Help with either Parallel Process or Sankey Diagram?

📊 Plotly Python
1

So I have some data that is read in from a CSV. Assume these are in dataframes named summer, fall, & winter.

data

The number of workers varies by season. As does the type of fruit picked.

What I want to create is something like this (but with real data and not in powerpoint lol):

parallel
parallel1116×627 10.5 KB

Each axis or row is a season. Each worker connects to their name next season, but even if they don’t work next season, their output is still shown. Connecting the workers names between the seasons is one line/connection per fruit picked the width being based of the count.

But I can’t for the life of me figure out how to get the data arranged in a way to get it graphing right. If I just do workers on the left and all the seasons on the far right it works, but I want to be be able to eventually show a filter for say apples and highlight the path of all apple picks by each person each season.

2

Hey @kardiff welcome to the forums. I don’t have experience with this graph type. It seems to be quite complex indeed. I tried with plotly.express, I leave my intents here just in case it might inspire you. Spoiler: it is not an answer to your question, unfortunately.

Data: csv_parcords - Google Drive

import pandas as pd
import plotly.express as px

spring = pd.read_csv('page-1_table-1.csv')
spring['season'] = 'spring'
summer = pd.read_csv('page-1_table-2.csv')
summer['season'] = 'summer'
fall = pd.read_csv('page-1_table-3.csv')
fall['season'] = 'fall'

df = pd.concat([spring, summer, fall])
df = pd.melt(df, id_vars=['Fruit', 'season'], var_name='name', value_vars=df.columns[1:-1])

fig = px.parallel_categories(
    df, 
    dimensions=['Fruit', 'season', 'name']
)
fig.show()

newplot(12)
newplot(12)1488×360 229 KB

I think you would need to rearrange the df so that you have each season as different column.

3

What you’ve drawn looks a bit like a series of bar charts to me, with one bar chart per worker (and x=season, y=production, color=fruit). Would that approach give you something close enough to what you are trying to do?