Reconstruct. This should only be utilised Should you have a single dtype for all columns. Or else, you'll end up having dtype object for all columns and changing them back needs additional dictionary work.df.columns = new where by new is the listing of new columns names is as simple as it will get. The downside of the strategy is the fact it needs