fix tests in regression

Author: Eight1911

src/regression/bench.jl

@@ -0,0 +1,47 @@
+
+
+include("../../src/DecisionTree.jl")
+
+function loaddata()
+    f = open("data/digits.csv")
+    data = readlines(f)[2:end]
+    data = [[parse(Float32, i) 
+        for i in split(row, ",")]
+        for row in data]
+    data = hcat(data...)
+    Y = Int.(data[1, 1:end]) .+ 1
+    X = convert(Matrix, transpose(data[2:end, 1:end]))
+    return X, Y
+end
+
+num_leaves(node::DecisionTree.Node) = num_leaves(node.left) + num_leaves(node.right)
+num_leaves(node::DecisionTree.Leaf) = 1
+
+X, Y = loaddata()
+
+# for compilation
+for i in 1:10
+    t = DecisionTree.build_tree(Y, X)
+end
+
+println("[ === CLASSIFICATION BENCHMARK === ]")
+for j in 1:3
+    @time for i in 1:100
+        tree = DecisionTree.build_tree(Y, X)
+    end
+end
+
+
+Y = float.(Y) # labels/targets to Float to enable regression
+
+# for compilation
+for i in 1:10
+    t = DecisionTree.build_tree(Y, X)
+end
+
+println("[ === REGRESSION BENCHMARK === ]")
+for j in 1:3
+    @time for i in 1:100
+        tree = DecisionTree.build_tree(Y, X)
+    end
+end

test/bench/trees.jl

@@ -29,7 +29,7 @@ min_samples_leaf = 5; max_depth = 3; n_subfeatures = 0;
 model = build_tree(labels, features, min_samples_leaf, n_subfeatures, max_depth)
 @test depth(model) == max_depth
 
-min_samples_leaf = 1; n_subfeatures = 0; max_depth = -1; min_samples_split = 100;
+min_samples_leaf = 1; n_subfeatures = 0; max_depth = -1; min_samples_split = 300;
 model = build_tree(labels, features, min_samples_leaf, n_subfeatures, max_depth, min_samples_split)
 preds = apply_tree(model, features);
 @test R2(labels, preds) < 0.8