R 如何自定义Shapviz图

发布于04月08日

我是{shapviz}包的最近用户，在自定义图表(如重要性图表)时遇到了一些困难.

例如，我想:

在beeswarm图的情况下改变调色板和在条形图的情况下改变 colored颜色；
[something like 100]更改图例(标题和尺寸)；
更改数字的文本/标签(show_numbers = TRUE)[as if I were using the geom_text( ), 101 or 102 functions].

由于Shapviz图属于ggplot类型，我试图通过访问列表或添加图层来更改它们，但没有成功.

谢谢！

推荐答案

使用一些玩具数据，以下是你如何改变这些美学:

library(xgboost)
library(shapviz)
library(viridis)
library(ggplot2)

# Toy data
X_train <- data.matrix(iris[, -1])
dtrain <- xgb.DMatrix(X_train, label = iris[, 1], nthread = 1)
fit <- xgb.train(data = dtrain, nrounds = 10, nthread = 1)
x <- shapviz(fit, X_pred = X_train)

# Modify shapviz plot
p <- sv_importance(x, kind = "both", show_numbers = TRUE, fill = "red") +
  scale_colour_gradient(
    low = "darkblue", high = "skyblue",
    breaks = c(0, 1), labels = c("low", "high")
  ) +
  labs(colour = "My Title") +
  theme(
    legend.key.size = unit(2, "cm"),
    legend.title.position = "right",
    legend.title = element_text(angle = 90, hjust = 0.5)
  )

# Add geom_label and remove the geom_text
p2 <- p +
  geom_label(aes(x, y, label = label), data = layer_data(p, 4), fontface = "italic")

p2$layers[[4]] <- NULL

p2