Code
include("../utils.jl")
import MLJ:predict,fit!,predict_mode,range
using DataFrames,MLJ,CSV,MLJModelInterface,GLMakie4-german-creditcard-logistics-reg
简介
- ref :german-creditcard
- scitype 转换 参考:autotype(d, :few_to_finite)方法
1. load package
2. data procsssing
Code
Xtrain, Xtest, ytrain, ytest,cat= load_german_creditcard();3. MLJ workflow
3.1 define model
Code
LogisticClassifier = @load LogisticClassifier pkg=MLJLinearModels
model=LogisticClassifier()
NuSVC = @load NuSVC pkg=LIBSVM
model2 = NuSVC()
KNNClassifier = @load KNNClassifier pkg=NearestNeighborModels
model3 = KNNClassifier(weights = NearestNeighborModels.Inverse())
"定义 几个 tune 参数的区间 "
k1 =range(model, :gamma, lower=0.1, upper=1.2);
k2 =range(model, :lambda, lower=0.1, upper=1.2);
k3 =range(model, :penalty, values=([:l2, :l1,:en,:none]));
k4 =range(model, :fit_intercept, values=([true, false]));
tuning_logistic = TunedModel(model=model,
resampling = CV(nfolds=4, rng=1234),
tuning = Grid(resolution=8),
range = [k1,k2],
measure=accuracy)
mach = machine(tuning_logistic, Xtrain, ytrain;scitype_check_level=0)|>fit![ Info: For silent loading, specify `verbosity=0`.
[ Info: For silent loading, specify `verbosity=0`.
[ Info: For silent loading, specify `verbosity=0`.
[ Info: Training machine(ProbabilisticTunedModel(model = LogisticClassifier(lambda = 2.220446049250313e-16, …), …), …).
[ Info: Attempting to evaluate 64 models.
Evaluating over 64 metamodels: 0%[> ] ETA: N/A┌ Warning: The number and/or types of data arguments do not match what the specified model
│ supports. Suppress this type check by specifying `scitype_check_level=0`.
│
│ Run `@doc MLJLinearModels.LogisticClassifier` to learn more about your model's requirements.
│
│ Commonly, but non exclusively, supervised models are constructed using the syntax
│ `machine(model, X, y)` or `machine(model, X, y, w)` while most other models are
│ constructed with `machine(model, X)`. Here `X` are features, `y` a target, and `w`
│ sample or class weights.
│
│ In general, data in `machine(model, data...)` is expected to satisfy
│
│ scitype(data) <: MLJ.fit_data_scitype(model)
│
│ In the present case:
│
│ scitype(data) = Tuple{Table{Union{AbstractVector{Continuous}, AbstractVector{OrderedFactor{33}}, AbstractVector{OrderedFactor{10}}, AbstractVector{OrderedFactor{5}}, AbstractVector{OrderedFactor{53}}, AbstractVector{OrderedFactor{3}}, AbstractVector{OrderedFactor{4}}, AbstractVector{OrderedFactor{2}}}}, AbstractVector{OrderedFactor{2}}}
│
│ fit_data_scitype(model) = Tuple{Table{<:AbstractVector{<:Continuous}}, AbstractVector{<:Finite}}
└ @ MLJBase ~/.julia/packages/MLJBase/fEiP2/src/machines.jl:230
Evaluating over 64 metamodels: 2%[> ] ETA: 0:13:55Evaluating over 64 metamodels: 3%[> ] ETA: 0:07:07Evaluating over 64 metamodels: 5%[=> ] ETA: 0:04:40Evaluating over 64 metamodels: 6%[=> ] ETA: 0:03:27Evaluating over 64 metamodels: 8%[=> ] ETA: 0:02:43Evaluating over 64 metamodels: 9%[==> ] ETA: 0:02:13Evaluating over 64 metamodels: 11%[==> ] ETA: 0:01:52Evaluating over 64 metamodels: 12%[===> ] ETA: 0:01:37Evaluating over 64 metamodels: 14%[===> ] ETA: 0:01:25Evaluating over 64 metamodels: 16%[===> ] ETA: 0:01:15Evaluating over 64 metamodels: 17%[====> ] ETA: 0:01:07Evaluating over 64 metamodels: 19%[====> ] ETA: 0:01:00Evaluating over 64 metamodels: 20%[=====> ] ETA: 0:00:54Evaluating over 64 metamodels: 22%[=====> ] ETA: 0:00:50Evaluating over 64 metamodels: 23%[=====> ] ETA: 0:00:45Evaluating over 64 metamodels: 25%[======> ] ETA: 0:00:42Evaluating over 64 metamodels: 27%[======> ] ETA: 0:00:38Evaluating over 64 metamodels: 28%[=======> ] ETA: 0:00:36Evaluating over 64 metamodels: 30%[=======> ] ETA: 0:00:33Evaluating over 64 metamodels: 31%[=======> ] ETA: 0:00:31Evaluating over 64 metamodels: 33%[========> ] ETA: 0:00:29Evaluating over 64 metamodels: 34%[========> ] ETA: 0:00:27Evaluating over 64 metamodels: 36%[========> ] ETA: 0:00:25Evaluating over 64 metamodels: 38%[=========> ] ETA: 0:00:23Evaluating over 64 metamodels: 39%[=========> ] ETA: 0:00:22Evaluating over 64 metamodels: 41%[==========> ] ETA: 0:00:20Evaluating over 64 metamodels: 42%[==========> ] ETA: 0:00:19Evaluating over 64 metamodels: 44%[==========> ] ETA: 0:00:18Evaluating over 64 metamodels: 45%[===========> ] ETA: 0:00:17Evaluating over 64 metamodels: 47%[===========> ] ETA: 0:00:16Evaluating over 64 metamodels: 48%[============> ] ETA: 0:00:15Evaluating over 64 metamodels: 50%[============> ] ETA: 0:00:14Evaluating over 64 metamodels: 52%[============> ] ETA: 0:00:13Evaluating over 64 metamodels: 53%[=============> ] ETA: 0:00:12Evaluating over 64 metamodels: 55%[=============> ] ETA: 0:00:12Evaluating over 64 metamodels: 56%[==============> ] ETA: 0:00:11Evaluating over 64 metamodels: 58%[==============> ] ETA: 0:00:10Evaluating over 64 metamodels: 59%[==============> ] ETA: 0:00:10Evaluating over 64 metamodels: 61%[===============> ] ETA: 0:00:09Evaluating over 64 metamodels: 62%[===============> ] ETA: 0:00:08Evaluating over 64 metamodels: 64%[================> ] ETA: 0:00:08Evaluating over 64 metamodels: 66%[================> ] ETA: 0:00:07Evaluating over 64 metamodels: 67%[================> ] ETA: 0:00:07Evaluating over 64 metamodels: 69%[=================> ] ETA: 0:00:06Evaluating over 64 metamodels: 70%[=================> ] ETA: 0:00:06Evaluating over 64 metamodels: 72%[=================> ] ETA: 0:00:06Evaluating over 64 metamodels: 73%[==================> ] ETA: 0:00:05Evaluating over 64 metamodels: 75%[==================> ] ETA: 0:00:05Evaluating over 64 metamodels: 77%[===================> ] ETA: 0:00:04Evaluating over 64 metamodels: 78%[===================> ] ETA: 0:00:04Evaluating over 64 metamodels: 80%[===================> ] ETA: 0:00:04Evaluating over 64 metamodels: 81%[====================> ] ETA: 0:00:03Evaluating over 64 metamodels: 83%[====================> ] ETA: 0:00:03Evaluating over 64 metamodels: 84%[=====================> ] ETA: 0:00:03Evaluating over 64 metamodels: 86%[=====================> ] ETA: 0:00:02Evaluating over 64 metamodels: 88%[=====================> ] ETA: 0:00:02Evaluating over 64 metamodels: 89%[======================> ] ETA: 0:00:02Evaluating over 64 metamodels: 91%[======================> ] ETA: 0:00:01Evaluating over 64 metamodels: 92%[=======================> ] ETA: 0:00:01Evaluating over 64 metamodels: 94%[=======================> ] ETA: 0:00:01Evaluating over 64 metamodels: 95%[=======================> ] ETA: 0:00:01Evaluating over 64 metamodels: 97%[========================>] ETA: 0:00:00Evaluating over 64 metamodels: 98%[========================>] ETA: 0:00:00Evaluating over 64 metamodels: 100%[=========================] Time: 0:00:14
┌ Warning: The number and/or types of data arguments do not match what the specified model
│ supports. Suppress this type check by specifying `scitype_check_level=0`.
│
│ Run `@doc MLJLinearModels.LogisticClassifier` to learn more about your model's requirements.
│
│ Commonly, but non exclusively, supervised models are constructed using the syntax
│ `machine(model, X, y)` or `machine(model, X, y, w)` while most other models are
│ constructed with `machine(model, X)`. Here `X` are features, `y` a target, and `w`
│ sample or class weights.
│
│ In general, data in `machine(model, data...)` is expected to satisfy
│
│ scitype(data) <: MLJ.fit_data_scitype(model)
│
│ In the present case:
│
│ scitype(data) = Tuple{Table{Union{AbstractVector{Continuous}, AbstractVector{OrderedFactor{33}}, AbstractVector{OrderedFactor{10}}, AbstractVector{OrderedFactor{5}}, AbstractVector{OrderedFactor{53}}, AbstractVector{OrderedFactor{3}}, AbstractVector{OrderedFactor{4}}, AbstractVector{OrderedFactor{2}}}}, AbstractVector{OrderedFactor{2}}}
│
│ fit_data_scitype(model) = Tuple{Table{<:AbstractVector{<:Continuous}}, AbstractVector{<:Finite}}
└ @ MLJBase ~/.julia/packages/MLJBase/fEiP2/src/machines.jl:230import MLJLinearModels ✔
import MLJLIBSVMInterface ✔
import NearestNeighborModels ✔trained Machine; does not cache data
model: ProbabilisticTunedModel(model = LogisticClassifier(lambda = 2.220446049250313e-16, …), …)
args:
1: Source @007 ⏎ Table{Union{AbstractVector{Continuous}, AbstractVector{OrderedFactor{33}}, AbstractVector{OrderedFactor{10}}, AbstractVector{OrderedFactor{5}}, AbstractVector{OrderedFactor{53}}, AbstractVector{OrderedFactor{3}}, AbstractVector{OrderedFactor{4}}, AbstractVector{OrderedFactor{2}}}}
2: Source @757 ⏎ AbstractVector{OrderedFactor{2}}3.2 predict test results
Code
yhat=predict_mode(mach, Xtest)|>Array
@info "german-creditcard 违约预测准确率"=>accuracy(ytest,yhat)|>d->round(d,digits=3)[ Info: "german-creditcard 违约预测准确率" => 0.74