ML Notes

Datasets for ML Practice

Collection of publicly available datasets for practicing machine learning algorithms and building projects.

A curated collection of datasets for learning and practicing ML algorithms.

Beginner-Friendly Datasets

Iris Dataset

Type: Classification | Samples: 150 | Features: 4

Famous flower classification dataset.

python example

from sklearn.datasets import load_iris
iris = load_iris()
X, y = iris.data, iris.target

Titanic Dataset

Type: Classification | Samples: 891 | Features: 11

Predict passenger survival. Great for feature engineering.

python example

import pandas as pd
titanic = pd.read_csv('titanic.csv')

Source: Kaggle Titanic

Wine Dataset

Type: Classification | Samples: 178 | Features: 13

Classify wine quality. Good for multi-class classification.

python example

from sklearn.datasets import load_wine
wine = load_wine()

Breast Cancer Dataset

Type: Classification | Samples: 569 | Features: 30

Medical diagnosis prediction.

python example

from sklearn.datasets import load_breast_cancer
cancer = load_breast_cancer()

Intermediate Datasets

Boston Housing

Type: Regression | Samples: 506 | Features: 13

Predict house prices (classic regression dataset).

python example

from sklearn.datasets import load_boston
boston = load_boston()
# Or: df = pd.read_csv('housing.csv')

MNIST Digits

Type: Image Classification | Samples: 70,000 | Features: 784

Handwritten digit recognition (0-9).

python example

from keras.datasets import mnist
(X_train, y_train), (X_test, y_test) = mnist.load_data()

Fashion MNIST

Type: Image Classification | Samples: 70,000 | Features: 784

Clothing item classification (10 categories).

python example

from keras.datasets import fashion_mnist
(X_train, y_train), (X_test, y_test) = fashion_mnist.load_data()

Iris-like Synthetic Data

Type: Classification | Samples: Configurable

Generate synthetic data for testing.

python example

from sklearn.datasets import make_classification, make_blobs

# Classification
X, y = make_classification(n_samples=1000, n_features=20, n_classes=3)

# Clustering
X, y = make_blobs(n_samples=300, centers=3, n_features=2)

Kaggle Competitions & Datasets

Popular Kaggle Datasets

Name	Type	Difficulty	Use Case
Titanic	Classification	Beginner	Feature engineering, baseline models
Iris	Classification	Beginner	Algorithm comparison
Housing	Regression	Beginner	Regression basics
Digit Recognition	Image	Intermediate	Deep learning, CNNs
Titanic	Survival	Intermediate	NLP features
Rossmann Store Sales	Time Series	Advanced	Forecasting, feature engineering
Airbnb Listings	Mixed	Intermediate	EDA, embeddings
MovieLens	Recommendation	Intermediate	Collaborative filtering

Access via: https://www.kaggle.com/datasets

UCI Machine Learning Repository

Repository of 600+ datasets: https://archive.ics.uci.edu/

Notable Datasets

Car Evaluation: Classification of cars

Example

Samples: 1,728 | Features: 6 | Classes: 4 Use: Decision trees, random forests

Mushroom: Edible/poisonous classification

Example

Samples: 8,124 | Features: 22 | Classes: 2 Use: Classification with categorical features

Abalone: Age prediction

Example

Samples: 4,177 | Features: 8 | Classes: 28 Use: Regression, ordinal classification

Adult Income: >50K income prediction

Example

Samples: 48,842 | Features: 14 | Classes: 2 Use: Feature engineering, imbalanced classification

NLP Datasets

Movie Reviews (IMDb)

Type: Text Classification | Samples: 50,000

Sentiment analysis dataset.

python example

from keras.datasets import imdb
(X_train, y_train), (X_test, y_test) = imdb.load_data()

20 Newsgroups

Type: Text Classification | Samples: 18,846

News article classification across 20 categories.

python example

from sklearn.datasets import fetch_20newsgroups
newsgroups = fetch_20newsgroups(subset='train')

Common Crawl

Type: Text Corpus | Size: 1TB+

Massive web text corpus for NLP research.

Source: https://commoncrawl.org/

WikiText

Type: Language Modeling | Size: 4GB

Wikipedia text for language model training.

Computer Vision Datasets

CIFAR-10 & CIFAR-100

Type: Image Classification | Samples: 60,000

10 or 100 object categories.

python example

from keras.datasets import cifar10
(X_train, y_train), (X_test, y_test) = cifar10.load_data()

ImageNet

Type: Image Classification | Samples: 14M

Massive image dataset (1,000 classes).

Source: http://www.image-net.org/

COCO (Common Objects in Context)

Type: Object Detection | Samples: 330K

Images with object annotations.

Source: https://cocodataset.org/

Pascal VOC

Type: Object Detection | Samples: 11K

Visual Object Classes dataset.

Source: http://host.robots.ox.ac.uk/pascal/VOC/

Time Series Datasets

Stock Market

Type: Time Series | Frequency: Daily/Hourly

Historical stock prices.

python example

import yfinance as yf
data = yf.download('AAPL', start='2020-01-01')

Weather Data

Type: Time Series | Frequency: Hourly

Historical weather observations.

Source: https://www.ncei.noaa.gov/products/

Energy Consumption

Type: Time Series | Frequency: Hourly

Electricity usage patterns.

AirBnB Pricing

Type: Time Series + Regression | Samples: 40K+

Time-varying price predictions.

Recommendation System Datasets

MovieLens

Type: Ratings | Ratings: 25M

Movie ratings dataset.

python example

# Download from: https://grouplens.org/datasets/movielens/
df = pd.read_csv('ratings.csv')

Last.fm

Type: Music Ratings | Interactions: 19M

Music listening patterns.

Jester

Type: Joke Ratings | Ratings: 5M

Rating jokes (for CF research).

Medical/Healthcare Datasets

Heart Disease

Type: Classification | Samples: 4,207

Predict presence of heart disease.

Diabetes (Pima Indians)

Type: Classification | Samples: 768

Diabetes prediction.

python example

from sklearn.datasets import load_diabetes
diabetes = load_diabetes()

COVID-19 Data

Type: Time Series | Daily Updates

Real-time case data.

Source: https://github.com/CSSEGISandData/COVID-19

Where to Find Datasets

Public Repositories

Site	Focus	Size
Kaggle	Competitions + general	100K+
UCI ML	Scientific + classic	600+
GitHub	Curated collections	Varies
Google Dataset	Search engine	Millions
AWS Open Data	Cloud-ready	Varies
Hugging Face	NLP/Vision	1000+

Academic Repositories

Stanford Large Network Dataset (SNAP)
UC Irvine ML Repository
Carnegie Mellon StatLib
Harvard Dataverse

Creating Your Own

If public datasets don't fit your needs:

Example

1. Web scraping (with permission) 2. APIs (Twitter, Reddit, etc.) 3. Collect from sensors/logs 4. Surveys and user studies 5. Synthetic data generation

Summary

Recommended learning path:

Start: Iris, MNIST, Titanic
Progress: Boston Housing, Fashion MNIST
Advanced: ImageNet, COCO, Kaggle competitions
Specialized: Domain-specific datasets

Always:

Understand data before modeling
Check for bias and imbalance
Cite data sources properly

Exam Focus

Revise definitions, diagrams, examples, and short-answer points for Datasets for ML Practice.

Interview Use

Prepare one clear explanation, one practical example, and one common mistake for this Machine Learning topic.

Search Terms

machine-learning, machine learning, machine, learning, resources, datasets, for, practice

Continue learning this concept

ResourcesImportant ML FormulasCollection of essential mathematical formulas used in machine learning algorithms and evaluation.ResourcesMachine Learning GlossaryComprehensive glossary of machine learning terms, definitions, and concepts.ResourcesRecommended ML Books and CoursesCurated list of best machine learning books, courses, and learning resources for all skill levels.Interview PreparationStatistics for Machine Learning InterviewsEssential statistics concepts for ML interviews: hypothesis testing, distributions, sampling, Bayesian thinking, and probability theory.MlopsCI/CD for Machine LearningImplementing continuous integration and continuous deployment pipelines for ML models including testing, validation, and automated deployment.MlopsDocker for Machine LearningContainerizing ML models and workflows using Docker for reproducibility, scalability, and deployment.

PreviousStatistics for Machine Learning Interviews NextImportant ML Formulas

ML Notes

Datasets for ML Practice

Collection of publicly available datasets for practicing machine learning algorithms and building projects.

A curated collection of datasets for learning and practicing ML algorithms.

Beginner-Friendly Datasets

Iris Dataset

Type: Classification | Samples: 150 | Features: 4

Famous flower classification dataset.

python example

from sklearn.datasets import load_iris
iris = load_iris()
X, y = iris.data, iris.target

Titanic Dataset

Type: Classification | Samples: 891 | Features: 11

Predict passenger survival. Great for feature engineering.

python example

import pandas as pd
titanic = pd.read_csv('titanic.csv')

Source: Kaggle Titanic

Wine Dataset

Type: Classification | Samples: 178 | Features: 13

Classify wine quality. Good for multi-class classification.

python example

from sklearn.datasets import load_wine
wine = load_wine()

Breast Cancer Dataset

Type: Classification | Samples: 569 | Features: 30

Medical diagnosis prediction.

python example

from sklearn.datasets import load_breast_cancer
cancer = load_breast_cancer()

Intermediate Datasets

Boston Housing

Type: Regression | Samples: 506 | Features: 13

Predict house prices (classic regression dataset).

python example

from sklearn.datasets import load_boston
boston = load_boston()
# Or: df = pd.read_csv('housing.csv')

MNIST Digits

Type: Image Classification | Samples: 70,000 | Features: 784

Handwritten digit recognition (0-9).

python example

from keras.datasets import mnist
(X_train, y_train), (X_test, y_test) = mnist.load_data()

Fashion MNIST

Type: Image Classification | Samples: 70,000 | Features: 784

Clothing item classification (10 categories).

python example

from keras.datasets import fashion_mnist
(X_train, y_train), (X_test, y_test) = fashion_mnist.load_data()

Iris-like Synthetic Data

Type: Classification | Samples: Configurable

Generate synthetic data for testing.

python example

from sklearn.datasets import make_classification, make_blobs

# Classification
X, y = make_classification(n_samples=1000, n_features=20, n_classes=3)

# Clustering
X, y = make_blobs(n_samples=300, centers=3, n_features=2)

Kaggle Competitions & Datasets

Popular Kaggle Datasets

Name	Type	Difficulty	Use Case
Titanic	Classification	Beginner	Feature engineering, baseline models
Iris	Classification	Beginner	Algorithm comparison
Housing	Regression	Beginner	Regression basics
Digit Recognition	Image	Intermediate	Deep learning, CNNs
Titanic	Survival	Intermediate	NLP features
Rossmann Store Sales	Time Series	Advanced	Forecasting, feature engineering
Airbnb Listings	Mixed	Intermediate	EDA, embeddings
MovieLens	Recommendation	Intermediate	Collaborative filtering

Access via: https://www.kaggle.com/datasets

UCI Machine Learning Repository

Repository of 600+ datasets: https://archive.ics.uci.edu/

Notable Datasets

Car Evaluation: Classification of cars

Example

Samples: 1,728 | Features: 6 | Classes: 4 Use: Decision trees, random forests

Mushroom: Edible/poisonous classification

Example

Samples: 8,124 | Features: 22 | Classes: 2 Use: Classification with categorical features

Abalone: Age prediction

Example

Samples: 4,177 | Features: 8 | Classes: 28 Use: Regression, ordinal classification

Adult Income: >50K income prediction

Example

Samples: 48,842 | Features: 14 | Classes: 2 Use: Feature engineering, imbalanced classification

NLP Datasets

Movie Reviews (IMDb)

Type: Text Classification | Samples: 50,000

Sentiment analysis dataset.

python example

from keras.datasets import imdb
(X_train, y_train), (X_test, y_test) = imdb.load_data()

20 Newsgroups

Type: Text Classification | Samples: 18,846

News article classification across 20 categories.

python example

from sklearn.datasets import fetch_20newsgroups
newsgroups = fetch_20newsgroups(subset='train')

Common Crawl

Type: Text Corpus | Size: 1TB+

Massive web text corpus for NLP research.

Source: https://commoncrawl.org/

WikiText

Type: Language Modeling | Size: 4GB

Wikipedia text for language model training.

Computer Vision Datasets

CIFAR-10 & CIFAR-100

Type: Image Classification | Samples: 60,000

10 or 100 object categories.

python example

from keras.datasets import cifar10
(X_train, y_train), (X_test, y_test) = cifar10.load_data()

ImageNet

Type: Image Classification | Samples: 14M

Massive image dataset (1,000 classes).

Source: http://www.image-net.org/

COCO (Common Objects in Context)

Type: Object Detection | Samples: 330K

Images with object annotations.

Source: https://cocodataset.org/

Pascal VOC

Type: Object Detection | Samples: 11K

Visual Object Classes dataset.

Source: http://host.robots.ox.ac.uk/pascal/VOC/

Time Series Datasets

Stock Market

Type: Time Series | Frequency: Daily/Hourly

Historical stock prices.

python example

import yfinance as yf
data = yf.download('AAPL', start='2020-01-01')

Weather Data

Type: Time Series | Frequency: Hourly

Historical weather observations.

Source: https://www.ncei.noaa.gov/products/

Energy Consumption

Type: Time Series | Frequency: Hourly

Electricity usage patterns.

AirBnB Pricing

Type: Time Series + Regression | Samples: 40K+

Time-varying price predictions.

Recommendation System Datasets

MovieLens

Type: Ratings | Ratings: 25M

Movie ratings dataset.

python example

# Download from: https://grouplens.org/datasets/movielens/
df = pd.read_csv('ratings.csv')

Last.fm

Type: Music Ratings | Interactions: 19M

Music listening patterns.

Jester

Type: Joke Ratings | Ratings: 5M

Rating jokes (for CF research).

Medical/Healthcare Datasets

Heart Disease

Type: Classification | Samples: 4,207

Predict presence of heart disease.

Diabetes (Pima Indians)

Type: Classification | Samples: 768

Diabetes prediction.

python example

from sklearn.datasets import load_diabetes
diabetes = load_diabetes()

COVID-19 Data

Type: Time Series | Daily Updates

Real-time case data.

Source: https://github.com/CSSEGISandData/COVID-19

Where to Find Datasets

Public Repositories

Site	Focus	Size
Kaggle	Competitions + general	100K+
UCI ML	Scientific + classic	600+
GitHub	Curated collections	Varies
Google Dataset	Search engine	Millions
AWS Open Data	Cloud-ready	Varies
Hugging Face	NLP/Vision	1000+

Academic Repositories

Stanford Large Network Dataset (SNAP)
UC Irvine ML Repository
Carnegie Mellon StatLib
Harvard Dataverse

Creating Your Own

If public datasets don't fit your needs:

Example

1. Web scraping (with permission) 2. APIs (Twitter, Reddit, etc.) 3. Collect from sensors/logs 4. Surveys and user studies 5. Synthetic data generation

Summary

Recommended learning path:

Start: Iris, MNIST, Titanic
Progress: Boston Housing, Fashion MNIST
Advanced: ImageNet, COCO, Kaggle competitions
Specialized: Domain-specific datasets

Always:

Understand data before modeling
Check for bias and imbalance
Cite data sources properly

Exam Focus

Revise definitions, diagrams, examples, and short-answer points for Datasets for ML Practice.

Interview Use

Prepare one clear explanation, one practical example, and one common mistake for this Machine Learning topic.

Search Terms

machine-learning, machine learning, machine, learning, resources, datasets, for, practice

Continue learning this concept

PreviousStatistics for Machine Learning Interviews NextImportant ML Formulas