design.puml

@startuml
'https://plantuml.com/class-diagram
skinparam classAttributeIconSize 0

abstract class RandomForest {
-num_trees : int
-min_size : int
-max_depth: int
-ratio_samples : float
-num_random_features : float
-multiprocessing : boolean
-extra_trees : boolean

+fit (x: float[][], y: float[])
+predict (X : float[][]) : float[]
+feature_importance(): dictionary
+print_trees():
-predict_standard (X : float[][]): float[]
-predict_multiprocessing (X : float[][]): float[]
-make_decision_trees_standard (dataset : Dataset)
-make_decision_trees_multiprocessing (dataset : Dataset)
-target_predict (x : float[] ) : int
-target_fit (dataset : Dataset) : Node
-make_node (dataset : Dataset, depth : int) : Node
-make_parent_or_leaf (dataset : Dataset, depth : int) : Node
//-make_leaf (dataset : Dataset, depth : int) : Leaf//
-best_split (idx_features : int[], dataset : Dataset) : Dataset[2]
-cart_cost (left_dataset : Dataset, right_dataset : Dataset) : float
//-make_impurity(name_impurity : String) : ImpurityMeasures//
//-combine_predictions(predictions: float[]):float//
}

class RandomForestClassifier extends RandomForest{
-make_impurity(name_impurity : String): ImpurityMeasures
-combine_predictions(predictions: float[]) : float
-make_leaf(dataset : Dataset): Leaf
}
class RandomForestRegressor extends RandomForest{
-make_impurity(name_impurity : String): ImpurityMeasures
-combine_predictions(predictions: float[]) : float
-make_leaf(dataset : Dataset): Leaf
}

class IsolationForest extends RandomForest{
-train_size : int
-test_size : int
+fit(X : float[][])
+predict(X : float[][]) : float[]
-combine_predictions(predictions: float[]) : float
-make_leaf(dataset : Dataset): Leaf
-best_split(idx_features : int[], dataset: Dataset) : Dataset[2]
}

interface Visitor{
//visit_parent(parent: Parent)://
//visit_leaf(leaf: Leaf)//
}

class FeatureImportance implements Visitor{
-occurrences: dictionary
+visit_parent(parent: Parent)
+visit_leaf(leaf: Leaf)
}

class PrinterTree implements Visitor{
-depth: int
+visit_parent(parent: Parent)
+visit_leaf(leaf: Leaf)
}

interface ImpurityMeasures{
//+compute_impurity(prob : float[])  : float//
}

class Gini implements ImpurityMeasures {
+comupte_impurity(dataset: Dataset()) :  float
}

class Entropy implements ImpurityMeasures {
+comupte_impurity(dataset: Dataset()) : float
}

class SumSquareError implements ImpurityMeasures{
+comupte_impurity(dataset: Dataset()) : float
}

class Dataset {
-X: float [][]
-y: int[]
-num_samples : int
-num_features : int
+random_sampling(ratio_samples : float) : Dataset
+split(idx : int, value : double) : Dataset[2]
+most_frequent_label() : int
+probability() : float[]
+mean_value(): float
}

interface Node {
//+predict(x:float[]) : int//
//+accept_visitor(visitor: Visitor) : int//
}

class Leaf implements Node{
-value_or_label_or_depth: float
+predict(x : float[]): float
+accept_visitor(visitor : Visitor)
}

class Parent implements Node{
-idx_features : int
-value : float
-predict(x: float[]) : float
+accept_visitor(visitor : Visitor)

}

note right of Leaf::predict
return label
end note

note left of Parent::predict
if x[feature] <= value:
    return left_child.predict(x)
else:
    return right_child.predict(x)
end note

note left of Node
design pattern
**composite**
end note
note right of ImpurityMeasures
design pattern
**strategy**
end note
note right of Visitor
design pattern
**visitor**
end note
note right of RandomForest
design pattern
**template method**
end note

Parent --> "2" Node: left_child, right_child
RandomForest *--> "1" ImpurityMeasures : impurity_measure
RandomForest *--left--> "1.." Node: decision_trees


@enduml