Practical Examples and Case Studies

Apply your skills with Pandas and NumPy to real-world scenarios through practical examples and detailed case studies.

Overview

Time to put it all together. We’ll work through one complete example, from generating data to a chart, then give you some projects to try on your own.

Worked example: robot battery log

Imagine your robot logs its battery level every hour for a week. We’ll use NumPy to make some realistic-looking data, then clean it, analyse it and chart it with Pandas.

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt

# 1. Make a week of hourly readings
rng = np.random.default_rng(seed=42)
times = pd.date_range('2024-01-01', periods=24 * 7, freq='h')
battery = 100 - (np.arange(len(times)) % 24) * 3 + rng.normal(0, 2, len(times))

log = pd.DataFrame({'time': times, 'battery': battery.round(1)})

# 2. Knock out a few readings to simulate a flaky sensor
log.loc[rng.choice(len(log), 10, replace=False), 'battery'] = np.nan

# 3. Clean: fill gaps by interpolating between neighbouring readings
print("Missing before:", log['battery'].isna().sum())
log['battery'] = log['battery'].interpolate()
print("Missing after:", log['battery'].isna().sum())

# 4. Analyse: lowest battery each day, and a 6-hour rolling average
daily_min = log.resample('D', on='time')['battery'].min()
print(daily_min)

log['smooth'] = log['battery'].rolling(window=6).mean()

# 5. Visualise
log.plot(x='time', y=['battery', 'smooth'], title='Robot battery level')
plt.ylabel('Battery %')
plt.show()

How it works

1. pd.date_range() makes 168 hourly timestamps. NumPy builds a battery level that drains through each day (the robot is recharged every 24 hours) plus a little random noise. The seed means you’ll get the same numbers every time
2. We set 10 random readings to NaN to pretend the sensor dropped out
3. interpolate() fills each gap with a value between its neighbours, which suits a smoothly changing reading like this
4. resample('D', on='time') groups the readings by day, and rolling(window=6).mean() averages each reading with the five before it, smoothing out the noise
5. plot() draws the raw and smoothed lines together

Projects to try

Financial data

Download historical prices for a stock or cryptocurrency as a CSV file. Load it with read_csv(), parse the dates, and plot the closing price with 20-day and 50-day rolling averages.

Survey data

Make a small CSV of survey answers with some blanks. Count missing values, fill or drop them, turn text answers like “Agree” into numbers with replace(), and use value_counts() to summarise each question.

Sales forecasting

Group a year of sales data by month with resample('MS'), plot the trend, and use a rolling average to estimate next month’s sales.

Your own sensors

Log readings from a real sensor, such as a Raspberry Pi Pico’s temperature sensor, to a CSV file. Then load and analyse it with everything you’ve learned.

Summary

This lesson worked through a complete analysis, from raw data to a chart, and suggested real-world projects to practise on. Real data is the best teacher, so pick a project and dive in!

< Previous Next >

You can use the arrows  ← → on your keyboard to navigate between lessons.