Preprocessing and datasets containing different levels of missing rates (event logs)

We provided preprocessed datasets containing missing values, targeting six public event logs and four missing percentage levels (15%, 20%, 25%, and 30%). The CNN method only repairs the single attribute value of event activity names, so we simulated missing values only for the activity names. The MIEC and DFI methods employed the same missing value simulation approach.Please note that we did not apply the masking mechanism to simulate missing values for timestamps, as they represent crucial temporal information in event logs.