// TestCSV
func main() {
var reader io.Reader
var err error
if reader, err = os.Open("pima-indians-diabetes.csv"); err != nil {
log.Fatalf("%v", err)
}
handler := &Handler{}
err = db.Import(reader, handler)
targetBool := make([]bool, handler.target.Len())
targetNumeric := make([]float64, handler.target.Len())
for i := 0; i < handler.target.Len(); i++ {
targetBool[i] = handler.target.NumericValue(i) == 1.0
targetNumeric[i] = handler.target.NumericValue(i)
}
if err != nil {
log.Fatal("Import returned error")
}
fmt.Printf("target: %v\n", handler.target)
rf := db.NewRandomForestRegressor(100)
oobScores := rf.Fit(handler.features, handler.target)
auc := db.ROCArea(oobScores, targetBool)
mse := db.MSE(oobScores, targetNumeric)
variance := stats.Variance(targetNumeric)
fmt.Printf("ROCArea: %v MSE: %v Var: %v\n", auc, mse, variance)
}
func TestRandomForest(test *testing.T) {
x := db.NewNumericFeature(nil)
x.Add(0, 1, 2, 3, 4, 5, 6, 7)
y := db.NewNumericFeature(nil)
y.Add(1, 2, 1, 2, 1, 2, 1, 2)
z := db.NewNumericFeature(nil)
z.Add(1, 1, 1, 1, 2, 2, 2, 2)
t := db.NewNumericFeature(nil)
// t.Add(3, 0, 3, 1, 7, 6, 5, -1)
t.Add(0, 0, 0, 0, 1, 1, 1, 1)
rfFeatures := []db.OrderedFeature{}
for _, f := range []*db.NumericFeature{x, y, z} {
rfFeatures = append(rfFeatures, f)
}
rf := db.NewRandomForestRegressor(10)
oob := rf.Fit(rfFeatures, t)
tEstimate := rf.Predict([]db.Feature{x, y, z})
if len(tEstimate) != t.Len() {
test.Errorf("rf.Predict() returned result of length: %d; expected %d", len(tEstimate), t.Len())
}
fmt.Printf("oob preds: %v\n", oob)
for i, te := range tEstimate {
fmt.Printf("predicted: %v; actual: %v\n", te, t.Value(i))
// if te != t.Value(i) {
// test.Errorf("Row %d: predicted %v; actual %v", i, te, t.Value///(i))
// }
}
fmt.Printf("feature importances (forest): %v\n", rf.Importances())
}