Using kNN with weighted dataset

Question

I have a dataset df:

	category	var 1	...	var 32	weighting	country
1	blue	1.0		54.2	3.0	US
2	pink	0.0		101.0	1.0	other
3	blue	1.0		49.9	3.0	US
4	green	1.0		72.2	9.0	US

I'm using the kNN classifier (on the country variable) but need it to take into account the current dataset weights I have included. After looking at the sklearn pack I can see the KNeighborsClassifier() does have a weight argument, can I set this argument 'weight = df.weighting'? or do I have to go about this another way?

Answer 1

you can explode samples by weight, for example, or you can think about creating custom weighted distance function

for weight, x_sample, y_sample in zip(sample_weights, X, y):
    weighted_X.extend([x_sample] * int(weight))
    weighted_y.extend([y_sample] * int(weight))