I think for all gaps in the dataset, the available attributes would still give us some information. Therefore, I think among the observations, I am going to calculate the similarities by unsupervised methods first and then use the similarity score as a feature to find the most prominent features.

After examining the data, I have decided to choose "Meteorite_Landings.csv" as my dataset for the challenge. The problem for the challenge is to find a way to recover some of the missing data so other people can use the data in various projects.