pandas groupby based on multi-columns, but keep the most repeated duplicates number on other column

You can combine .groupby with Series.mode:

x = df.groupby(["lon", "lat"])["output"].apply(lambda x: x.mode()[0])
print(x.reset_index())

Prints:

         lon        lat       output
0 -47.812224 -19.043365  1890.283215
1 -47.812014 -19.007094  1813.785728
2 -47.811177 -19.008053  1763.091936
3 -47.774424 -19.008700  2172.781911
4 -47.774166 -19.043847  2897.092502
5 -47.773958 -19.007574  2179.670924
6 -47.773121 -19.008533  2236.769862

We can use the .groupby aggregate methods as an alternative to the Andrej's method of using .apply and computing for each row.

While it does solve our problem .apply methods tend to become slower for a large dataset due to lack of vectorization.

Also, reset_index works much faster when used along with 'inplace=True'.

%%timeit
df.groupby(['lat', 'lon']).agg(pd.Series.mode).reset_index(inplace=True)

Hope that helps!!