# Frequency of occurrence - dummy variables

I am thinking about it not the first time, namely if I have a variable that I want to convert later to the variable dummy (cities in this case), should I delete lines that occur less often than N times?

For example, the value of new york has occurred 400+ times but there are cities that only appeared once or twice.

What should I do with values ​​that have appeared only once or twice?

print(df[cities].value_counts())


Output:

city1         424
city2         107
city3          35
city4          33
city5          28
city6          24
city7          15
city8           7
city9           4
city10          3
city11          2
city12          1
city13          1
city14          1
city15          1
city16          1
city17          1


Having said that, I think that a good approach is to treat the cities with a lot of occurrences each on its own, and then group all others under a 'other' category.
Going further, you could have multiple 'other' groups, grouped by various criteria, for example, geographical criteria, or anything that makes sense in your context.