Open nesrnesr opened 2 years ago
Unfortunately, I have the same issue and I think nesrnesr is absolutely right. If the prediction does not include any ones the precision should simply be zero. Luckily the case is easy to catch and fix.
try something like ts_precision([0,1,0] , [0,0,0])
@sbuse if you are in a hurry, you can use the official C++ implementation [https://github.com/IntelLabs/TSAD-Evaluator]().
Hello, thank you for the implementation. However, as the title indicates, when I want to assess the performance of a prediction (that says that the full time-serie is normal) compared to a ground-truth that says otherwise (contains anomalies), the code raises an assertion error. Shouldn't the score be indicating 0 since the predictions failed to predict that there are anomalies that were not detected?