nicolasesnis

import seaborn as sns
import pandas as pd
import plotly.graph_objects as go
import chart_studio.plotly as py
import plotly

# Path to the raw data: https://gist.github.com/nicolasesnis/eb3b35545e97926ab53e0617c5e4b639

data = pd.read_csv('your/path/to/the/raw/data.csv')[
    ['user_id', 'time_install', 'event_name', 'time_event']]

nicolasesnis

import plotly.graph_objects as go
import chart_studio.plotly as py
import plotly

fig = go.Figure(data=[go.Sankey(
    node=dict(
        thickness=10,  # default is 20
        line=dict(color="black", width=0.5),
        label=labels,
        color=colors

nicolasesnis

targets = []
sources = []
values = []
time_to_next = []

for source_key, source_value in output['links_dict'].items():
    for target_key, target_value in output['links_dict'][source_key].items():
        sources.append(source_key)
        targets.append(target_key)
        values.append(target_value['unique_users'])

nicolasesnis

# Working on the links_dict

output.update({'links_dict': dict()})

# Group the DataFrame by user_id and rank_event
grouped = data.groupby(['user_id', 'rank_event'])

# Define a function to read the souces, targets, values and time from event to next_event:
def update_source_target(user):
    try:

nicolasesnis

import seaborn as sns

# Working on the nodes_dict

all_events = list(data.event_name.unique())

# Create a set of colors that you'd like to use in your plot.
palette = ['50BE97', 'E4655C', 'FCC865',
           'BFD6DE', '3E5066', '353A3E', 'E6E6E6']
#  Here, I passed the colors as HEX, but we need to pass it as RGB. This loop will convert from HEX to RGB:

nicolasesnis

user_id	event_name	time_event	rank_action	next_action	time_to_next
001...	install	2020-03-05 15:48:00	1	signup	0 days 00:03:00
001...	signup	2020-03-05 15:51:00	2	NaN	NaT
003...	install	2020-03-02 04:38:00	1	signup	0 days 00:04:00
003...	signup	2020-03-02 04:42:00	2	reopen	2 days 02:42:00
003...	reopen	2020-03-04 07:24:00	3	NaN	NaT

nicolasesnis

import pandas as pd

data = pd.read_csv('raw.csv')[
    ['user_id', 'time_install', 'event_name', 'time_event']]

# Start with making sure that time_event and time_insrall are Pandas Datetime types:
data['time_event'] = pd.to_datetime(data['time_event'], unit='s') # Unit = "s" is required here since the raw data format is unique timestamp. Remove this param if your data is already a datetime like data.
data['time_install'] = pd.to_datetime(data['time_install'], unit='s')

# Make sure that there's no event occurring before time_install

nicolasesnis

user_id,time_install,event_name,time_event
4c6065c9466bc68e324e316edfb0227ff7cccc6c,43892.9781365741,purchase,43892.9838078704
f3049eac4788ffd4482390f8333d7e1adbf4c5a1,43896.2203703704,signup,43896.2207986111
f3049eac4788ffd4482390f8333d7e1adbf4c5a1,43896.2203703704,purchase,43896.2218865741
9c1e35e89a374207409ad05da6d69d43f427c5f2,43896.7880555556,reopen,43896.1077083333
723fba1295b9a7c8321bbc433f87629b90660582,43892.0000462963,reopen,43893.5774884259
723fba1295b9a7c8321bbc433f87629b90660582,43892.0000462963,signup,43892.0025231481
97963ab19ae86316d639e3a181a156c6c9dd5be0,43892.9100231481,purchase,43893.0026157407
97963ab19ae86316d639e3a181a156c6c9dd5be0,43892.9100231481,signup,43892.9137152778
97963ab19ae86316d639e3a181a156c6c9dd5be0,43892.9100231481,purchase,43892.9711689815

nicolasesnis

user_id	time_install	event_name	time_event
4c6065c9466bc6ad...	2020-03-02 23:28:31	purchase	2020-03-02 23:36:41
f3049eac4788ffak...	2020-03-06 05:17:20	signup	2020-03-06 05:17:57
f3049eac4788ffak...	2020-03-06 05:17:20	purchase	2020-03-06 05:19:31
9c1e35e89a3742e1...	2020-03-06 18:54:48	reopen	2020-03-06 02:35:06
723fba1295b9a7fb...	2020-03-02 00:00:04	reopen	2020-03-03 13:51:35