📟
Charl•se

Plot beautiful time series with Python seaborn

Setup

For this example we extract dataset contains daily values for the Nasdaq 100 and Dow Jones indices, between 01/01/2019 and 19/05/2020. Which was quite an interesting period, due to the Covid crash.

# Import libraries
import pandas as pd
import seaborn as sns

# Setup
%matplotlib inline
sns.set(rc={'figure.figsize':(12,8)})

# Read sample data, parsing date field
df = pd.read_csv('data/finance_data.csv', parse_dates=['Date'])
df.tail()
Name
Date
Value
691
Nasdaq100
2020-05-13
9000.08
692
Nasdaq100
2020-05-14
9094.42
693
Nasdaq100
2020-05-15
9152.64
694
Nasdaq100
2020-05-18
9331.93
695
Nasdaq100
2020-05-19
9390.94

Plot relative variation with emphasis on specific periods

# Normalize values to the first date
df['Var'] = (
    df
    .groupby('Name')
    .apply(lambda x: x['Value']/(x.head(1)['Value'].values))
    .values-1
)

# Plot all values
sns.lineplot(
    data=df, x='Date', y='Var', units='Name', 
    estimator=None, color='grey', alpha=0.3
)

# Plot solid line within date range
sns.lineplot(
    data=df.query('Name == "Nasdaq100" & Date >= "2019-11-04" & Date <= "2020-03-06"'), 
    x='Date', y='Var', color='peru'
)
sns.lineplot(
    data=df.query('Name == "DowJones" & Date >= "2019-05-02" & Date <= "2019-12-12"'), 
    x='Date', y='Var', color='steelblue'
)

# Plot points at first and last date
sns.scatterplot(
    data=df.query('Name == "Nasdaq100" & Date in ["2019-11-04", "2020-03-06"]'), 
    x='Date', y='Var', color='peru'
)
sns.scatterplot(
    data=df.query('Name == "DowJones" & Date in ["2019-05-02", "2019-12-12"]'), 
    x='Date', y='Var', color='steelblue'
);
image

Plot moving average

# Compute moving average over 15 days (centered)
df['MovAvg'] = (
    df
    .groupby('Name')
    .apply(lambda x: x['Var'].rolling(window=15, center=True).mean())
    .values
)

# Plot moving average
sns.lineplot(data=df, x='Date', y='MovAvg', hue='Name', linestyle='--')

# Plot daily values
sns.lineplot(
    data=df, x='Date', y='Var', units='Name', color='grey', 
    estimator=None, alpha=0.3, legend=None
);
image