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: ggplot2 3.5.1 | R 4.4.1
#' Quality: 92/100 | Created: 2026-09-05
library(ggplot2)
library(dplyr)
library(tidyr)
library(scales)
library(ragg)
set.seed(42)
# --- Theme tokens -----------------------------------------------------------
THEME <- Sys.getenv("ANYPLOT_THEME", "light")
PAGE_BG <- if (THEME == "light") "#FAF8F1" else "#1A1A17"
INK <- if (THEME == "light") "#1A1A17" else "#F0EFE8"
INK_SOFT <- if (THEME == "light") "#4A4A44" else "#B8B7B0"
INK_MUTED <- if (THEME == "light") "#6B6A63" else "#A8A79F"
IMPRINT_PALETTE <- c("#009E73", "#C475FD", "#4467A3", "#BD8233",
"#AE3030", "#2ABCCD", "#954477", "#99B314")
# --- Data ---------------------------------------------------------------
# Fraud detection: 5% of transactions are fraudulent (heavily imbalanced),
# comparing a weak baseline classifier against a tuned one.
n <- 3000
positive_rate <- 0.05
y_true <- rbinom(n, 1, positive_rate)
baseline_scores <- plogis(ifelse(y_true == 1, rnorm(n, 1.0, 1.0), rnorm(n, -1.0, 1.0)))
tuned_scores <- plogis(ifelse(y_true == 1, rnorm(n, 2.2, 1.0), rnorm(n, -2.2, 1.0)))
scores_df <- bind_rows(
tibble(model = "Tuned model", y_true = y_true, y_score = tuned_scores),
tibble(model = "Baseline model", y_true = y_true, y_score = baseline_scores)
)
# Precision/recall at every threshold, walking scores from high to low.
pr_curve <- scores_df %>%
arrange(model, desc(y_score)) %>%
group_by(model) %>%
mutate(
tp = cumsum(y_true),
fp = cumsum(1 - y_true),
precision = tp / (tp + fp),
recall = tp / sum(y_true),
recall_prev = lag(recall, default = 0)
)
# Average precision: AP = sum_k (R_k - R_{k-1}) * P_k (sklearn convention).
ap_scores <- pr_curve %>%
summarise(ap = sum((recall - recall_prev) * precision), .groups = "drop")
pr_curve <- pr_curve %>%
ungroup() %>%
left_join(ap_scores, by = "model") %>%
mutate(model_label = sprintf("%s (AP = %.2f)", model, ap))
model_labels <- pr_curve %>%
distinct(model, model_label) %>%
arrange(match(model, c("Tuned model", "Baseline model")))
baseline_df <- tibble(recall = c(0, 1), precision = positive_rate)
# Iso-F1 reference curves: F1 = 2PR / (P+R) solved for P at fixed F1 levels.
f1_levels <- c(0.2, 0.4, 0.6, 0.8)
iso_f1 <- expand_grid(f1 = f1_levels, recall = seq(0.02, 1, length.out = 300)) %>%
mutate(precision = f1 * recall / (2 * recall - f1)) %>%
filter(precision > 0, precision <= 1)
iso_f1_labels <- iso_f1 %>%
group_by(f1) %>%
slice_max(recall, n = 1) %>%
ungroup() %>%
mutate(label = sprintf("F1=%.1f", f1))
# --- Plot ---------------------------------------------------------------
plot_title <- "Fraud Detection Model Comparison · precision-recall · r · ggplot2 · anyplot.ai"
title_fontsize <- max(8, round(12 * min(1, 67 / nchar(plot_title))))
p <- ggplot() +
geom_line(
data = iso_f1, aes(x = recall, y = precision, group = f1),
color = INK_MUTED, linewidth = 0.35, linetype = "dotted", alpha = 0.6
) +
geom_text(
data = iso_f1_labels, aes(x = recall, y = precision, label = label),
color = INK_MUTED, size = 2.6, hjust = 0, nudge_x = 0.015
) +
geom_line(
data = baseline_df, aes(x = recall, y = precision, linetype = "Baseline (random)"),
color = INK_MUTED, linewidth = 0.7
) +
geom_step(
data = pr_curve, aes(x = recall, y = precision, color = model_label),
linewidth = 1.1
) +
scale_color_manual(values = IMPRINT_PALETTE[1:2], breaks = model_labels$model_label, name = NULL) +
scale_linetype_manual(values = c("Baseline (random)" = "dashed"), name = NULL) +
scale_x_continuous(expand = expansion(mult = c(0.01, 0.1))) +
scale_y_continuous(expand = expansion(mult = c(0.02, 0.05))) +
coord_cartesian(xlim = c(0, 1), ylim = c(0, 1), clip = "off") +
labs(x = "Recall", y = "Precision", title = plot_title) +
theme_minimal(base_size = 8) +
theme(
plot.background = element_rect(fill = PAGE_BG, color = PAGE_BG),
panel.background = element_rect(fill = PAGE_BG, color = NA),
panel.grid.major = element_line(color = alpha(INK, 0.15), linewidth = 0.3),
panel.grid.minor = element_blank(),
axis.title = element_text(color = INK, size = 10),
axis.text = element_text(color = INK_SOFT, size = 8),
axis.line = element_line(color = INK_SOFT),
plot.title = element_text(color = INK, size = title_fontsize),
legend.text = element_text(color = INK_SOFT, size = 8),
legend.title = element_text(color = INK, size = 10),
legend.background = element_blank(),
legend.key = element_rect(fill = PAGE_BG, color = NA),
legend.position = "inside",
legend.position.inside = c(0.02, 0.05),
legend.justification = c(0, 0)
)
# --- Save -----------------------------------------------------------------
ggsave(
filename = sprintf("plot-%s.png", THEME),
plot = p,
device = ragg::agg_png,
width = 8,
height = 4.5,
units = "in",
dpi = 400
)
Runnable source as JSON, for any HTTP client: https://api.anyplot.ai/specs/precision-recall/ggplot2/code. Any spec id and library id listed in llms-full.txt fit the same URL shape; every URL below is complete and callable.
{
"spec_id": "precision-recall",
"language": "r",
"library": "ggplot2",
"page": "https://anyplot.ai/precision-recall/r/ggplot2",
"hub": "https://anyplot.ai/precision-recall",
"code_json": "https://api.anyplot.ai/specs/precision-recall/ggplot2/code",
"spec_json": "https://api.anyplot.ai/specs/precision-recall",
"render_light_png": "https://storage.googleapis.com/anyplot-images/plots/precision-recall/r/ggplot2/plot-light.png",
"render_dark_png": "https://storage.googleapis.com/anyplot-images/plots/precision-recall/r/ggplot2/plot-dark.png",
"quality_score": 92.0,
"license": "MIT",
"guide": "https://anyplot.ai/llms.txt"
}Part of Precision-Recall Curve on anyplot.ai.