{ Analytica Model Python_decision_tree_example, encoding="UTF-8" }
SoftwareVersion 7.1.8


{ System Variables with non-default values: }
SampleSize := 1000
TypeChecking := 1
Checking := 1
SaveOptions := 2
SaveValues := 0
PythonModelEnvironment := Analytica_Iris_Tree
WindState PythonModelEnvironment: 2,98,82,720,350
PythonStartupCode := import numpy~
import numpy as np~
from sklearn.datasets import load_iris
WindState PythonStartupCode: 2,98,82,720,350

{!-50299|DiagramColor Model: 0x00ffffff}
{!-50299|DiagramColor Module: 0x00ffffff}
{!-50299|DiagramColor LinkModule: 0x00ffffff}
{!-50299|DiagramColor Library: 0x00ffffff}
{!-50299|DiagramColor LinkLibrary: 0x00ffffff}
{!-50299|DiagramColor Form: 0x00ffffff}
NodeInfo FormNode: 1,0,0,,0,0,,,,0,,,0
{!-50299|NodeColor Text: 0x00f3f3f3}




Model Python_decision_tree_example
Title: Python Decision Tree Example
Description: Documentation example showing how to load scikit-learn's Iris data, train a CART decision tree in Python, and use its predictions in Analytica. Its setup button creates the required Analytica_Iris_Tree Conda environment without saving any machine-specific paths.
Author: Assista~
Lumina Decision Systems
Date: Thu, Jul 16, 2026 3:18 PM
DiagState: 2,0,1,1560,656,17,10
WindState: 2,98,82,720,350
FontStyle: Arial,15
FileInfo: 0,Model Python_decision_tree_example,2,2,0,0,W:\Testmodels\Python Decision Tree Example.ana
{!70104|Att_GUID: d71e036b-3fa7-460a-877f-067b3aa36e90}

Module Iris_Decision_Tree
Title: Iris decision tree
Description: Setup, data viewing, and decision-tree training for a Python-integration documentation example. The setup button creates the model-specific Conda environment Analytica_Iris_Tree locally when Anaconda or Miniconda is available.
Author: Assista~
Lumina Decision Systems
Date: Thu, Jul 16, 2026 4:13 PM
NodeLocation: 160,512,1
NodeSize: 80,24
NodeInfo: 1,,,,,,,0,,,,,,0
DiagState: 2,1055,99,1560,768,17,10
WindState: 2,756,209,720,350

Module Iris_data
Title: Iris data
Description: The Iris measurements, species labels, and their Observation and Measurement dimensions.
NodeLocation: 208,544,2
NodeSize: 96,32
NodeInfo: 1,,,,,,,0,,,,,,0
DiagState: 2,0,1,1560,656,17,10
WindState: 2,98,82,720,350

Index Observation
Title: Observation
Description: The 150 Iris flower observations supplied with scikit-learn.
Definition: 1..150
NodeLocation: 96,64,2
NodeSize: 80,24
NodeInfo: 1,,,,,,,0,,,,,,0

Index Feature
Title: Measurement
Description: The four flower measurements used as predictor features.
Definition: ['Sepal length (cm)','Sepal width (cm)','Petal length (cm)','Petal width (cm)']
NodeLocation: 288,64,2
NodeSize: 80,24
NodeInfo: 1,,,,,,,0,,,,,,0

Variable Iris_measurements
Title: Iris measurements
Units: cm
Description: The 150 by 4 numerical measurements loaded from scikit-learn's built-in Iris data set.
Definition: PyExplode(~
    PyEval('load_iris().data'),~
    Observation,~
    Feature~
)
NodeLocation: 512,136,2
NodeSize: 96,24
NodeInfo: 1,,,,,,,0,,,,,,0
WindState: 2,98,82,720,350
ValueState: 2,662,56,752,303,,MIDM
Aliases: FormNode Iris_measurements1148013867

Variable Iris_species_code
Title: Actual species code
Description: Integer target class for each Iris observation: 0=setosa, 1=versicolor, 2=virginica.
Definition: PyExplode(~
    PyEval('load_iris().target'),~
    Observation~
)
NodeLocation: 512,208,2
NodeSize: 96,24
NodeInfo: 1,,,,,,,0,,,,,,0

Variable Iris_species
Title: Actual species
Description: Species name for each Iris observation, loaded from scikit-learn's built-in Iris data set.
Definition: PyExplode(~
    PyEval('load_iris().target_names[load_iris().target]'),~
    Observation~
)
NodeLocation: 512,280,2
NodeSize: 96,24
NodeInfo: 1,,,,,,,0,,,,,,0
ValueState: 2,100,106,416,303,,MIDM
Aliases: FormNode Iris_species611142955

Close Iris_data

Module Tree_training
Title: Train decision tree
Description: Training controls and Python components used to fit the Iris CART decision tree.
NodeLocation: 448,544,2
NodeSize: 96,32
NodeInfo: 1,,,,,,,0,,,,,,0
DiagState: 2,0,1,1560,656,17,10
WindState: 2,98,82,720,350

Decision Tree_max_depth
Title: Maximum tree depth
Description: Maximum depth of the fitted CART decision tree. Increase it to allow a more complex tree.
Definition: 3
NodeLocation: 96,176,2
NodeSize: 96,24
NodeInfo: 1,,,,,,,0,,,,,,0
Aliases: FormNode Tree_max_depth1148013867

Decision Random_seed
Title: Random seed
Description: Seed supplied to scikit-learn's tree estimator, for reproducible fitting.
Definition: 7
NodeLocation: 96,248,2
NodeSize: 96,24
NodeInfo: 1,,,,,,,0,,,,,,0
Aliases: FormNode Random_seed1148013867

{!70000|Callable Fit_tree}
{!70000|Title: Fit CART tree (Python)}
{!70000|Description: Python Callable that fits and returns a scikit-learn CART DecisionTreeClassifier.}
{!70000|Definition: PyExec('import numpy as np~
from sklearn.tree import DecisionTreeClassifier~
~
def fit_tree(X, y, max_depth, seed, test_positions):~
    test_pos = np.asarray(test_positions, dtype=int) - 1~
    train_mask = np.ones(len(y), dtype=bool)~
    train_mask[test_pos] = False~
    tree = DecisionTreeClassifier(~
        criterion="gini",~
        max_depth=int(max_depth),~
        random_state=int(seed)~
    )~
    return tree.fit(X[train_mask], y[train_mask])', flags:0)}
{!70000|NodeLocation: 560,280,2}
{!70000|NodeSize: 96,24}
{!70000|NodeInfo: 1,,,,,,,0,,,,,,0}

Variable Fitted_tree
Title: Fitted Python decision tree
Description: Opaque Python scikit-learn DecisionTreeClassifier obtained from Tree_as_struct's python_model member. Retained as an optional technical view; model calculations use the native Decision_tree instance.
Definition: Tree_as_struct -> python_model
NodeLocation: 1064,176,2
NodeSize: 104,24
NodeInfo: 1,,,,,,,0,,,,,,0
ValueState: 2,678,121,647,436,,MIDM
Aliases: FormNode Fitted_tree611142955

{!70000|Callable Export_tree_JSON}
{!70000|Title: Export tree as JSON}
{!70000|Description: Python Callable that converts a fitted scikit-learn decision tree into nested JSON suitable for ParseJSON and inspection as native Analytica Structs.}
{!70000|Definition: PyExec('import json~
import numpy as np~
~
def export_tree_json(model, feature_names):~
    tree = model.tree_~
~
    def visit(node_id):~
        counts = [int(x) for x in tree.value[node_id][0]]~
        result = {~
            "node": int(node_id),~
            "samples": int(tree.n_node_samples[node_id]),~
            "class_counts": counts,~
            "prediction": int(np.argmax(counts))~
        }~
        if tree.children_left[node_id] == tree.children_right[node_id]:~
            result["kind"] = "leaf"~
        else:~
            feature_id = int(tree.feature[node_id])~
            result.update({~
                "kind": "split",~
                "feature": str(feature_names[feature_id]),~
                "threshold": float(tree.threshold[node_id]),~
                "left": visit(int(tree.children_left[node_id])),~
                "right": visit(int(tree.children_right[node_id]))~
            })~
        return result~
~
    return json.dumps({~
        "classes": [int(x) for x in model.classes_],~
        "root": visit(0)~
    })', flags:0)}
{!70000|NodeLocation: 96,80,2}
{!70000|NodeSize: 80,24}
{!70000|NodeInfo: 1,,,,,,,0,,,,,,0}

Objective Tree_as_struct
Title: Fitted decision tree
Description: Native fitted Decision_tree. It is trained only on the 120-observation training set; its Fit and Predict member functions own the model operations, and python_model holds the opaque scikit-learn estimator.
Definition: Tree_template -> Fit(~
    PyArray(Iris_measurements, Observation, Feature),~
    PyArray(Iris_species_code, Observation, dtype:'int64'),~
    Tree_max_depth,~
    Random_seed,~
    PyList(Test_observation),~
    'Iris CART tree'~
)
NodeLocation: 824,176,2
NodeSize: 96,24
NodeInfo: 1,,,,,,,0,,,,,,0
WindState: 2,753,287,720,350
ValueState: 2,1042,111,400,600,0,MIDM
Aliases: FormNode Tree_as_struct2037599531

{!70000|Struct Decision_tree( python_model : atom ; feature_names : atom ; display_name : text atom )}
{!70000|Title: Decision tree}
{!70000|Description: Native Analytica decision-tree datatype. Its instances hold the opaque fitted Python estimator plus native top-level tree metadata: classes, root, and a configurable display name.}
{!70000|Definition: If IsNull(python_model) Then Self Else (~
    Local tree_data := ParseJSON(~
        Export_tree_JSON(python_model, feature_names)~
    );~
    Self -> classes := tree_data -> classes;~
    Self -> root := tree_data -> root;~
    Self~
)}
{!70000|NodeLocation: 320,80,2}
{!70000|NodeSize: 96,24}
{!70000|NodeInfo: 1,,,,,,,0,,,,,,0}
{!70000|DiagState: 2,0,1,1560,656,17,10}
{!70000|WindState: 2,98,82,720,350}

Function DisplayName( Self : Decision_tree atom )
Title: Display name
Description: Returns the display name used when a Decision_tree instance appears in a result table.
Definition: Self->display_name
NodeLocation: 160,64,1
NodeSize: 64,24
NodeInfo: 1,,,,,,,0,,,,,,0
WindState: 2,98,82,720,350

Function Fit( Self : Decision_tree atom ; X : atom ; y : atom ; max_depth : atom ; seed : atom ; test_positions : atom ; display_name : text atom )
Title: Fit tree
Description: Fits a CART classifier and returns a new Decision_tree instance. The fitted Python estimator is stored in its python_model member.
Definition: Decision_tree(~
    Fit_tree(X, y, max_depth, seed, test_positions),~
    Self -> feature_names,~
    display_name~
)
NodeLocation: 352,64,1
NodeSize: 64,24
NodeInfo: 1,,,,,,,0,,,,,,0

Function Predict( Self : Decision_tree atom ; X : atom )
Title: Predict species code
Description: Uses this fitted tree's python_model member to predict species codes for a NumPy measurement matrix.
Definition: Self -> python_model -> predict(X)
NodeLocation: 544,64,1
NodeSize: 64,24
NodeInfo: 1,,,,,,,0,,,,,,0

Function Predict_holdout( Self : Decision_tree atom ; X : atom ; test_positions : atom )
Title: Predict holdout
Description: Predicts the selected 1-based observation positions from a full NumPy measurement matrix, using this instance's python_model member.
Definition: Predict_tree(~
    Self -> python_model,~
    X,~
    test_positions~
)
NodeLocation: 744,64,1
NodeSize: 72,24
NodeInfo: 1,,,,,,,0,,,,,,0

Function Score_holdout( Self : Decision_tree atom ; X : atom ; y : atom ; test_positions : atom )
Title: Score holdout accuracy
Description: Computes classification accuracy for selected 1-based observation positions from full NumPy measurement and target arrays, using this instance's python_model member.
Definition: Score_tree(~
    Self -> python_model,~
    X,~
    y,~
    test_positions~
)
NodeLocation: 928,64,1
NodeSize: 64,24
NodeInfo: 1,,,,,,,0,,,,,,0

Close Decision_tree

Variable Tree_template
Title: Unfitted decision tree template
Description: Internal untrained Decision_tree instance used to invoke its Fit member function.
Definition: Decision_tree(~
    Null,~
    PyList(Feature, Feature),~
    'Unfitted Iris CART tree'~
)
NodeLocation: 560,80,2
NodeSize: 112,24
NodeInfo: 1,,,,,,,0,,,,,,0

Close Tree_training

Module New_flower_classification
Title: Classify new flower
Description: Classify a new Iris measurement using the fitted decision tree.
NodeLocation: 448,672,2
NodeSize: 104,32
NodeInfo: 1,,,,,,,0,,,,,,0
DiagState: 2,0,1,1560,656,17,10

Decision New_measurements
Title: New measurements
Units: cm
Description: Enter the four flower measurements to classify, in the order shown by the Measurement index.
Definition: Table(Feature)(5.1,3.5,1.4,0.2)
NodeLocation: 96,160,2
NodeSize: 96,24
NodeInfo: 1,,,,,,,0,,,,,,0
DefnState: 2,612,618,416,303,0,DFNM
Aliases: FormNode New_measurements2084261163

Variable New_species_code
Title: Predicted species code
Description: Species code predicted from New_measurements by Tree_as_struct's Predict member function.
Definition: Local x := PyArray(New_measurements, Feature);~
Local prediction := Tree_as_struct -> Predict(~
    PyEval('np.asarray(x).reshape(1, -1)')~
);~
PyEval('int(prediction[0])')
NodeLocation: 352,160,2
NodeSize: 96,24
NodeInfo: 1,,,,,,,0,,,,,,0

Objective New_species
Title: Predicted species
Description: Species name corresponding to the prediction for New_measurements.
Definition: Local code := New_species_code;~
PyEval('load_iris().target_names[int(code)]')
NodeLocation: 608,160,2
NodeSize: 80,24
NodeInfo: 1,,,,,,,0,,,,,,0
Aliases: FormNode New_species2084261163

Close New_flower_classification

Module Holdout_validation
Title: Holdout validation
Description: Stratified holdout observations, predictions, and classification-accuracy assessment.
NodeLocation: 208,672,2
NodeSize: 96,32
NodeInfo: 1,,,,,,,0,,,,,,0
DiagState: 2,0,1,1560,656,17,10

Variable Test_observation
Title: Test observations
Description: A reproducible, stratified 20% holdout: 10 observations from each Iris species. It changes with Random_seed.
Definition: Local seed := Random_seed;~
PyExplode(~
    PyEval(~
        'np.sort(np.concatenate([np.random.default_rng(int(seed) + k).choice(np.arange(1 + 50*k, 51 + 50*k), size=10, replace=False) for k in range(3)]))'~
    )~
)
NodeLocation: 104,120,2
NodeSize: 88,24
NodeInfo: 1,,,,,,,0,,,,,,0

Variable Test_actual_species_code
Title: Actual test species code
Description: Actual species codes for the 30-observation holdout, in the same row order as Test observations.
Definition: Local test_positions := PyList(Test_observation);~
PyExplode(~
    PyEval(~
        'load_iris().target[np.asarray(test_positions, dtype=int) - 1]'~
    )~
)
NodeLocation: 368,80,2
NodeSize: 112,24
NodeInfo: 1,,,,,,,0,,,,,,0
ValueState: 2,580,586,416,303,,MIDM
Aliases: FormNode Test_actual_species_code2084261163

Variable Test_predicted_species_code
Title: Predicted test species code
Description: Species codes predicted for the holdout observations by Tree_as_struct's Predict_holdout member function.
Definition: PyExplode(~
    Tree_as_struct -> Predict_holdout(~
        PyArray(Iris_measurements, Observation, Feature),~
        PyList(Test_observation)~
    )~
)
NodeLocation: 368,160,2
NodeSize: 120,24
NodeInfo: 1,,,,,,,0,,,,,,0
ValueState: 2,596,602,416,303,,MIDM
Aliases: FormNode Test_predicted_species_code2084261163

Objective Test_accuracy
Title: Test classification accuracy
Units: %
Description: Percentage of the 30 stratified holdout observations classified correctly. It is evaluated by Tree_as_struct's Score_holdout member function.
Definition: 100 * Tree_as_struct -> Score_holdout(~
    PyArray(Iris_measurements, Observation, Feature),~
    PyArray(Iris_species_code, Observation, dtype:'int64'),~
    PyList(Test_observation)~
)
NodeLocation: 656,120,2
NodeSize: 104,24
NodeInfo: 1,,,,,,,0,,,,,,0
ValueState: 2,628,634,416,303,,MIDM
Aliases: FormNode Test_accuracy2084261163

{!70000|Callable Predict_tree}
{!70000|Title: Predict tree holdout (Python)}
{!70000|Description: Internal Python helper used by the Decision_tree Predict_holdout member function.}
{!70000|Definition: PyExec('import numpy as np~
~
def predict_tree(model, X, positions):~
    return model.predict(X[np.asarray(positions, dtype=int) - 1])', flags:0)}
{!70000|NodeLocation: 104,264,2}
{!70000|NodeSize: 96,24}
{!70000|NodeInfo: 1,,,,,,,0,,,,,,0}
{!70000|ValueState: 2,644,650,416,303,,MIDM}

{!70000|Callable Score_tree}
{!70000|Title: Score tree holdout (Python)}
{!70000|Description: Internal Python helper used by the Decision_tree Score_holdout member function.}
{!70000|Definition: PyExec('import numpy as np~
~
def score_tree(model, X, y, positions):~
    pos = np.asarray(positions, dtype=int) - 1~
    return float(np.mean(model.predict(X[pos]) == y[pos]))', flags:0)}
{!70000|NodeLocation: 368,264,2}
{!70000|NodeSize: 96,24}
{!70000|NodeInfo: 1,,,,,,,0,,,,,,0}

Close Holdout_validation

Close Iris_Decision_Tree

FormNode Tree_max_depth1148013867
Definition: 0
NodeLocation: 620,104,2
NodeSize: 140,24
NodeInfo: 1,,,,,,,96,,,,,,0
Original: Tree_max_depth

FormNode Random_seed1148013867
Definition: 0
NodeLocation: 620,168,2
NodeSize: 140,24
NodeInfo: 1,,,,,,,96,,,,,,0
Original: Random_seed

FormNode Iris_measurements1148013867
Definition: 1
NodeLocation: 248,284,2
NodeSize: 144,20
NodeInfo: 1,,,,,,,96,,,,,,0
Original: Iris_measurements

FormNode Iris_species611142955
Definition: 1
NodeLocation: 248,348,2
NodeSize: 144,20
NodeInfo: 1,,,,,,,96,,,,,,0
Original: Iris_species

FormNode Fitted_tree611142955
Definition: 1
NodeLocation: 620,292,2
NodeSize: 160,24
NodeInfo: 1,,,,,,,72,,,,,,0
Original: Fitted_tree

Text Setup_panel
Title: 1. Python setup: Analytica_Iris_Tree
Description: One-time setup: Click the button below to create the Analytica_Iris_Tree environment. It requires Anaconda or Miniconda; if neither is found, it displays installation instructions.
NodeLocation: 236,136,1
NodeSize: 172,88
NodeInfo: 1,,,,1,1,,0,,,,,,0,,,,8
NodeColor: 0xfff4f7fa

Text Controls_panel
Title: 2. Training controls
NodeLocation: 620,136,1
NodeSize: 180,88
NodeInfo: 1,,,,1,1,,0,,,,,,0,,,,8
NodeColor: 0xfff4f7fa

Text Data_panel
Title: 3. View Iris data
NodeLocation: 236,348,1
NodeSize: 172,108
NodeInfo: 1,,,,1,1,,0,,,,,,0,,,,8
NodeColor: 0xfff4f7fa

Text Model_panel
Title: 4. Train
NodeLocation: 620,348,1
NodeSize: 180,108
NodeInfo: 1,,,,1,1,,0,,,,,,0,,,,8
NodeColor: 0xfff4f7fa

Button Create_Iris_Python_environment
Title: Create Python environment
Description: Creates the Analytica_Iris_Tree Conda environment with Python, NumPy, and scikit-learn when needed.
NodeLocation: 236,186,2
NodeSize: 116,18
NodeInfo: 1,,,,,,1,0,,,,,,0
WindState: 2,98,82,992,602
NodeFont: ,10
OnClick: Local envName := 'Analytica_Iris_Tree';~
Local pythonEnvs := GetProcessInfo('Python Environments');~
Local envExists := Sum(~
    FindInText(envName, pythonEnvs, caseInsensitive: true) > 0~
) > 0;~
Local installations := GetProcessInfo('Python Installations');~
Local isConda := FindInText(~
    'conda',~
    installations,~
    caseInsensitive: true~
) > 0;~
If envExists Then~
    MsgBox(~
        'The ' & envName & ' environment is already available.',~
        buttons: 0,~
        title: 'Python environment'~
    )~
Else If Sum(isConda) = 0 Then~
    MsgBox(~
        'Anaconda or Miniconda was not found. Install one of them, then click this button again.',~
        buttons: 0,~
        title: 'Python environment unavailable'~
    )~
Else~
    RunConsoleProcess(~
        If isConda Then~
            installations & '\Scripts\conda.exe'~
        Else~
            Null,~
        'conda.exe create -n Analytica_Iris_Tree -y python=3.12 numpy scikit-learn',~
        '',~
        true,~
        '',~
        0,~
        2~
    )

FormNode Tree_as_struct2037599531
Definition: 1
NodeLocation: 620,348,2
NodeSize: 160,24
NodeInfo: 1,,,,,,,108,,,,,,0
Original: Tree_as_struct

FormNode New_measurements2084261163
Definition: 0
NodeLocation: 1040,108,2
NodeSize: 184,28
NodeInfo: 1,,,,,,,128,,,,,,0
Original: New_measurements

FormNode New_species2084261163
Definition: 1
NodeLocation: 1040,164,2
NodeSize: 184,28
NodeInfo: 1,,,,,,,128,,,,,,0
Original: New_species

FormNode Test_accuracy2084261163
Definition: 1
NodeLocation: 1040,300,2
NodeSize: 184,28
NodeInfo: 1,,,,,,,128,,,,,,0
Original: Test_accuracy

FormNode Test_predicted_species_code2084261163
Definition: 1
NodeLocation: 1040,412,2
NodeSize: 184,28
NodeInfo: 1,,,,,,,128,,,,,,0
Original: Test_predicted_species_code

FormNode Test_actual_species_code2084261163
Definition: 1
NodeLocation: 1040,356,2
NodeSize: 184,28
NodeInfo: 1,,,,,,,128,,,,,,0
Original: Test_actual_species_code

Text Classify_panel
Title: 5. Classify a new Iris flower
NodeLocation: 1040,132,1
NodeSize: 208,84
NodeInfo: 1,,,,1,1,,0,,,,,,0,,,,8
NodeColor: 0xffeaf2fa
{!50400|NodeBorderColor: 0xff6b8cae}

Text Test_panel
Title: 6. Holdout test performance
NodeLocation: 1040,348,1
NodeSize: 208,108
NodeInfo: 1,,,,1,0,,0,,,,,,0,,,,8
NodeColor: 0xfff3f7ec
{!50400|NodeBorderColor: 0xff7d9b5d}

Close Python_decision_tree_example