Optiscope Classification Analysis¶

This notebook demonstrates the solution classification methods in optiscope, including:

1. Classification by Best/Worst Objective
2. Physical Programming (PRUF) Classification

In [1]:

import numpy as np
import pandas as pd

from optiscope import OptimizationResult, ProblemMetadata
from optiscope.analysis.classification import classify_result
from optiscope.core import Objective, OptimizationDirection
from optiscope.plotting import plot_parallel_coordinates, plot_scatter_matrix

import numpy as np import pandas as pd from optiscope import OptimizationResult, ProblemMetadata from optiscope.analysis.classification import classify_result from optiscope.core import Objective, OptimizationDirection from optiscope.plotting import plot_parallel_coordinates, plot_scatter_matrix

Data Generation¶

We generate a synthetic multi-objective optimization problem.

In [2]:

def generate_test_problem(n_points: int = 200, seed: int = 42):
    """Generate synthetic multi-objective optimization problem."""
    np.random.seed(seed)

    # Design variables
    x1 = np.random.uniform(0, 10, n_points)
    x2 = np.random.uniform(0, 10, n_points)
    x3 = np.random.uniform(0, 10, n_points)

    # Objectives
    f1 = x1
    f2 = (x2 + 1) / (x1 + 0.1)
    f3 = x3 * x1

    # Create result
    result = OptimizationResult(
        problem_metadata=ProblemMetadata(
            name="Classification Test",
            solver="Random",
            n_design_variables=3,
            n_objectives=3,
        ),
        design_variables=pd.DataFrame({"x1": x1, "x2": x2, "x3": x3}),
        objectives=pd.DataFrame({"f1": f1, "f2": f2, "f3": f3}),
    )

    result.add_variable_metadata(Objective(name="f1", direction=OptimizationDirection.MINIMIZE))
    result.add_variable_metadata(Objective(name="f2", direction=OptimizationDirection.MINIMIZE))
    result.add_variable_metadata(Objective(name="f3", direction=OptimizationDirection.MINIMIZE))

    return result


print("Generating test problem...")
result = generate_test_problem(n_points=200)
print(f"Created problem: {result.problem_metadata.name}")

def generate_test_problem(n_points: int = 200, seed: int = 42): """Generate synthetic multi-objective optimization problem.""" np.random.seed(seed) # Design variables x1 = np.random.uniform(0, 10, n_points) x2 = np.random.uniform(0, 10, n_points) x3 = np.random.uniform(0, 10, n_points) # Objectives f1 = x1 f2 = (x2 + 1) / (x1 + 0.1) f3 = x3 * x1 # Create result result = OptimizationResult( problem_metadata=ProblemMetadata( name="Classification Test", solver="Random", n_design_variables=3, n_objectives=3, ), design_variables=pd.DataFrame({"x1": x1, "x2": x2, "x3": x3}), objectives=pd.DataFrame({"f1": f1, "f2": f2, "f3": f3}), ) result.add_variable_metadata(Objective(name="f1", direction=OptimizationDirection.MINIMIZE)) result.add_variable_metadata(Objective(name="f2", direction=OptimizationDirection.MINIMIZE)) result.add_variable_metadata(Objective(name="f3", direction=OptimizationDirection.MINIMIZE)) return result print("Generating test problem...") result = generate_test_problem(n_points=200) print(f"Created problem: {result.problem_metadata.name}")

Generating test problem...
Created problem: Classification Test

1. Classification by Objective Performance¶

Identify which objective is the best or worst performing for each solution.

In [3]:

# Classify by best objective
best_obj_class = classify_result(
    result, method="best_obj", add_to_result=True, new_column_name="best_objective"
)
print("Classification by Best Objective (first 10):")
print(best_obj_class[:10])

# Classify by worst objective
worst_obj_class = classify_result(
    result, method="worst_obj", add_to_result=True, new_column_name="worst_objective"
)
print("\nClassification by Worst Objective (first 10):")
print(worst_obj_class[:10])

# Classify by best objective best_obj_class = classify_result( result, method="best_obj", add_to_result=True, new_column_name="best_objective" ) print("Classification by Best Objective (first 10):") print(best_obj_class[:10]) # Classify by worst objective worst_obj_class = classify_result( result, method="worst_obj", add_to_result=True, new_column_name="worst_objective" ) print("\nClassification by Worst Objective (first 10):") print(worst_obj_class[:10])

Classification by Best Objective (first 10):
['f2' 'f2' 'f2' 'f2' 'f3' 'f2' 'f3' 'f3' 'f2' 'f2']

Classification by Worst Objective (first 10):
['f1' 'f1' 'f1' 'f1' 'f1' 'f1' 'f1' 'f1' 'f1' 'f1']

2. Physical Programming (PRUF) Classification¶

Classify solutions into desirability categories (Highly Desirable to Highly Undesirable).

In [4]:

# PRUF Worst (Pessimistic)
pruf_worst = classify_result(
    result, method="pruf_worst", add_to_result=True, new_column_name="pruf_class"
)
print("PRUF Classification (Pessimistic) - first 10:")
print(pruf_worst[:10])

# Count categories
unique, counts = np.unique(pruf_worst, return_counts=True)
print("\nCategory Counts:")
for cat, count in zip(unique, counts):
    print(f"  {cat}: {count}")

# PRUF Worst (Pessimistic) pruf_worst = classify_result( result, method="pruf_worst", add_to_result=True, new_column_name="pruf_class" ) print("PRUF Classification (Pessimistic) - first 10:") print(pruf_worst[:10]) # Count categories unique, counts = np.unique(pruf_worst, return_counts=True) print("\nCategory Counts:") for cat, count in zip(unique, counts): print(f" {cat}: {count}")

PRUF Classification (Pessimistic) - first 10:
['Desirable' 'Highly Undesirable' 'Undesirable' 'Undesirable'
 'Highly Desirable' 'Highly Desirable' 'Highly Desirable'
 'Highly Undesirable' 'Undesirable' 'Undesirable']

Category Counts:
  Desirable: 50
  Highly Desirable: 38
  Highly Undesirable: 43
  Tolerable: 31
  Undesirable: 38

3. Visualization¶

Visualize the results colored by their classification.

In [5]:

# Parallel Coordinates colored by PRUF class
fig = plot_parallel_coordinates(
    result,
    include_objectives=True,
    color_by="pruf_class",
    title="Parallel Coordinates (Colored by PRUF Class)",
)
fig.show()

# Parallel Coordinates colored by PRUF class fig = plot_parallel_coordinates( result, include_objectives=True, color_by="pruf_class", title="Parallel Coordinates (Colored by PRUF Class)", ) fig.show()

In [6]:

# Scatter Matrix colored by Best Objective
fig2 = plot_scatter_matrix(
    result,
    include_objectives=True,
    color_by="best_objective",
    title="Scatter Matrix (Colored by Best Objective)",
)
fig2.show()

# Scatter Matrix colored by Best Objective fig2 = plot_scatter_matrix( result, include_objectives=True, color_by="best_objective", title="Scatter Matrix (Colored by Best Objective)", ) fig2.show()