23  Upset Plot

library(ggplot2)
library(tidyr)
library(patchwork)
library(scales)
library(dplyr)

# 1. 模拟数据准备
df_models <- data.frame(
  var1 = c(TRUE,  FALSE, FALSE, TRUE,  TRUE,  FALSE, TRUE),
  var2 = c(FALSE, TRUE,  FALSE, TRUE,  FALSE, TRUE,  TRUE),
  var3 = c(FALSE, FALSE, TRUE,  FALSE, TRUE,  TRUE,  TRUE),
  Metric = c(0.125, 0.182, 0.085, 0.254, 0.312, 0.221, 0.458) # 小数格式的指标
)

# 按指标从大到小排序，并创建用于对齐的 model_id
df_models <- df_models %>% 
  arrange(desc(Metric)) %>% 
  mutate(model_id = factor(row_number(), levels = row_number()))

# 转换成长数据供底部点阵图使用
df_long <- df_models %>% 
  pivot_longer(cols = c(var1, var2, var3), names_to = "Variable", values_to = "Selected") %>% 
  mutate(Variable = factor(Variable, levels = c("var3", "var2", "var1"))) # 控制Y轴变量顺序

# ==========================================
# 📊 组件一：顶部 Metric 柱状图（geom_col + 百分比）
# ==========================================
p_top <- ggplot(df_models, aes(x = model_id, y = Metric)) +
  geom_col(fill = "#2b5c8f", width = 0.55, alpha = 0.9) +
  # 柱子上方显示精确的百分比标签
  geom_text(aes(label = percent(Metric, accuracy = 0.1)), vjust = -0.5, size = 3.8, fontface = "bold") +
  # Y轴使用百分比格式化
  scale_y_continuous(labels = percent_format(accuracy = 1), expand = expansion(mult = c(0, 0.15))) +
  labs(title = "不同变量组合模型的性能评估", x = "", y = "性能提升 (%)") +
  theme_minimal(base_size = 12) +
  theme(
    axis.text.x = element_blank(),       # 隐藏横轴文字，交给底部点阵图
    axis.ticks.x = element_blank(),
    panel.grid.major.x = element_blank(),
    panel.grid.minor = element_blank(),
    plot.title = element_text(face = "bold", hjust = 0.5, size = 14)
  )

# ==========================================
# 🔴 组件二：底部交集点阵图（标准 UpSet 样式）
# ==========================================
p_bottom <- ggplot(df_long, aes(x = model_id, y = Variable)) +
  # 背景灰色横向轨道线
  geom_line(aes(group = Variable), color = "#f2f2f2", linewidth = 3) +
  # 未选中的浅灰色点
  geom_point(data = filter(df_long, !Selected), color = "#e0e0e0", size = 4.5) +
  # 连线：把同一个模型中选中的变量连起来
  geom_line(data = filter(df_long, Selected), aes(group = model_id), color = "#2c3e50", linewidth = 1.2) +
  # 已选中的深色点
  geom_point(data = filter(df_long, Selected), color = "#2c3e50", size = 5.5) +
  labs(x = "模型组合 (已按性能从高到低排序)", y = "输入变量") +
  theme_minimal(base_size = 12) +
  theme(
    panel.grid.major = element_blank(),
    panel.grid.minor = element_blank(),
    axis.text.x = element_text(face = "bold"), # 显示模型组合的序号
    axis.text.y = element_text(face = "bold", size = 12)
  )

# ==========================================
# 📐 完美上下垂直拼接并对齐
# ==========================================
# 使用 patchwork 的 / 运算符直接上下拼接，指定高度比例为 6:4
p_final <- p_top / p_bottom + plot_layout(heights = c(1.5, 1))

# 渲染图片
print(p_final)