added simple classify method for kmeans (#255)

Author: ExpandingMan

Project.toml

@@ -1,6 +1,6 @@
 name = "OnlineStats"
 uuid = "a15396b6-48d5-5d58-9928-6d29437db91e"
-version = "1.5.14"
+version = "1.5.15"
 
 [deps]
 AbstractTrees = "1520ce14-60c1-5f80-bbc7-55ef81b5835c"

src/stats/stats.jl

@@ -265,6 +265,8 @@ Approximate K-Means clustering of `k` clusters.
     o = fit!(KMeans(2, 2), eachrow(x))
 
     sort!(o; rev=true)  # Order clusters by number of observations
+
+    classify(o, x[1])  # returns index of cluster closest to x[1]
 """
 mutable struct KMeans{T, C <: NTuple{N, Cluster{T}} where N, W} <: OnlineStat{VectorOb}
     value::C
@@ -305,6 +307,8 @@ function _fit!(o::KMeans{T}, x) where {T}
     end
 end
 
+classify(o::KMeans, x) = findmin(c -> norm(x .- c.value), o.value)[2]
+
 #-----------------------------------------------------------------------# MovingTimeWindow
 """
     MovingTimeWindow{T<:TimeType, S}(window::Dates.Period)

test/runtests.jl

@@ -299,6 +299,10 @@ end
     o = fit!(KMeans(2), eachrow(ymat))
     sort!(o, rev=true)
     @test o.value[1].n ≥ o.value[2].n
+    
+    x = [repeat([[1.0, 1.0]], 3); repeat([[-1.0, -1.0]], 3)]
+    o = fit!(KMeans(2), (ξ for ξ ∈ x))
+    @test classify(o, x[1]) ≠ classify(o, x[4])
 end
 #-----------------------------------------------------------------------# LinReg
 @testset "LinReg" begin