Tutorial to visualize customer behavior. More details here: https://github.com/hadinh1306/blog-posts/blob/master/tutorials/Visualize_Customer_Behavior.ipynb

visualize_customer_behavior.py

import pandas as pd
import numpy as np

import matplotlib.pyplot as plt
import matplotlib.dates as mdates

import warnings
warnings.filterwarnings("ignore")

# Load Data
republished_df = pd.read_csv("../data/visualize-customer-behavior/republished_sample.csv")
login_df = pd.read_csv("../data/visualize-customer-behavior/login_sample.csv")
subscription_info_df = pd.read_csv("../data/visualize-customer-behavior/subscription_info.csv")

# Transform Data
republished_df['action_date'] = pd.to_datetime(republished_df['action_date'])
login_df['action_date'] = pd.to_datetime(login_df['action_date'])
subscription_info_df['subscription_starts_at'] = pd.to_datetime(subscription_info_df['subscription_starts_at'])
subscription_info_df['subscription_ends_at'] = pd.to_datetime(subscription_info_df['subscription_ends_at'])

sample_subscription = subscription_info_df[subscription_info_df['AccountCode'] == 'a']
sample_republished = republished_df[republished_df['AccountCode'] == 'a']
sample_login = login_df[login_df['AccountCode'] == 'a']

# this is a constant for visualization purpose
sample_subscription['vizline'] = 0.5
sample_republished['vizline'] = 0.5
sample_login['vizline'] = 0.5

fig, ax = plt.subplots(figsize=(20, 5))

# Plot subscription starts and ends
ax.plot(sample_subscription['subscription_starts_at'], sample_subscription['vizline'], 
        marker='|', linewidth = 0.1,
        markersize=50, mew=2, alpha=0.5,
        color='royalblue', label='Subscription Starts')

no_expire_mask = ~sample_subscription['subscription_ends_at'].isnull()
ax.plot(sample_subscription[no_expire_mask]['subscription_ends_at'], sample_subscription[no_expire_mask]['vizline'],
        linewidth = 0.1, marker='|',
        markersize=50, mew=2, alpha=0.5,
        color='crimson', label='Subscription Ends')

# Plot login and republish events
ax.plot(sample_login['action_date'], sample_login['vizline'],
        marker='o', markersize=11,
        alpha=0.3, color='darkseagreen',
        linewidth=0.1, label='Login')

ax.plot(sample_republished['action_date'], sample_republished['vizline'],
        marker='^', markersize=8,
        alpha=0.5, color='teal',
        linewidth=0.1, label='Republish')

# Limit date range
datemin = pd.to_datetime('2019/01/01').date()
datemax = pd.to_datetime('2019/12/31').date()
ax.set_xlim(datemin, datemax)

# Format date
date_form = mdates.DateFormatter("%Y/%m/%d")
ax.xaxis.set_major_formatter(date_form)
# Ensure ticks fall once every other week (interval=2) 
ax.xaxis.set_major_locator(mdates.WeekdayLocator(interval=2))
ax.xaxis.set_tick_params(rotation=40)

ax.legend(loc='upper left', ncol=4)
ax.set_title("Customer Behavior")

ax.get_yaxis().set_visible(False)

# Make it reproducible with functions
def _get_sample_data(AccountCode):
    """This function gets subscription info, login events and republish events for the AccountCode input.
    
    Args:
        AccountCode (str): Account unique identification.

    Returns:
        pandas.core.frame.DataFrame: 3 dataframes with subscription info, login and republish events.
    """
    sample_subscription = subscription_info_df[subscription_info_df['AccountCode'] == AccountCode]
    sample_republished = republished_df[republished_df['AccountCode'] == AccountCode]
    sample_login = login_df[login_df['AccountCode'] == AccountCode]

    # this is a constant for visualization purpose
    sample_subscription['vizline'] = 0.5
    sample_republished['vizline'] = 0.5
    sample_login['vizline'] = 0.5
    
    return sample_subscription, sample_republished, sample_login

def _visualize_customer_behavior(AccountCode):
    """This function visualizes customer behavior using subscription, login and republish events of a customer.
    
    Args: 
        AccountCode (str): Account unique identification.
        
    Returns:
        matplotlib.figure.Figure: a visualization with subscription, login and republish events of a customer.
    """
    sample_subscription, sample_republished, sample_login = _get_sample_data(AccountCode)
    
    fig, ax = plt.subplots(figsize=(20, 5))

    # Plot subscription starts and ends
    ax.plot(sample_subscription['subscription_starts_at'], sample_subscription['vizline'], 
            marker='|', linewidth = 0.1,
            markersize=50, mew=2, alpha=0.5,
            color='royalblue', label='Subscription Starts')

    no_expire_mask = ~sample_subscription['subscription_ends_at'].isnull()
    ax.plot(sample_subscription[no_expire_mask]['subscription_ends_at'], sample_subscription[no_expire_mask]['vizline'],
            linewidth = 0.1, marker='|',
            markersize=50, mew=2, alpha=0.5,
            color='crimson', label='Subscription Ends')

    # Plot login and republish events
    ax.plot(sample_login['action_date'], sample_login['vizline'],
            marker='o', markersize=11,
            alpha=0.3, color='darkseagreen',
            linewidth=0.1, label='Login')

    ax.plot(sample_republished['action_date'], sample_republished['vizline'],
            marker='^', markersize=8,
            alpha=0.5, color='teal',
            linewidth=0.1, label='Republish')

    # Limit date range
    datemin = pd.to_datetime('2019/01/01').date()
    datemax = pd.to_datetime('2019/12/31').date()
    ax.set_xlim(datemin, datemax)

    # Show weekly date
    date_form = mdates.DateFormatter("%Y/%m/%d")
    ax.xaxis.set_major_formatter(date_form)
    # Ensure ticks fall once every other week (interval=2) 
    ax.xaxis.set_major_locator(mdates.WeekdayLocator(interval=2))
    ax.xaxis.set_tick_params(rotation=40)

    ax.legend(loc='upper left', ncol=4)
    ax.set_title("Customer Behavior")

    ax.get_yaxis().set_visible(False)
    
    return fig