Skip to main content

Sorting Data

In this tutorial, we'll explore how to efficiently sort data in a Pandas DataFrame using the nlargest and nsmallest functions. These functions are particularly useful when you need to identify the largest or smallest values in a DataFrame based on one or multiple columns.

1. Importing Necessary Libraries and Loading Data​

First, let's import the Pandas library and load our dataset. We'll specify the columns we're interested in, which are the series title, rating, and the number of votes.

import pandas as pd

# Define columns of interest
columns = ['title', 'rating', 'num_votes']

# Load dataset
df = pd.read_csv('your_dataset.csv', usecols=columns)

Replace 'your_dataset.csv' with the path to your dataset file.

2. Sorting with nlargest​

2.1 Sorting by a Single Column​

To sort the DataFrame by a single column, such as the number of votes, we use the nlargest function.

# Sorting by the number of votes (top 10)
top_10_votes = df.nlargest(10, 'num_votes')
print(top_10_votes)

2.2 Sorting by Multiple Columns​

To sort by multiple columns, such as rating and number of votes, we pass a list of columns to the nlargest function.

# Sorting by rating and then by number of votes
top_rated = df.nlargest(10, ['rating', 'num_votes'])
print(top_rated)

3. Sorting with nsmallest​

Sorting with nsmallest works similarly to nlargest, but it retrieves the smallest values instead.

# Getting the bottom 10 smallest values for number of votes
bottom_10_votes = df.nsmallest(10, 'num_votes')
print(bottom_10_votes)

4. Handling Duplicates with the keep Parameter​

The keep parameter determines how duplicates are handled when retrieving the largest or smallest values.

# Keeping all duplicates for the top rated movies
all_duplicates = df.nlargest(10, 'rating', keep='all')
print(all_duplicates)

5. Using sort_values as an Alternative​

Though less efficient, you can also use sort_values to achieve similar results.

# Sorting using sort_values (less efficient)
sorted_df = df.sort_values(by=['rating', 'num_votes'], ascending=False).head(10)
print(sorted_df)