I have a task that involves imputation of anomaly rows. I have a dataset that look like this, Test Time Voltage Cycle 1970-01-01 01:58:25.535599999 1.746142 1 1970-01-01 01:58:25.606800000 1.723656 1 1970-01-01 01:58:25.679500000 1.701954 1 1970-01-01 01:58:25.753600000 1.679174 1 1970-01-01 01:58:25.812500000 1.657140 1 1970-01-01 01:58:25.873300000 1.636981 ... ... 1970-01-01 19:34:56.799600 -1.356987 13 1970-01-01 19:35:01.808400 -1.357799 13 Basically every cycle has it's own full time series cycle but cycles don't have the same row counts and as you can see the time interval are not equal.Here is plot look like for this sample data, Anomaly dataset image I tried resampling methods but I am ending up having different row count for each cycle and for that reason neither time series(arima, sarima) or ml models generate right results. What should be the approach for this dataset ? I have seen that some libraries in R capable of handling unequal time series but none of them have reliable working packages in python. Continue reading...
