1-catboost-classfication

dataset

dataset 参见 clustering-exercises dataset

1. load package

Code

include("../utils.jl")
import Plots:scatter!,contourf
import MLJ:predict,predict_mode,measures
using Plots, MLJ, CSV, DataFrames
using CatBoost.MLJCatBoostInterface

2. load data

Code

  df=load_csv("basic1")
  cat=df[:,:color]|>levels|>length # 类别
  ytrain, Xtrain =  unpack(df, ==(:color), rng=123);
  first(df,10)

10×3 DataFrame

Row	x	y	color
	Float64	Float64	Int64
1	79.4083	152.834	0
2	98.0463	186.911	0
3	240.579	48.4737	1
4	109.687	277.946	0
5	249.626	229.753	1
6	100.785	281.983	0
7	235.33	109.54	1
8	262.352	64.5746	1
9	76.5589	204.296	0
10	245.558	134.502	1

3. MLJ workflow

3.1 fitting model

Code

    catboost = CatBoostClassifier(iterations=2,learning_rate=0.20)
    mach = machine(catboost, Xtrain, ytrain;scitype_check_level=0)|>fit!
    tx,ty,xtest=boundary_data(df)  # boudary data and xtest 
    ytest = predict_mode(mach, xtest)[:,1]|>Array

[ Info: Training machine(CatBoostClassifier(iterations = 2, …), …).

40000-element Vector{Int64}:
 0
 0
 0
 0
 0
 0
 0
 0
 0
 0
 0
 0
 0
 ⋮
 3
 3
 3
 3
 3
 3
 3
 3
 3
 3
 3
 3

3.2 plot results

Code

contourf(tx,ty,ytest,levels=cat,color=cgrad(:redsblues),alpha=0.7)
p1=scatter!(df[:,:x],df[:,:y],group=df[:,:color],label=false,ms=3,alpha=0.3)