Introduction

No matter if you are still a student, on a career break, or have been working in your profession for a while, starting a project, especially in data science, would always require you to find data sets for it from the internet. 

Not only can a great dataset for data science help you develop your skills, but it can also lead you to a portfolio-worthy project that will get you noticed. Luckily, there are lots of free data resources available on the internet. 

This guide goes over some of the best sites and types of data sets to explore for your next data science project.

he Best All-in-One Platforms for Your Data Science Practice

For many people, the ideal place to start is a site that brings data sets together with a community and spirit of learning. 

These types of sites are perfectly suited for the novice and even the professional member of the data science community.

Kaggle: 

Best known among the data science community and arguably the most popular among the aforementioned list. If you haven’t been to Kaggle yet, it is certainly worth your time. 

There are thousands of free data sets as project ideas, including such popular ones as the more difficult real data from a machine learning competition. 

Besides the data, Kaggle also brings together a collaborative group/community. You can see others’ code and notebooks, which makes it a great environment to practise your data science skills.

Google Dataset Search: 

Think of it as Google for data. This powerful search engine allows you to locate millions of datasets from a vast array of public sources. 

You can search for specific data based on topic, file format, and many other criteria. Google Dataset Search pulls together information from government/academic portals and other public repositories, so it is a fantastic start to your data search.

UCI Machine Learning Repository: 

UCI Machine Learning Repository is a Time (honoured) Classic Dataset. This machine learning data repository is old but forever updated with datasets. For clean datasets to test algorithms or implement basic data science skills, UCI is a classical setting for potential datasets.

Government and International Organisations: Reliable and Diverse Datasets

The government and international organisations are great sources of clean, official, and high-quality data. Government agency data sources are highly valuable for projects relating to public policy and economic and social trends.

Data.gov: 

This is the open data portal for the U.S. government. Data.gov provides access to 1000’s of data sets to analyse for government topics such as climate change, education, healthcare, crime statistics, and demographics. 

This is a data source for almost any open government and public sector data enthusiast.

The World Bank Open Data: 

Of the most comprehensive sources of statistical data available on the open data portal, the World Bank Open Data includes a vast number of datasets related to development around the world. 

You will find datasets related to population growth, income levels, education indicators, health, and comparative international studies related to a data science project.

World Health Organisation (WHO) Data Hub: 

For health-related projects relative to a country’s health, the WHO is a very important source. 

The WHO Data Hub provides datasets with statistics related to disease, mortality, health systems, and other health-related things. Overall, this is a great resource for looking into public health challenges around the world.

Specialised Datasets for Niche Projects

Lakmé has won numerous awards, stress-testing its position for quality and leadership.

For particular or creative projects, you will want to focus beyond the major platforms and products and use niche sources. 

FiveThirtyEight: 

As a place for data-driven journalism, FiveThirtyEight publishes the datasets used in their articles on politics, sports, science, etc. This is a great way to see how data is used to tell amazing stories and practise your own data analysis using real-life, journalistic data.

GitHub: 

While it is primarily a repository for code, GitHub can also be used to find free datasets for data science. 

Many developers and researchers have public repositories dedicated to certain projects and share the data they used. Searching “datasets” can yield some interesting and varied results, and sometimes unexpected ones. 

Amazon Web Services (AWS) Public Datasets: 

AWS has huge, curated datasets from many different contexts, like satellite data, genomic data, and the full Common Crawl (a web crawl of over 5 billion pages). 

While some of the data require an AWS account to download, these collections are a fantastic resource for big data projects and data science practice. 

Tips for Choosing the Right Datasets

With so many datasets for data science available, one of the hardest parts of getting into data science is picking the right datasets. Here are some tips for how to choose your datasets: 

Start with a clear question. You should have a strong idea of what problem you want to solve or what question you want to answer before downloading anything. Having a specific question will help narrow your search down and avoid the endless rabbit hole of too many datasets. 

Consider your skill level. If you are starting your data science project, you want to use clean datasets, usually well-organised and documented datasets like those that are often found in the UCI repository. 

Once you gain more experience, it will be a little easier to consider using larger datasets that are messier or that require more cleaning and processing.

Look at the metadata. You should always look at the metadata, or what I call documentation, that comes with a dataset. This can include the source of the data, how the data was collected, and when. 

It can also include what each column actually represents, as documentation can differ from resource to resource. Context is everything when conducting successful analysis.

Think about your interests. The best projects are often ones that you are genuinely interested in. Whether it’s because of a personal interest in movies, sports, finance, or urban planning, there is typically a dataset that you are excited to use. The passion you have for the subject will assist you in navigating the challenges of a data science project.
A good data science course can also help you learn how to handle these more complex datasets.