如何用gnuplot绘制多类别基于Rank的累积占比曲线？

Hey there! Let's break this down into two clear, actionable steps: first calculating that cumulative proportion column using Python, then plotting the three category curves in gnuplot.

Step 1: Calculate Cumulative Proportions with Python #

Since you're comfortable with bash but need to handle this data calculation, Python's pandas library makes this process quick and clean. Here's a script that reads your raw data, computes the cumulative percentage for each category, and saves the result to a new file:

import pandas as pd

# Load your input data (update the filename to match your actual file)
# Assuming your input has no header row; adjust names if your data has headers
df = pd.read_csv("raw_data.csv", header=None, names=["Rank", "Category"])

# First, get the total number of entries for each category
total_per_category = df["Category"].value_counts()

# Calculate cumulative count per category (add 1 because cumcount starts at 0)
df["Cumulative_Pct"] = df.groupby("Category").cumcount() + 1

# Divide by total to get the proportion (0-1 range)
df["Cumulative_Pct"] = df["Cumulative_Pct"] / df["Category"].map(total_per_category)

# Save the result to a new CSV (this will have all three columns)
df.to_csv("processed_data.csv", index=False)

Quick breakdown of the logic: #

• groupby("Category").cumcount() tracks how many times the current category has appeared before the current row; adding 1 gives the total count up to and including the current row.
• map(total_per_category) looks up the total number of entries for each category, so dividing the cumulative count by this total gives the cumulative proportion you need.

Step 2: Plot the Curves with Gnuplot #

Once you have the processed CSV with all three columns, you can use gnuplot to plot each category as a distinct curve. Here's a sample gnuplot script tailored to your needs:

# Set output to a PNG file (adjust size/format as needed)
set terminal pngcairo enhanced font 'Arial,12' size 1000,700
set output 'category_cumulative_plot.png'

# Configure axis labels and plot title
set xlabel "Rank" font 'Arial,14'
set ylabel "Cumulative Proportion" font 'Arial,14'
set title "Cumulative Distribution of Categories by Rank" font 'Arial,16'

# Set y-axis to range from 0 to 1 (matches your proportion values)
set yrange [0:1]
set grid ytics lt 0 lw 1 lc rgb '#eeeeee'

# Plot each category as a distinct line
plot 'processed_data.csv' using 1:($2 eq "type1" ? $3 : NaN) \
     title 'Type 1' with lines lt 1 lc rgb '#1f77b4' lw 2, \
     '' using 1:($2 eq "type2" ? $3 : NaN) \
     title 'Type 2' with lines lt 1 lc rgb '#ff7f0e' lw 2, \
     '' using 1:($2 eq "type3" ? $3 : NaN) \
     title 'Type 3' with lines lt 1 lc rgb '#2ca02c' lw 2

Key details in the plot script: #

• The using 1:($2 eq "type1" ? $3 : NaN) clause filters rows to only include those where the category is "type1", using Rank (column 1) as the X-axis and Cumulative_Pct (column 3) as the Y-axis. Rows that don't match are ignored (marked as NaN).
• lt sets line type, lc rgb sets a unique color for each category, and lw increases line width for better readability.
• The subtle grid lines make it easier to interpret the y-axis proportion values.

To run this, save the script as plot_script.gp and execute it with gnuplot plot_script.gp in your terminal.

内容的提问来源于stack exchange，提问作者B. du Garay