Merge pull request #48 from JuliaAI/add-tests-categorical-features

Add generic MLJ interface tests for categorical features

test/runtests.jl

@@ -8,19 +8,41 @@ using MLJGLMInterface
 using GLM: coeftable
 import GLM
 import MLJTestInterface
+using Tables
 
 using  Distributions: Normal, Poisson, Uniform
 import StableRNGs
 using Tables
 
 expit(X) = 1 ./ (1 .+ exp.(-X))
 
-# TODO: Add more datasets to the following generic interface tests after #45 is merged
+# synthesize small data sets with mixed features:
+
+n = 100
+X_regression, y_regression = MLJBase.make_regression(n, 3)
+outlook = categorical(rand(["sunny", "overcast", "rainy"], n))
+temperature = categorical(
+    rand(["cold", "mild", "hot"], n),
+    ordered=true,
+    levels=["cold", "mild", "hot"],
+    )
+X = merge(
+    Tables.columntable(X_regression),
+    (; outlook, temperature),
+)
+y_binary = categorical((temperature .== "mild") .| (outlook .== "sunny"))
+y_count = map(X.x1) do x
+    floor(Int, 10*abs(x))
+end
+mixed_binary = (X, y_binary)
+mixed_count = (X, y_count)
+mixed_regression = (X, y_regression)
 
 @testset "generic interface tests" begin
     @testset "LinearRegressor" begin
         for data in [
-                        MLJTestInterface.make_regression(),
+            MLJTestInterface.make_regression(),
+            mixed_regression,
             ]
             failures, summary = MLJTestInterface.test(
                 [LinearRegressor,],
@@ -36,6 +58,7 @@ expit(X) = 1 ./ (1 .+ exp.(-X))
     @testset "LinearCountRegressor" begin
         for data in [
             MLJTestInterface.make_count(),
+            mixed_count,
             ]
             failures, summary = MLJTestInterface.test(
                 [LinearCountRegressor,],
@@ -51,6 +74,7 @@ expit(X) = 1 ./ (1 .+ exp.(-X))
     @testset "LinearBinaryClassifier" begin
         for data in [
             MLJTestInterface.make_binary(),
+            mixed_binary,
             ]
             failures, summary = MLJTestInterface.test(
                 [LinearBinaryClassifier,],