Introduction to Sankey Diagram

Many times, we need a situation where we must visualize how data flows between entities. For example, take how residents move from one country to another. Here's a demonstration of how many residents moved from England to Northern Ireland, Scotland and Wales.

It is clear from this Sankey visualization that more residents moved from England to Wales than from Scotland or Northern Ireland.

What is a Sankey diagram?

Sankey diagrams usually depict the flow of data from one entity (or node) to another entity (or node).

The entities to which data flows are called nodes. The node where the data flow originates is the source node (for example, England on the left), and the node where the flow ends is the target node (for example, Wales on the right). Source and target nodes are usually represented as labeled rectangles.

The flow itself is represented by straight or curved paths, called links. The width of a stream/link is directly proportional to the volume/number of streams. In the example above, the movement from England to Wales (i.e. the migration of residents) is more extensive (i.e. the migration of residents) than the movement from England to Scotland or Northern Ireland (i.e. the migration of residents), indicating that more residents move to Wales than to other countries .

Sankey diagrams can be used to represent the flow of energy, money, costs, and anything that has a flow concept.

Minard's classic chart of Napoleon's invasion of Russia is probably the most famous example of a Sankey chart. This visualization using a Sankey diagram shows very effectively how the French army progressed (or decreased?) on its way to Russia and back.

#In this article, we use python’s plotly to draw a Sankey diagram.

How to draw a Sankey diagram?

This article uses the 2021 Olympic Games data set to draw a Sankey diagram. The dataset contains detailed information about the total number of medals - country, total number of medals, and individual totals for gold, silver, and bronze medals. We plot a Sankey chart to find out how many gold, silver and bronze medals a country has won.

df_medals = pd.read_excel("data/Medals.xlsx")
print(df_medals.info())
df_medals.rename(columns={'Team/NOC':'Country', 'Total': 'Total Medals', 'Gold':'Gold Medals', 'Silver': 'Silver Medals', 'Bronze': 'Bronze Medals'}, inplace=True)
df_medals.drop(columns=['Unnamed: 7','Unnamed: 8','Rank by Total'], inplace=True)

df_medals

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 93 entries, 0 to 92
Data columns (total 9 columns):
 # Column Non-Null CountDtype
--------- -------------------
 0 Rank 93 non-null int64
 1 Team/NOC 93 non-null object 
 2 Gold 93 non-null int64
 3 Silver 93 non-null int64
 4 Bronze 93 non-null int64
 5 Total93 non-null int64
 6 Rank by Total93 non-null int64
 7 Unnamed: 7 0 non-nullfloat64
 8 Unnamed: 8 1 non-nullfloat64
dtypes: float64(2), int64(6), object(1)
memory usage: 6.7+ KB
None

Sankey diagram drawing basics

Use plotly's go.Sankey, this method takes 2 parameters - nodes and links (nodes and links ).

Note: All nodes - source and target should have unique identifiers.

In the case of the Olympic medals data set in this article:

Source is the country. Consider the first 3 countries (United States, China, and Japan) as source nodes. Label these source nodes with the following (unique) identifiers, labels, and colors:

• 0: United States: Green
• 1: China: Blue
• 2: Japan: Orange

Target is gold, silver or bronze. Label these target nodes with the following (unique) identifiers, labels, and colors:

• 3: Gold Medal: Gold
• 4: Silver Medal: Silver
• 5 : Bronze: Brown

Link (between source node and target node) is the number of medals of each type. In each source there are 3 links, each ending with a target - Gold, Silver and Bronze. So there are 9 links in total. The width of each link should be the number of gold, silver and bronze medals. Tag these links to targets, values ??and colors with the following sources:

• 0 (US) to 3,4,5 : 39, 41, 33
• 1 (China) to 3 ,4,5 : 38, 32, 18
• 2 (Japan) to 3,4,5 : 27, 14, 17

You need to instantiate 2 python dict objects to Represents

• nodes (source and target): labels and colors as separate lists and
• links: source node, target node, value (width) and color of the link as separate List

and pass it to plotly's go.Sankey.

Each index of the list (label, source, target, value and color) corresponds to a node or link.

NODES = dict( 
# 0 1 23 4 5 
label = ["United States of America", "People's Republic of China", "Japan", "Gold", "Silver", "Bronze"],
color = ["seagreen", "dodgerblue", "orange", "gold", "silver", "brown" ],)
LINKS = dict( 
source = [0,0,0,1,1,1,2,2,2], # 鏈接的起點(diǎn)或源節(jié)點(diǎn)
target = [3,4,5,3,4,5,3,4,5], # 鏈接的目的地或目標(biāo)節(jié)點(diǎn)
value =[ 39, 41, 33, 38, 32, 18, 27, 14, 17], # 鏈接的寬度（數(shù)量）
# 鏈接的顏色
# 目標(biāo)節(jié)點(diǎn)： 3-Gold4-Silver5-Bronze
color = [ 
"lightgreen", "lightgreen", "lightgreen",# 源節(jié)點(diǎn)：0 - 美國(guó) States of America
"lightskyblue", "lightskyblue", "lightskyblue",# 源節(jié)點(diǎn)：1 - 中華人民共和國(guó)China
"bisque", "bisque", "bisque"],)# 源節(jié)點(diǎn)：2 - 日本
data = go.Sankey(node = NODES, link = LINKS)
fig = go.Figure(data)
fig.show()

This is a very basic Sankey diagram. But have you noticed that the chart is too wide and the silver medals appear before the gold medals?

Here’s how to adjust the position and width of nodes.

Adjust node positions and chart width

Add x and y positions to nodes to explicitly specify the node's position. Value should be between 0 and 1.

NODES = dict( 
# 0 1 23 4 5 
label = ["United States of America", "People's Republic of China", "Japan", "Gold", "Silver", "Bronze"],
color = ["seagreen", "dodgerblue", "orange", "gold", "silver", "brown" ],)
x = [ 0,0,0,0.5,0.5,0.5],
y = [ 0,0.5,1,0.1,0.5,1],)
data = go.Sankey(node = NODES, link = LINKS)
fig = go.Figure(data)
fig.update_layout(title="Olympics - 2021: Country &Medals",font_size=16)
fig.show()

So we got a compact Sankey diagram:

Let’s take a look at how the various parameters passed in the code are mapped to the nodes and nodes in the graph. Link.

代碼如何映射到?；鶊D

添加有意義的懸停標(biāo)簽

我們都知道plotly繪圖是交互的，我們可以將鼠標(biāo)懸停在節(jié)點(diǎn)和鏈接上以獲取更多信息。

帶有默認(rèn)懸停標(biāo)簽的?；鶊D

當(dāng)將鼠標(biāo)懸停在圖上，將會(huì)顯示詳細(xì)信息。懸停標(biāo)簽中顯示的信息是默認(rèn)文本：節(jié)點(diǎn)、節(jié)點(diǎn)名稱、傳入流數(shù)、傳出流數(shù)和總值。

例如：

• 節(jié)點(diǎn)美國(guó)共獲得11枚獎(jiǎng)牌(=39金+41銀+33銅)
• 節(jié)點(diǎn)金牌共有104枚獎(jiǎng)牌(=美國(guó)39枚，中國(guó)38枚，日本27枚)

如果我們覺(jué)得這些標(biāo)簽太冗長(zhǎng)了，我們可以對(duì)此進(jìn)程改進(jìn)。使用hovertemplate參數(shù)改進(jìn)懸停標(biāo)簽的格式

• 對(duì)于節(jié)點(diǎn)，由于hoverlabels 沒(méi)有提供新信息，通過(guò)傳遞一個(gè)空hovertemplate = ""來(lái)去掉hoverlabel
• 對(duì)于鏈接，可以使標(biāo)簽簡(jiǎn)潔，格式為-
• 對(duì)于節(jié)點(diǎn)和鏈接，讓我們使用后綴"Medals"顯示值。例如 113 枚獎(jiǎng)牌而不是 113 枚。這可以通過(guò)使用具有適當(dāng)valueformat和valuesuffix的update_traces函數(shù)來(lái)實(shí)現(xiàn)。

NODES = dict( 
# 0 1 23 4 5
label = ["United States of America", "People's Republic of China", "Japan", "Gold", "Silver", "Bronze"],
color = ["seagreen", "dodgerblue","orange", "gold", "silver", "brown" ],
x = [ 0,0, 0,0.5,0.5,0.5],
y = [ 0,0.5, 1,0.1,0.5,1],
hovertemplate=" ",)

LINK_LABELS = []
for country in ["USA","China","Japan"]:
for medal in ["Gold","Silver","Bronze"]:
LINK_LABELS.append(f"{country}-{medal}")
LINKS = dict(source = [0,0,0,1,1,1,2,2,2], 
 # 鏈接的起點(diǎn)或源節(jié)點(diǎn)
 target = [3,4,5,3,4,5,3,4,5], 
 # 鏈接的目的地或目標(biāo)節(jié)點(diǎn)
 value =[ 39, 41, 33, 38, 32, 18, 27, 14, 17], 
 # 鏈接的寬度（數(shù)量） 
 # 鏈接的顏色
 # 目標(biāo)節(jié)點(diǎn)：3-Gold4 -Silver5-Bronze
 color = ["lightgreen", "lightgreen", "lightgreen", # 源節(jié)點(diǎn)：0 - 美國(guó)
"lightskyblue", "lightskyblue", "lightskyblue", # 源節(jié)點(diǎn)：1 - 中國(guó)
"bisque", "bisque", "bisque"],# 源節(jié)點(diǎn)：2 - 日本
 label = LINK_LABELS, 
 hovertemplate="%{label}",)

data = go.Sankey(node = NODES, link = LINKS)
fig = go.Figure(data)
fig.update_layout(title="Olympics - 2021: Country &Medals",
font_size=16, width=1200, height=500,)
fig.update_traces(valueformat='3d', 
valuesuffix='Medals', 
selector=dict(type='sankey'))
fig.update_layout(hoverlabel=dict(bgcolor="lightgray",
font_size=16,
font_family="Rockwell"))
fig.show("png") #fig.show()

帶有改進(jìn)的懸停標(biāo)簽的?；鶊D

對(duì)多個(gè)節(jié)點(diǎn)和級(jí)別進(jìn)行泛化相對(duì)于鏈接，節(jié)點(diǎn)被稱為源和目標(biāo)。作為一個(gè)鏈接目標(biāo)的節(jié)點(diǎn)可以是另一個(gè)鏈接的源。

該代碼可以推廣到處理數(shù)據(jù)集中的所有國(guó)家。

還可以將圖表擴(kuò)展到另一個(gè)層次，以可視化各國(guó)的獎(jiǎng)牌總數(shù)。

NUM_COUNTRIES = 5
X_POS, Y_POS = 0.5, 1/(NUM_COUNTRIES-1)
NODE_COLORS = ["seagreen", "dodgerblue", "orange", "palevioletred", "darkcyan"]
LINK_COLORS = ["lightgreen", "lightskyblue", "bisque", "pink", "lightcyan"]

source = []
node_x_pos, node_y_pos = [], []
node_labels, node_colors = [], NODE_COLORS[0:NUM_COUNTRIES]
link_labels, link_colors, link_values = [], [], [] 

# 第一組鏈接和節(jié)點(diǎn)
for i in range(NUM_COUNTRIES):
source.extend([i]*3)
node_x_pos.append(0.01)
node_y_pos.append(round(i*Y_POS+0.01,2))
country = df_medals['Country'][i]
node_labels.append(country) 
for medal in ["Gold", "Silver", "Bronze"]:
link_labels.append(f"{country}-{medal}")
link_values.append(df_medals[f"{medal} Medals"][i])
link_colors.extend([LINK_COLORS[i]]*3)

source_last = max(source)+1
target = [ source_last, source_last+1, source_last+2] * NUM_COUNTRIES
target_last = max(target)+1

node_labels.extend(["Gold", "Silver", "Bronze"])
node_colors.extend(["gold", "silver", "brown"])
node_x_pos.extend([X_POS, X_POS, X_POS])
node_y_pos.extend([0.01, 0.5, 1])

# 最后一組鏈接和節(jié)點(diǎn)
source.extend([ source_last, source_last+1, source_last+2])
target.extend([target_last]*3)
node_labels.extend(["Total Medals"])
node_colors.extend(["grey"])
node_x_pos.extend([X_POS+0.25])
node_y_pos.extend([0.5])

for medal in ["Gold","Silver","Bronze"]:
link_labels.append(f"{medal}")
link_values.append(df_medals[f"{medal} Medals"][:i+1].sum())
link_colors.extend(["gold", "silver", "brown"])

print("node_labels", node_labels)
print("node_x_pos", node_x_pos); print("node_y_pos", node_y_pos)

node_labels ['United States of America', "People's Republic of China", 
 'Japan', 'Great Britain', 'ROC', 'Gold', 'Silver', 
 'Bronze', 'Total Medals']
node_x_pos [0.01, 0.01, 0.01, 0.01, 0.01, 0.5, 0.5, 0.5, 0.75]
node_y_pos [0.01, 0.26, 0.51, 0.76, 1.01, 0.01, 0.5, 1, 0.5]

# 顯示的圖
NODES = dict(pad= 20, thickness = 20, 
 line = dict(color = "lightslategrey",
 width = 0.5),
 hovertemplate=" ",
 label = node_labels, 
 color = node_colors,
 x = node_x_pos, 
 y = node_y_pos, )
LINKS = dict(source = source, 
 target = target, 
 value = link_values, 
 label = link_labels, 
 color = link_colors,
 hovertemplate="%{label}",)
data = go.Sankey(arrangement='snap', 
 node = NODES, 
 link = LINKS)
fig = go.Figure(data)
fig.update_traces(valueformat='3d', 
valuesuffix=' Medals', 
selector=dict(type='sankey'))
fig.update_layout(title="Olympics - 2021: Country &Medals",
font_size=16,
width=1200,
height=500,)
fig.update_layout(hoverlabel=dict(bgcolor="grey", 
font_size=14, 
font_family="Rockwell"))
fig.show("png") 

The above is the detailed content of Python draws stunning Sankey diagrams, have you learned it?. For more information, please follow other related articles on the PHP Chinese website!