Time changes everything ⏰

Many aspects of your data are just that: Aspects of your data. test Time, however, time changes everything.

When you are dealing with a time series or panel data, the rules of the game change.

You can look at nation’s developing across centuries, or stocks getting sold within seconds.

Time introduces patterns and trends that reveal important insights. Understanding seasonality, cyclical patterns, and long-term trends helps to understand the phenomenon.

You have to deal with date formats, periods, and missing days.

Time changes the nature of data. Your data is now dynamic, it evolves. Time orders your data.

Could I have known this already? Or am I modeling today, based on the tomorrow?

For neural networks, be careful with your mini-batch: Shuffling around the data is not valid anymore, the batches need to preserve the sequence of time.

Drawing samples becomes more difficult. If you want to downsample the data to have a smaller dataset, ensure you get data from every period.

More time is not the same as more observations/people/units. A person playing Russian Roulette a hundred times is different from a hundred people playing Russian Roulette once.

Keep an eye on the time.