Kaggle — Competitions and Datasets

What Kaggle Actually Is

Kaggle is a website built around practicing and sharing data science work. It's free to join, and once you're in you'll mainly bump into four things: competitions you can enter, datasets anyone can browse or download, notebooks (hosted, runnable Jupyter notebooks other users publish), and a community of people discussing all of it. Owned by Google since 2017, it's the single most common place beginners go after a course like this one to keep practicing on data that isn't hand-built for them.

None of this replaces what you've learned in this course — pandas, Matplotlib/Seaborn, and scikit-learn are exactly what you'll use on Kaggle too. What Kaggle adds is a constant supply of new, real datasets and problems, plus a way to see how other people approached the exact same problem.

Finding a Beginner-Friendly Dataset or Competition

Kaggle hosts competitions at every level, from casual practice to serious prize money with strict rules. As a beginner, you want the ones Kaggle itself labels as "Getting Started" — they never close, award no prize money, and exist purely so newcomers have somewhere low-stakes to practice the full workflow. Two of them are practically rites of passage:

Learning From Public Notebooks

Every competition and dataset has a Code (formerly "Kernels") tab listing notebooks other users have published against it. Sort by "Most Votes" and you'll find well-explained walkthroughs near the top — often written specifically to teach, not just to score well.

import pandas as pd

# Kaggle notebooks can read competition files straight from
# the attached /kaggle/input/ directory — no manual download needed
train = pd.read_csv('/kaggle/input/titanic/train.csv')
test = pd.read_csv('/kaggle/input/titanic/test.csv')

print(train.shape, test.shape)
train.head()

Making Your First Submission

A "Getting Started" competition gives you three files: train.csv (features plus the answer, for training), test.csv (features only, for prediction), and sample_submission.csv (a template showing exactly the column names and row order your submission file must match). Your job is to predict the missing answer for every row in test.csv and format it exactly like the sample.

from sklearn.ensemble import RandomForestClassifier

# same pattern as the Student Performance Predictor in Lesson 30:
# clean -> engineer features -> train -> predict on the unseen test set
model = RandomForestClassifier(n_estimators=200, random_state=42)
model.fit(X_train, y_train)

predictions = model.predict(X_test)

# must match sample_submission.csv exactly: same columns, same row order
submission = pd.DataFrame({
    'PassengerId': test['PassengerId'],
    'Survived': predictions
})
submission.to_csv('submission.csv', index=False)

Kaggle's Notebook Environment Basics

Kaggle Notebooks are a browser-based, hosted Jupyter environment — you don't need Python installed locally to use them. They come with pandas, NumPy, Matplotlib, and scikit-learn already available, and they attach directly to any dataset or competition without a manual download.

Lesson Summary