I have visualized missing data in my Kaggle project:

To me, it clearly shows that there is a pattern in missing data (probably human-made to challenge people working on the project).

However, I just don’t know how to handle it. I could just impute mean, median or mode for each variable but I wonder if there is a best practice in this case and additional insights to get from the visualization I didn’t get.