func TestCancer(t *testing.T) {
X, y := datasets.Load("cancer")
datasets.RandomShuffle(X, y)
XTrain, XTest := X[:67], X[67:]
yTrain, yTest := y[:67], y[67:]
beta := LinearRegression(XTrain, yTrain)
// validate on held out data
yPred := matrix.VecMult(XTest, beta)
fmt.Println("cancer error", metrics.MeanSquaredError(yPred, yTest))
}
func testDataset(X [][]float64, y []float64) float64 {
runtime.GOMAXPROCS(runtime.NumCPU())
datasets.RandomShuffle(X, y)
kFolds := 5
foldSize := int(float64(len(X)) / float64(kFolds))
foldSum := 0.0
for k := 0; k < kFolds; k++ {
start := k * foldSize
for j := 0; j < foldSize; j++ {
X[j], X[start+j] = X[start+j], X[j]
y[j], y[start+j] = y[start+j], y[j]
}
tree, _ := DecisionTree(1, GINI)
tree.Fit(X[foldSize:], y[foldSize:])
yPred := tree.Classify(X[:foldSize])
foldSum += metrics.Accuracy(yPred, y[:foldSize])
}
return foldSum / float64(kFolds)
}
