A Precision-Recall curve plots precision (positive predictive value) against recall (sensitivity) at various classification thresholds. This visualization is essential for evaluating binary classifiers on imbalanced datasets where accuracy alone is misleading. The area under the curve (Average Precision) summarizes classifier performance, with higher values indicating better performance.
""" anyplot.ai
precision-recall: Precision-Recall Curve
Library: letsplot 4.9.0 | Python 3.13.13
Quality: 91/100 | Updated: 2026-05-10
"""
import os
import numpy as np
import pandas as pd
from lets_plot import *
from lets_plot import ggsave
LetsPlot.setup_html()
# Theme tokens
THEME = os.getenv("ANYPLOT_THEME", "light")
PAGE_BG = "#FAF8F1" if THEME == "light" else "#1A1A17"
ELEVATED_BG = "#FFFDF6" if THEME == "light" else "#242420"
INK = "#1A1A17" if THEME == "light" else "#F0EFE8"
INK_SOFT = "#4A4A44" if THEME == "light" else "#B8B7B0"
INK_MUTED = "#6B6A63" if THEME == "light" else "#A8A79F"
RULE = "rgba(26,26,23,0.10)" if THEME == "light" else "rgba(240,239,232,0.10)"
# Okabe-Ito palette (first series always #009E73)
BRAND = "#009E73"
SECONDARY = "#C475FD"
# Generate synthetic classification data with imbalanced classes
np.random.seed(42)
n_samples = 500
# Simulate imbalanced dataset: 20% positive, 80% negative
positive_ratio = 0.2
n_positive = int(n_samples * positive_ratio)
n_negative = n_samples - n_positive
# True labels
y_true = np.concatenate([np.ones(n_positive), np.zeros(n_negative)])
# Simulate classifier scores - good classifier gives higher scores to positives
positive_scores = np.random.beta(5, 2, n_positive) # Skewed higher
negative_scores = np.random.beta(2, 5, n_negative) # Skewed lower
y_scores = np.concatenate([positive_scores, negative_scores])
# Calculate precision-recall curve
# Sort by scores descending
desc_score_indices = np.argsort(y_scores)[::-1]
y_scores_sorted = y_scores[desc_score_indices]
y_true_sorted = y_true[desc_score_indices]
# Get unique thresholds
distinct_value_indices = np.where(np.diff(y_scores_sorted))[0]
threshold_idxs = np.concatenate([[0], distinct_value_indices + 1])
# Calculate TP, FP cumulative sums
tps = np.cumsum(y_true_sorted)
fps = np.cumsum(1 - y_true_sorted)
# Calculate precision and recall at each threshold
precision_vals = tps / (tps + fps)
recall_vals = tps / tps[-1]
# Use thresholds at distinct values (add starting point: recall=0, precision=1)
precision = np.concatenate([[1], precision_vals[threshold_idxs]])
recall = np.concatenate([[0], recall_vals[threshold_idxs]])
# Calculate Average Precision (area under curve)
recall_diff = np.diff(recall)
average_precision = np.sum(precision[1:] * recall_diff)
# Create step-wise data for proper step visualization
recall_step = []
precision_step = []
for i in range(len(recall) - 1):
recall_step.extend([recall[i], recall[i + 1]])
precision_step.extend([precision[i], precision[i]])
recall_step.append(recall[-1])
precision_step.append(precision[-1])
df_curve = pd.DataFrame(
{"recall": recall_step, "precision": precision_step, "model": f"Classifier (AP = {average_precision:.3f})"}
)
# Create iso-F1 curves data
f1_scores = [0.2, 0.4, 0.6, 0.8]
f1_data = []
for f1_score in f1_scores:
x = np.linspace(f1_score / 2 + 0.01, 1, 100)
y = f1_score * x / (2 * x - f1_score)
mask = (y >= 0) & (y <= 1)
for xi, yi in zip(x[mask], y[mask], strict=False):
f1_data.append({"recall": xi, "precision": yi, "f1": f"F1={f1_score:.1f}"})
df_f1 = pd.DataFrame(f1_data)
# Labels for iso-F1 curves (at the end of each curve)
f1_labels = []
for f1_score in f1_scores:
x = np.linspace(f1_score / 2 + 0.01, 1, 100)
y = f1_score * x / (2 * x - f1_score)
mask = (y >= 0) & (y <= 1)
if np.any(mask):
f1_labels.append({"recall": x[mask][-1] + 0.01, "precision": y[mask][-1], "label": f"F1={f1_score:.1f}"})
df_f1_labels = pd.DataFrame(f1_labels)
# Build the plot
plot = (
ggplot()
# Iso-F1 curves (background)
+ geom_line(
data=df_f1,
mapping=aes(x="recall", y="precision", group="f1"),
color=INK_SOFT,
alpha=0.3,
size=1,
linetype="dotted",
)
# F1 labels
+ geom_text(
data=df_f1_labels, mapping=aes(x="recall", y="precision", label="label"), color=INK_MUTED, size=10, alpha=0.6
)
# Main precision-recall curve
+ geom_area(data=df_curve, mapping=aes(x="recall", y="precision"), fill=BRAND, alpha=0.2)
+ geom_line(data=df_curve, mapping=aes(x="recall", y="precision", color="model"), size=1.5)
# Baseline: random classifier
+ geom_hline(yintercept=positive_ratio, color=SECONDARY, size=1.2, linetype="dashed")
+ geom_text(
data=pd.DataFrame(
{"x": [0.85], "y": [positive_ratio + 0.03], "label": [f"Random Baseline ({positive_ratio:.0%})"]}
),
mapping=aes(x="x", y="y", label="label"),
color=INK_SOFT,
size=12,
)
# Labels and title
+ labs(x="Recall (Sensitivity)", y="Precision (PPV)", title="precision-recall · letsplot · anyplot.ai", color="")
+ scale_x_continuous(limits=[0, 1.0])
+ scale_y_continuous(limits=[0, 1.05])
+ scale_color_manual(values=[BRAND])
# Size for 4800x2700 at scale=3
+ ggsize(1600, 900)
# Theme
+ theme_minimal()
+ theme(
plot_background=element_rect(fill=PAGE_BG, color=PAGE_BG),
panel_background=element_rect(fill=PAGE_BG),
panel_grid_major_y=element_line(color=RULE, size=0.3),
axis_title=element_text(size=20, color=INK),
axis_text=element_text(size=16, color=INK_SOFT),
plot_title=element_text(size=24, color=INK),
legend_text=element_text(size=16, color=INK_SOFT),
legend_title=element_text(color=INK),
legend_background=element_rect(fill=ELEVATED_BG, color=INK_SOFT),
legend_position="top",
)
)
# Save as PNG (scale 3x = 4800 x 2700 px) and HTML
ggsave(plot, f"plot-{THEME}.png", scale=3, path=".")
ggsave(plot, f"plot-{THEME}.html", path=".")
Part of Precision-Recall Curve on anyplot.ai.