使用Holoviews和Bokeh按指定顺序排列桑基图的源与目标

To enforce your desired Target order (DD → LC → NT → VU → EN → CE), use one of these straightforward solutions:

1. Convert Target to Ordered Categorical #

This is the simplest approach. By making the Target column an ordered categorical, Holoviews will respect your specified sequence when arranging nodes.

import pandas as pd
import holoviews as hv
import colorcet as cc

# Define your desired Target order
target_order = ["DD", "LC", "NT", "VU", "EN", "CE"]

# Convert Target column to ordered categorical type
sorted_df0['Target'] = pd.Categorical(sorted_df0['Target'], categories=target_order, ordered=True)

# Create Sankey diagram
sankey2 = hv.Sankey(sorted_df0, kdims=["Source", "Target"], vdims=["Value"])

# Set color palette and diagram options
color_palette = cc.glasbey_dark[:len(sorted_df0)]
opts.defaults(opts.Sankey(edge_color=hv.Cycle(values=color_palette)))
sankey2.opts(
    cmap="PuBuGn_r", 
    label_position='outer',
    width=1000, 
    height=700,
    title="AAAA"
)

2. Explicitly Define Node Order #

If the categorical method doesn’t work, directly specify the node_order in the Sankey options to force Target nodes into your desired sequence.

import holoviews as hv
import colorcet as cc

# Define desired Target order
target_order = ["DD", "LC", "NT", "VU", "EN", "CE"]

# Get unique Source nodes (preserves their original order from the dataset)
source_nodes = sorted_df0['Source'].unique().tolist()

# Combine Source nodes with Target nodes in your preferred order
node_order = source_nodes + target_order

# Create Sankey diagram
sankey2 = hv.Sankey(sorted_df0, kdims=["Source", "Target"], vdims=["Value"])

# Set options including custom node order
color_palette = cc.glasbey_dark[:len(sorted_df0)]
opts.defaults(opts.Sankey(edge_color=hv.Cycle(values=color_palette)))
sankey2.opts(
    cmap="PuBuGn_r", 
    label_position='outer',
    width=1000, 
    height=700,
    title="AAAA",
    node_order=node_order  # Enforce the custom node sequence
)

Key Notes: #

• Remove plt.tight_layout() from your code—it’s a Matplotlib function and unnecessary for Holoviews/Bokeh plots.
• Ensure all Target values in your dataset are included in the target_order list to avoid missing nodes.

内容的提问来源于stack exchange，提问作者Sajna v.h