search

KDD-OpenSource / DeepADoTS

Repository of the paper "A Systematic Evaluation of Deep Anomaly Detection Methods for Time Series".
MIT License
569 stars 117 forks source link

interpolate nans #139

Closed xasetl closed 6 years ago

WGierke commented 6 years ago

I'd argue that the global mean of the time series is prone to outliers and there are cases where imputing with it can create an additional extreme outlier. How about a simple interpolation using the two neighboring points or using pd.rolling_mean?

danthe96 commented 6 years ago

I would also prefer rolling mean or linear interpolation