The "probability" of the nearest K-type neighbor

Question

I have a small set of data points (about 10) in a 2D space, and each of them has a category label. I want to classify a new data point based on existing data point labels and also associate a “probability” for belonging to any label class.

Is it appropriate to designate a new point based on the label to the nearest neighbor (for example, the nearest neighbor K, K = 1)? To get the probability, I want to rearrange all the marks and calculate the entire minimum distance of the unknown point and the rest and find the proportion of cases when the minimum distance is less than or equal to the distance that was used to mark it.

thank

Answer 1

Roweis KNN . "" , , j ,

,

d_ij - j.

Answer 2

Nearest Neighbor , , K-. , K, , , . :

P ( | z) = P (z | ) P ()/P (z) = K ()/K

, . , :

VP (x) = K/N ( V)

P (x) = K/NV ()

P (x = label) = K ()/N () V ( K () N () - )

P () = N ()/N.

K, , , , .

Answer 3

K- , , SVM. , . 1. K 2. K, .

LibSVM.

Answer 4

10 ; 1 5 6... — ,
1 1 1 5 6... — 3 ,
... .
10 , . 1 - (/) ** p,
. ( p .)

, .

: , 0 1, -. "" , , ,
-.

: academar.google.com "(single | 1) " > 300 ; "k " 3000.
( ), 10 k-NN ,
, 9 . .
, stats.stackexchange.com,

Answer 5

: .

, , X, Y . , .

Now suppose X, Y is a lat / long job for this person. The person working is not related to sharing the label (last name).

So, it depends on the semantics of your tags and axes.

NTN!