func main() {
var tree base.Classifier
rand.Seed(44111342)
// Load in the iris dataset
iris, err := base.ParseCSVToInstances("/home/kralli/go/src/github.com/sjwhitworth/golearn/examples/datasets/iris_headers.csv", true)
if err != nil {
panic(err)
}
// Discretise the iris dataset with Chi-Merge
filt := filters.NewChiMergeFilter(iris, 0.999)
for _, a := range base.NonClassFloatAttributes(iris) {
filt.AddAttribute(a)
}
filt.Train()
irisf := base.NewLazilyFilteredInstances(iris, filt)
// Create a 60-40 training-test split
//testData
trainData, _ := base.InstancesTrainTestSplit(iris, 0.60)
findBestSplit(trainData)
//fmt.Println(trainData)
//fmt.Println(testData)
fmt.Println(tree)
fmt.Println(irisf)
}
func (lr *LogisticRegression) Predict(X base.FixedDataGrid) base.FixedDataGrid {
// Only support 1 class Attribute
classAttrs := X.AllClassAttributes()
if len(classAttrs) != 1 {
panic(fmt.Sprintf("%d Wrong number of classes", len(classAttrs)))
}
// Generate return structure
ret := base.GeneratePredictionVector(X)
classAttrSpecs := base.ResolveAttributes(ret, classAttrs)
// Retrieve numeric non-class Attributes
numericAttrs := base.NonClassFloatAttributes(X)
numericAttrSpecs := base.ResolveAttributes(X, numericAttrs)
// Allocate row storage
row := make([]float64, len(numericAttrSpecs))
X.MapOverRows(numericAttrSpecs, func(rowBytes [][]byte, rowNo int) (bool, error) {
for i, r := range rowBytes {
row[i] = base.UnpackBytesToFloat(r)
}
val := Predict(lr.model, row)
vals := base.PackFloatToBytes(val)
ret.Set(classAttrSpecs[0], rowNo, vals)
return true, nil
})
return ret
}
func BenchmarkBaggingRandomForestPredict(t *testing.B) {
inst, err := base.ParseCSVToInstances("../examples/datasets/iris_headers.csv", true)
if err != nil {
t.Fatal("Unable to parse CSV to instances: %s", err.Error())
}
rand.Seed(time.Now().UnixNano())
filt := filters.NewChiMergeFilter(inst, 0.90)
for _, a := range base.NonClassFloatAttributes(inst) {
filt.AddAttribute(a)
}
filt.Train()
instf := base.NewLazilyFilteredInstances(inst, filt)
rf := new(BaggedModel)
for i := 0; i < 10; i++ {
rf.AddModel(trees.NewRandomTree(2))
}
rf.Fit(instf)
t.ResetTimer()
for i := 0; i < 20; i++ {
rf.Predict(instf)
}
}
func TestRandomForest1(testEnv *testing.T) {
inst, err := base.ParseCSVToInstances("../examples/datasets/iris_headers.csv", true)
if err != nil {
panic(err)
}
rand.Seed(time.Now().UnixNano())
trainData, testData := base.InstancesTrainTestSplit(inst, 0.6)
filt := filters.NewChiMergeFilter(inst, 0.90)
for _, a := range base.NonClassFloatAttributes(inst) {
filt.AddAttribute(a)
}
filt.Train()
trainDataf := base.NewLazilyFilteredInstances(trainData, filt)
testDataf := base.NewLazilyFilteredInstances(testData, filt)
rf := new(BaggedModel)
for i := 0; i < 10; i++ {
rf.AddModel(trees.NewRandomTree(2))
}
rf.Fit(trainDataf)
fmt.Println(rf)
predictions := rf.Predict(testDataf)
fmt.Println(predictions)
confusionMat := eval.GetConfusionMatrix(testDataf, predictions)
fmt.Println(confusionMat)
fmt.Println(eval.GetMacroPrecision(confusionMat))
fmt.Println(eval.GetMacroRecall(confusionMat))
fmt.Println(eval.GetSummary(confusionMat))
}
func TestRandomTreeClassificationAfterDiscretisation(t *testing.T) {
Convey("Predictions on filtered data with a Random Tree", t, func() {
instances, err := base.ParseCSVToInstances("../examples/datasets/iris_headers.csv", true)
So(err, ShouldBeNil)
trainData, testData := base.InstancesTrainTestSplit(instances, 0.6)
filter := filters.NewChiMergeFilter(instances, 0.9)
for _, a := range base.NonClassFloatAttributes(instances) {
filter.AddAttribute(a)
}
filter.Train()
filteredTrainData := base.NewLazilyFilteredInstances(trainData, filter)
filteredTestData := base.NewLazilyFilteredInstances(testData, filter)
verifyTreeClassification(filteredTrainData, filteredTestData)
})
}
func processData(x base.FixedDataGrid) instances {
_, rows := x.Size()
result := make(instances, rows)
// Retrieve numeric non-class Attributes
numericAttrs := base.NonClassFloatAttributes(x)
numericAttrSpecs := base.ResolveAttributes(x, numericAttrs)
// Retrieve class Attributes
classAttrs := x.AllClassAttributes()
if len(classAttrs) != 1 {
panic("Only one classAttribute supported!")
}
// Check that the class Attribute is categorical
// (with two values) or binary
classAttr := classAttrs[0]
if attr, ok := classAttr.(*base.CategoricalAttribute); ok {
if len(attr.GetValues()) != 2 {
panic("To many values for Attribute!")
}
} else if _, ok := classAttr.(*base.BinaryAttribute); ok {
} else {
panic("Wrong class Attribute type!")
}
// Convert each row
x.MapOverRows(numericAttrSpecs, func(row [][]byte, rowNo int) (bool, error) {
// Allocate a new row
probRow := make([]float64, len(numericAttrSpecs))
// Read out the row
for i, _ := range numericAttrSpecs {
probRow[i] = base.UnpackBytesToFloat(row[i])
}
// Get the class for the values
class := base.GetClass(x, rowNo)
instance := instance{class, probRow}
result[rowNo] = instance
return true, nil
})
return result
}
func TestRandomForest(t *testing.T) {
Convey("Given a valid CSV file", t, func() {
inst, err := base.ParseCSVToInstances("../examples/datasets/iris_headers.csv", true)
So(err, ShouldBeNil)
Convey("When Chi-Merge filtering the data", func() {
filt := filters.NewChiMergeFilter(inst, 0.90)
for _, a := range base.NonClassFloatAttributes(inst) {
filt.AddAttribute(a)
}
filt.Train()
instf := base.NewLazilyFilteredInstances(inst, filt)
Convey("Splitting the data into test and training sets", func() {
trainData, testData := base.InstancesTrainTestSplit(instf, 0.60)
Convey("Fitting and predicting with a Random Forest", func() {
rf := NewRandomForest(10, 3)
err = rf.Fit(trainData)
So(err, ShouldBeNil)
predictions, err := rf.Predict(testData)
So(err, ShouldBeNil)
confusionMat, err := evaluation.GetConfusionMatrix(testData, predictions)
So(err, ShouldBeNil)
Convey("Predictions should be somewhat accurate", func() {
So(evaluation.GetAccuracy(confusionMat), ShouldBeGreaterThan, 0.35)
})
})
})
})
Convey("Fitting with a Random Forest with too many features compared to the data", func() {
rf := NewRandomForest(10, len(base.NonClassAttributes(inst))+1)
err = rf.Fit(inst)
Convey("Should return an error", func() {
So(err, ShouldNotBeNil)
})
})
})
}
func TestBaggedModelRandomForest(t *testing.T) {
Convey("Given data", t, func() {
inst, err := base.ParseCSVToInstances("../examples/datasets/iris_headers.csv", true)
So(err, ShouldBeNil)
Convey("Splitting the data into training and test data", func() {
trainData, testData := base.InstancesTrainTestSplit(inst, 0.6)
Convey("Filtering the split datasets", func() {
rand.Seed(time.Now().UnixNano())
filt := filters.NewChiMergeFilter(inst, 0.90)
for _, a := range base.NonClassFloatAttributes(inst) {
filt.AddAttribute(a)
}
filt.Train()
trainDataf := base.NewLazilyFilteredInstances(trainData, filt)
testDataf := base.NewLazilyFilteredInstances(testData, filt)
Convey("Fitting and Predicting with a Bagged Model of 10 Random Trees", func() {
rf := new(BaggedModel)
for i := 0; i < 10; i++ {
rf.AddModel(trees.NewRandomTree(2))
}
rf.Fit(trainDataf)
predictions := rf.Predict(testDataf)
confusionMat, err := evaluation.GetConfusionMatrix(testDataf, predictions)
So(err, ShouldBeNil)
Convey("Predictions are somewhat accurate", func() {
So(evaluation.GetAccuracy(confusionMat), ShouldBeGreaterThan, 0.5)
})
})
})
})
})
}
func convertInstancesToProblemVec(X base.FixedDataGrid) [][]float64 {
// Allocate problem array
_, rows := X.Size()
problemVec := make([][]float64, rows)
// Retrieve numeric non-class Attributes
numericAttrs := base.NonClassFloatAttributes(X)
numericAttrSpecs := base.ResolveAttributes(X, numericAttrs)
// Convert each row
X.MapOverRows(numericAttrSpecs, func(row [][]byte, rowNo int) (bool, error) {
// Allocate a new row
probRow := make([]float64, len(numericAttrSpecs))
// Read out the row
for i, _ := range numericAttrSpecs {
probRow[i] = base.UnpackBytesToFloat(row[i])
}
// Add the row
problemVec[rowNo] = probRow
return true, nil
})
return problemVec
}
func TestRandomForest1(testEnv *testing.T) {
inst, err := base.ParseCSVToInstances("../examples/datasets/iris_headers.csv", true)
if err != nil {
panic(err)
}
filt := filters.NewChiMergeFilter(inst, 0.90)
for _, a := range base.NonClassFloatAttributes(inst) {
filt.AddAttribute(a)
}
filt.Train()
instf := base.NewLazilyFilteredInstances(inst, filt)
trainData, testData := base.InstancesTrainTestSplit(instf, 0.60)
rf := NewRandomForest(10, 3)
rf.Fit(trainData)
predictions := rf.Predict(testData)
fmt.Println(predictions)
confusionMat := eval.GetConfusionMatrix(testData, predictions)
fmt.Println(confusionMat)
fmt.Println(eval.GetSummary(confusionMat))
}
func main() {
var tree base.Classifier
rand.Seed(44111342)
// Load in the iris dataset
iris, err := base.ParseCSVToInstances("../datasets/iris_headers.csv", true)
if err != nil {
panic(err)
}
// Discretise the iris dataset with Chi-Merge
filt := filters.NewChiMergeFilter(iris, 0.999)
for _, a := range base.NonClassFloatAttributes(iris) {
filt.AddAttribute(a)
}
filt.Train()
irisf := base.NewLazilyFilteredInstances(iris, filt)
// Create a 60-40 training-test split
trainData, testData := base.InstancesTrainTestSplit(irisf, 0.60)
//
// First up, use ID3
//
tree = trees.NewID3DecisionTree(0.6)
// (Parameter controls train-prune split.)
// Train the ID3 tree
err = tree.Fit(trainData)
if err != nil {
panic(err)
}
// Generate predictions
predictions, err := tree.Predict(testData)
if err != nil {
panic(err)
}
// Evaluate
fmt.Println("ID3 Performance (information gain)")
cf, err := evaluation.GetConfusionMatrix(testData, predictions)
if err != nil {
panic(fmt.Sprintf("Unable to get confusion matrix: %s", err.Error()))
}
fmt.Println(evaluation.GetSummary(cf))
tree = trees.NewID3DecisionTreeFromRule(0.6, new(trees.InformationGainRatioRuleGenerator))
// (Parameter controls train-prune split.)
// Train the ID3 tree
err = tree.Fit(trainData)
if err != nil {
panic(err)
}
// Generate predictions
predictions, err = tree.Predict(testData)
if err != nil {
panic(err)
}
// Evaluate
fmt.Println("ID3 Performance (information gain ratio)")
cf, err = evaluation.GetConfusionMatrix(testData, predictions)
if err != nil {
panic(fmt.Sprintf("Unable to get confusion matrix: %s", err.Error()))
}
fmt.Println(evaluation.GetSummary(cf))
tree = trees.NewID3DecisionTreeFromRule(0.6, new(trees.GiniCoefficientRuleGenerator))
// (Parameter controls train-prune split.)
// Train the ID3 tree
err = tree.Fit(trainData)
if err != nil {
panic(err)
}
// Generate predictions
predictions, err = tree.Predict(testData)
if err != nil {
panic(err)
}
// Evaluate
fmt.Println("ID3 Performance (gini index generator)")
cf, err = evaluation.GetConfusionMatrix(testData, predictions)
if err != nil {
panic(fmt.Sprintf("Unable to get confusion matrix: %s", err.Error()))
}
fmt.Println(evaluation.GetSummary(cf))
//
// Next up, Random Trees
//
// Consider two randomly-chosen attributes
tree = trees.NewRandomTree(2)
err = tree.Fit(testData)
if err != nil {
panic(err)
}
predictions, err = tree.Predict(testData)
if err != nil {
panic(err)
}
fmt.Println("RandomTree Performance")
cf, err = evaluation.GetConfusionMatrix(testData, predictions)
if err != nil {
panic(fmt.Sprintf("Unable to get confusion matrix: %s", err.Error()))
}
fmt.Println(evaluation.GetSummary(cf))
//
// Finally, Random Forests
//
tree = ensemble.NewRandomForest(70, 3)
err = tree.Fit(trainData)
if err != nil {
panic(err)
}
predictions, err = tree.Predict(testData)
if err != nil {
panic(err)
}
fmt.Println("RandomForest Performance")
cf, err = evaluation.GetConfusionMatrix(testData, predictions)
if err != nil {
panic(fmt.Sprintf("Unable to get confusion matrix: %s", err.Error()))
}
fmt.Println(evaluation.GetSummary(cf))
}
func main() {
var tree base.Classifier
rand.Seed(time.Now().UTC().UnixNano())
// Load in the iris dataset
iris, err := base.ParseCSVToInstances("../datasets/iris_headers.csv", true)
if err != nil {
panic(err)
}
// Discretise the iris dataset with Chi-Merge
filt := filters.NewChiMergeFilter(iris, 0.99)
for _, a := range base.NonClassFloatAttributes(iris) {
filt.AddAttribute(a)
}
filt.Train()
irisf := base.NewLazilyFilteredInstances(iris, filt)
// Create a 60-40 training-test split
trainData, testData := base.InstancesTrainTestSplit(irisf, 0.60)
//
// First up, use ID3
//
tree = trees.NewID3DecisionTree(0.6)
// (Parameter controls train-prune split.)
// Train the ID3 tree
tree.Fit(trainData)
// Generate predictions
predictions := tree.Predict(testData)
// Evaluate
fmt.Println("ID3 Performance")
cf := eval.GetConfusionMatrix(testData, predictions)
fmt.Println(eval.GetSummary(cf))
//
// Next up, Random Trees
//
// Consider two randomly-chosen attributes
tree = trees.NewRandomTree(2)
tree.Fit(testData)
predictions = tree.Predict(testData)
fmt.Println("RandomTree Performance")
cf = eval.GetConfusionMatrix(testData, predictions)
fmt.Println(eval.GetSummary(cf))
//
// Finally, Random Forests
//
tree = ensemble.NewRandomForest(100, 3)
tree.Fit(trainData)
predictions = tree.Predict(testData)
fmt.Println("RandomForest Performance")
cf = eval.GetConfusionMatrix(testData, predictions)
fmt.Println(eval.GetSummary(cf))
}
func TestRandomTreeClassification(t *testing.T) {
Convey("Predictions on filtered data with a Random Tree", t, func() {
instances, err := base.ParseCSVToInstances("../examples/datasets/iris_headers.csv", true)
So(err, ShouldBeNil)
trainData, testData := base.InstancesTrainTestSplit(instances, 0.6)
filter := filters.NewChiMergeFilter(instances, 0.9)
for _, a := range base.NonClassFloatAttributes(instances) {
filter.AddAttribute(a)
}
filter.Train()
filteredTrainData := base.NewLazilyFilteredInstances(trainData, filter)
filteredTestData := base.NewLazilyFilteredInstances(testData, filter)
Convey("Using InferID3Tree to create the tree and do the fitting", func() {
Convey("Using a RandomTreeRule", func() {
randomTreeRuleGenerator := new(RandomTreeRuleGenerator)
randomTreeRuleGenerator.Attributes = 2
root := InferID3Tree(filteredTrainData, randomTreeRuleGenerator)
Convey("Predicting with the tree", func() {
predictions, err := root.Predict(filteredTestData)
So(err, ShouldBeNil)
confusionMatrix, err := evaluation.GetConfusionMatrix(filteredTestData, predictions)
So(err, ShouldBeNil)
Convey("Predictions should be somewhat accurate", func() {
So(evaluation.GetAccuracy(confusionMatrix), ShouldBeGreaterThan, 0.5)
})
})
})
Convey("Using a InformationGainRule", func() {
informationGainRuleGenerator := new(InformationGainRuleGenerator)
root := InferID3Tree(filteredTrainData, informationGainRuleGenerator)
Convey("Predicting with the tree", func() {
predictions, err := root.Predict(filteredTestData)
So(err, ShouldBeNil)
confusionMatrix, err := evaluation.GetConfusionMatrix(filteredTestData, predictions)
So(err, ShouldBeNil)
Convey("Predictions should be somewhat accurate", func() {
So(evaluation.GetAccuracy(confusionMatrix), ShouldBeGreaterThan, 0.5)
})
})
})
})
Convey("Using NewRandomTree to create the tree", func() {
root := NewRandomTree(2)
Convey("Fitting with the tree", func() {
err = root.Fit(filteredTrainData)
So(err, ShouldBeNil)
Convey("Predicting with the tree, *without* pruning first", func() {
predictions, err := root.Predict(filteredTestData)
So(err, ShouldBeNil)
confusionMatrix, err := evaluation.GetConfusionMatrix(filteredTestData, predictions)
So(err, ShouldBeNil)
Convey("Predictions should be somewhat accurate", func() {
So(evaluation.GetAccuracy(confusionMatrix), ShouldBeGreaterThan, 0.5)
})
})
Convey("Predicting with the tree, pruning first", func() {
root.Prune(filteredTestData)
predictions, err := root.Predict(filteredTestData)
So(err, ShouldBeNil)
confusionMatrix, err := evaluation.GetConfusionMatrix(filteredTestData, predictions)
So(err, ShouldBeNil)
Convey("Predictions should be somewhat accurate", func() {
So(evaluation.GetAccuracy(confusionMatrix), ShouldBeGreaterThan, 0.4)
})
})
})
})
})
}
