Lesson 4 - Multiple Lines and Series

Comparing Multiple Datasets#

You can create and customize single-line plots. Now you will learn to plot multiple lines on the same graph, load real-world time series data, and compare trends visually.

By the end of this lesson, you will be able to:

• Load time series data from CSV files using Pandas
• Convert string dates to datetime objects
• Plot time series data with proper date formatting
• Create multiple lines on the same graph
• Add legends to distinguish different lines
• Work with real-world bike-sharing data

Comparing multiple series on one graph reveals relationships, differences, and patterns that single plots cannot show.

What is Time Series Data?#

Understanding Time Series#

Time series data is any data collected over time at regular intervals:

• Stock prices (every second, minute, or day)
• Weather measurements (hourly, daily)
• Sales figures (daily, monthly, yearly)
• Website traffic (hourly, daily)
• Bike rentals (hourly, daily)

Why It Matters#

• Most business data is time series
• Helps identify trends and patterns
• Critical for forecasting and planning
• Reveals seasonal effects
• Shows growth or decline over time

Time series analysis is one of the most common tasks in data science.

Loading Data with Pandas#

We will use the Bike Sharing Dataset - real data from Washington D.C.’s Capital Bikeshare system (2011-2012).

About the Dataset#

The dataset contains:

• Date of each day (731 days total)
• Number of bike rentals
• Weather conditions (temperature, humidity, windspeed)
• Whether it was a holiday or working day
• Casual vs registered users

Load the Data#

import pandas as pd
import matplotlib.pyplot as plt

# Load the bike sharing data
bikes = pd.read_csv('day.csv')

# Look at the first few rows
print(bikes.head())

Key Columns#

• dteday - The date (currently a string)
• cnt - Total bike rentals that day
• casual - Rentals by casual (non-registered) users
• registered - Rentals by registered users
• temp - Normalized temperature
• weathersit - Weather situation (1=Clear, 2=Mist, 3=Light Rain/Snow)

Notice: The dteday column contains dates like “2011-01-01”, but they are stored as text (strings), not dates.

Converting Strings to Datetime#

To properly work with dates, we need to convert them from strings to datetime objects.

Why Convert?#

• Matplotlib can automatically format date axes
• You can do date math (“what was 30 days ago?”)
• You can extract parts like month, year, day of week
• Enables proper time-based operations

Use pd.to_datetime()#

import pandas as pd
import matplotlib.pyplot as plt

bikes = pd.read_csv('day.csv')

# Convert dteday column from string to datetime
bikes['dteday'] = pd.to_datetime(bikes['dteday'])

# Check the data type
print("Data type:", bikes['dteday'].dtype)
print("\nFirst 5 dates:")
print(bikes['dteday'].head())

What Changed?#

• Before: dtype: object (string)
• After: dtype: datetime64[ns] (datetime)

Now Python knows these are dates, not just random text.

Plotting Time Series Data#

Now we can plot bike rentals over time. Let’s see how demand changed across 2 years:

import pandas as pd
import matplotlib.pyplot as plt

bikes = pd.read_csv('day.csv')
bikes['dteday'] = pd.to_datetime(bikes['dteday'])

# Plot total rentals over time
plt.plot(bikes['dteday'], bikes['cnt'])
plt.title('Daily Bike Rentals in Washington D.C. (2011-2012)')
plt.xlabel('Date')
plt.ylabel('Number of Rentals')

# Rotate x-axis labels for better readability
plt.xticks(rotation=45)

plt.show()

What Do You Notice?#

• Clear seasonal pattern - high in summer, low in winter
• Overall upward trend - more rentals in 2012 than 2011
• Lots of daily variation (the “noise”)

About plt.xticks(rotation=45)#

• Dates are long text (“2011-01-01”)
• Rotating by 45 degrees prevents overlap
• Makes the graph much more readable

Common rotation values:

• 0° (horizontal) - default
• 45° (diagonal) - good for dates
• 90° (vertical) - extreme cases

Multiple Lines on One Graph#

Let’s compare casual users vs registered users. Do they have different patterns?

Plotting Two Lines#

To plot multiple lines, call plt.plot() multiple times before plt.show():

import pandas as pd
import matplotlib.pyplot as plt

bikes = pd.read_csv('day.csv')
bikes['dteday'] = pd.to_datetime(bikes['dteday'])

# Plot both casual and registered users
plt.plot(bikes['dteday'], bikes['casual'], label='Casual Users')
plt.plot(bikes['dteday'], bikes['registered'], label='Registered Users')

plt.title('Casual vs Registered Bike Users (2011-2012)')
plt.xlabel('Date')
plt.ylabel('Number of Rentals')
plt.xticks(rotation=45)

# Add legend to show which line is which
plt.legend()

plt.show()

Understanding the Code#

• label='Casual Users' - Names the first line
• label='Registered Users' - Names the second line
• plt.legend() - Shows a box with the labels

What Patterns Do You See?#

• Registered users (blue) are consistently higher
• Both show seasonal patterns
• Casual users have more extreme peaks (summer weekends?)
• Registered users more stable year-round

Hypothesis: Casual users are tourists/recreational riders (weather-dependent). Registered users are commuters (ride year-round regardless of weather).

Customizing Multiple Lines#

Make each line distinctive with colors, styles, and markers:

import pandas as pd
import matplotlib.pyplot as plt

bikes = pd.read_csv('day.csv')
bikes['dteday'] = pd.to_datetime(bikes['dteday'])

# Plot with custom styling
plt.plot(bikes['dteday'], bikes['casual'],
         color='orange',
         linestyle='--',
         linewidth=2,
         label='Casual Users')

plt.plot(bikes['dteday'], bikes['registered'],
         color='blue',
         linestyle='-',
         linewidth=2,
         label='Registered Users')

plt.title('Casual vs Registered Bike Users (2011-2012)')
plt.xlabel('Date')
plt.ylabel('Number of Rentals')
plt.xticks(rotation=45)
plt.legend()

plt.show()

Styling Tips for Multiple Lines#

Use different colors:

• Makes lines easy to distinguish
• Choose contrasting colors

Use different line styles:

• Solid vs dashed helps colorblind readers
• Provides redundant encoding

Consistent line width:

• Use same width for equal importance
• Use thicker lines for emphasis

Practice Exercises#

Apply time series and multiple line techniques.

Exercise 1: First 3 Months#

Create a line plot showing only the first 3 months of 2011 (January - March).

Hint: Use slicing to get the first 90 rows (approximately 3 months)

bikes_3months = bikes.head(90)

Requirements:

1. Plot total rentals (cnt) for the first 90 days
2. Title: “Bike Rentals - First Quarter 2011”
3. Proper axis labels
4. Rotated x-axis labels

# Your code here

import pandas as pd
import matplotlib.pyplot as plt

bikes = pd.read_csv('day.csv')
bikes['dteday'] = pd.to_datetime(bikes['dteday'])

bikes_3months = bikes.head(90)

plt.plot(bikes_3months['dteday'], bikes_3months['cnt'])
plt.title('Bike Rentals - First Quarter 2011')
plt.xlabel('Date')
plt.ylabel('Number of Rentals')
plt.xticks(rotation=45)

plt.show()

Exercise 2: Seasonal Comparison#

The dataset has a season column:

• 1 = Spring
• 2 = Summer
• 3 = Fall
• 4 = Winter

Task: Create separate plots for Summer and Winter to compare them.

Hint: Filter the data like this:

summer = bikes[bikes['season'] == 2]
winter = bikes[bikes['season'] == 4]

Requirements:

1. Plot both seasons on the same graph with different colors
2. Add a legend
3. Use meaningful title and labels

# Your code here

import pandas as pd
import matplotlib.pyplot as plt

bikes = pd.read_csv('day.csv')
bikes['dteday'] = pd.to_datetime(bikes['dteday'])

summer = bikes[bikes['season'] == 2]
winter = bikes[bikes['season'] == 4]

plt.plot(summer['dteday'], summer['cnt'],
         color='orange', label='Summer', marker='o', markersize=3)
plt.plot(winter['dteday'], winter['cnt'],
         color='blue', label='Winter', marker='s', markersize=3)

plt.title('Bike Rentals: Summer vs Winter')
plt.xlabel('Date')
plt.ylabel('Number of Rentals')
plt.xticks(rotation=45)
plt.legend()

plt.show()

Exercise 3: Growth Analysis#

Compare the same month across two years to see year-over-year growth.

Task: Plot July 2011 vs July 2012 on the same graph.

Hint: Filter by month:

bikes['month'] = bikes['dteday'].dt.month
july_2011 = bikes[(bikes['month'] == 7) & (bikes['dteday'].dt.year == 2011)]
july_2012 = bikes[(bikes['month'] == 7) & (bikes['dteday'].dt.year == 2012)]

Challenge: Can you calculate the average growth percentage?

# Your code here

import pandas as pd
import matplotlib.pyplot as plt

bikes = pd.read_csv('day.csv')
bikes['dteday'] = pd.to_datetime(bikes['dteday'])
bikes['month'] = bikes['dteday'].dt.month

july_2011 = bikes[(bikes['month'] == 7) & (bikes['dteday'].dt.year == 2011)]
july_2012 = bikes[(bikes['month'] == 7) & (bikes['dteday'].dt.year == 2012)]

plt.plot(july_2011['dteday'].dt.day, july_2011['cnt'],
         label='July 2011', marker='o')
plt.plot(july_2012['dteday'].dt.day, july_2012['cnt'],
         label='July 2012', marker='s')

plt.title('Year-over-Year Growth: July Comparison')
plt.xlabel('Day of Month')
plt.ylabel('Number of Rentals')
plt.legend()

# Calculate growth
avg_2011 = july_2011['cnt'].mean()
avg_2012 = july_2012['cnt'].mean()
growth = ((avg_2012 - avg_2011) / avg_2011) * 100
print(f"Average growth from July 2011 to July 2012: {growth:.1f}%")

plt.show()

Summary#

You now visualize and compare time series data. Let’s review the key concepts.

Key Concepts#

Load CSV Data

• Use pd.read_csv('filename.csv')
• Returns a DataFrame
• Check with .head() to see structure

Convert Dates

• Use pd.to_datetime(df['date_column'])
• Enables proper date handling
• Automatic axis formatting in plots

Rotate Labels

• Use plt.xticks(rotation=45) for readability
• Essential for date labels
• Prevents overlapping text

Multiple Lines

• Call plt.plot() multiple times
• Use label= parameter for each line
• Add plt.legend() to show labels
• Use different colors and styles

Time Series Patterns

• Trends: Overall direction (upward/downward)
• Seasonality: Repeating patterns (yearly, weekly, daily)
• Noise: Daily/random fluctuations

Filter Data

• Focus on specific time periods
• .head(n) for first n rows
• df[df['column'] == value] for conditions
• Extract date parts: .dt.year, .dt.month, .dt.day

Syntax Reference#

# Load and prepare data
import pandas as pd
import matplotlib.pyplot as plt

df = pd.read_csv('data.csv')
df['date'] = pd.to_datetime(df['date'])

# Plot multiple lines
plt.plot(df['date'], df['series1'], label='Series 1')
plt.plot(df['date'], df['series2'], label='Series 2')

# Formatting
plt.title('Title Here')
plt.xlabel('Date')
plt.ylabel('Value')
plt.xticks(rotation=45)
plt.legend()

plt.show()

Common Questions#

Q: My dates show weird formats like “2011-01-01 00:00:00”. How do I fix this? A: This is normal. Datetime objects include time. Matplotlib usually formats them nicely automatically.

Q: What if my CSV has dates in a different format like “01/15/2023”? A: pd.to_datetime() is smart and handles most formats automatically. For unusual formats, use: pd.to_datetime(df['date'], format='%m/%d/%Y')

Q: Can I plot more than 2 lines? A: Yes! Plot as many as needed, but more than 5-6 lines gets hard to read. Consider using multiple subplots instead (coming later).

Next Steps#

You have completed Module 1: Line Plots and Time Series. You now know how to create, customize, and compare line plots with real-world data.

In Module 2: Scatter Plots and Correlation, you will learn to explore relationships between variables, identify correlations, and understand causation.

Master Time Series Visualization#

You can now load, process, and visualize time series data from real-world sources. These skills apply to sales data, website traffic, stock prices, and any metric that changes over time.

Use these techniques to reveal trends and patterns in your data!