the parameter estimates. We could probably help you more if you posted some sample data and the algorithm you're using to detect anomalies?

Hi：I use the official example, and the result of each run is different.Can u help me?The [upper & lower ] of results are different each time. yhat is all the same.

I have the same problem. I used a basic model that didn't adjust any parameters. The yhat is the same, but the lower and upper values continue to change.