jianlin666
commented
4 years ago 补充:
Warning: The scitype of X, in machine(model, X, ...) is incompatible with model=XGBoostClassifier @658:
│ scitype(X) = Table{Union{AbstractArray{ScientificTypes.Continuous,1}, AbstractArray{Count,1}, AbstractArray{Multiclass{2},1}, AbstractArray{Multiclass{3},1}, AbstractArray{OrderedFactor{53},1}, AbstractArray{OrderedFactor{66},1}, AbstractArray{OrderedFactor{2},1}}}
│ input_scitype(model) = Table{var"#s13"} where var"#s13"<:(AbstractArray{var"#s45",1} where var"#s45"<:ScientificTypes.Continuous).
└ @ MLJBase C:\Users\y84157557.julia\packages\MLJBase\cJmIS\src\machines.jl:76
nesteiner
Hi, 你们做的项目真的太棒了
在应用MLJ进行预测时, 碰到了模型不兼容科学类型的问题
数据来自kaggle
` using MLJ,Queryverse,PrettyPrinting,Random,StatsKit,LossFunctions,Plots train_data = Queryverse.load("D://juliacode//archive//train.csv") |> DataFrame id = train_data.id #用户id变量单独存放 train_data = select(train_data, Not(:id)) #剔除用户id变量
StatsKit.countmap(train_data.Response) #查看正负样本是否均衡
y, X = unpack(train_data, ==(:Response), colname -> true) #拆包自变量与目标变量
转换科学类型
X = coerce(X, autotype(X)) y = coerce(y, autotype(y)) X |> MLJ.schema `
│ Gender │ CategoricalValue{String,UInt32} │ Multiclass{2} │ │ Age │ CategoricalValue{Int64,UInt32} │ OrderedFactor{66} │ │ Driving_License │ CategoricalValue{Int64,UInt32} │ OrderedFactor{2} │ │ Region_Code │ CategoricalValue{Float64,UInt32} │ OrderedFactor{53} │ │ Previously_Insured │ CategoricalValue{Int64,UInt32} │ OrderedFactor{2} │ │ Vehicle_Age │ CategoricalValue{String,UInt32} │ Multiclass{3} │ │ Vehicle_Damage │ CategoricalValue{String,UInt32} │ Multiclass{2} │ │ Annual_Premium │ Float64 │ ScientificTypes.Continuous │ │ Policy_Sales_Channel │ Float64 │ ScientificTypes.Continuous │ │ Vintage │ Int64 │ Count
@load XGBoostClassifier xgb = XGBoostClassifier() mach = machine(xgb, X, y) fit!(mach) #XGB报错不兼容科学类型以下是报错内容:
MethodError: no method matching XGBoost.DMatrix(::Array{Any,2}; label=Bool[1, 0, 1, 0, 0, 0, 0, 1, 0, 0 … 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]) Closest candidates are: XGBoost.DMatrix(!Matched::Ptr{Nothing}) at C:\Users\y84157557.julia\packages\XGBoost\fI0vs\src\xgboost_lib.jl:21 got unsupported keyword argument "label" XGBoost.DMatrix(!Matched::String; silent) at C:\Users\y84157557.julia\packages\XGBoost\fI0vs\src\xgboost_lib.jl:27 got unsupported keyword argument "label" XGBoost.DMatrix(!Matched::SparseArrays.SparseMatrixCSC{K,V}) where {K<:Real, V<:Integer} at C:\Users\y84157557.julia\packages\XGBoost\fI0vs\src\xgboost_lib.jl:34 got unsupported keyword argument "label" ... in include_string at base\loading.jl:1088 in top-level scope at archive.jl:36 in fit! at MLJBase\cJmIS\src\machines.jl:479 in #fit!#87 at MLJBase\cJmIS\src\machines.jl:481 in fit_only! at MLJBase\cJmIS\src\machines.jl:389 in #fit_only!#84 at MLJBase\cJmIS\src\machines.jl:436
请问已经按照MLJ的规则进行科学类型的变换, 为何无法兼容XGB模型呢, 谢谢!