A dataset is a large file of organized (structured) or unorganized data containing everything from text and numbers to images, video and sound. As a general rule, datasets contain enormous amounts of data to perform data analysis and extract patterns (a branch of big data ) or train Artificial Intelligence. However, some data sets are more significant than others.
When a dataset is organized coherently, it greatly facilitates the analysis and understanding process.
Components of a Dataset
In addition to data, we can find the following elements in a structured dataset.
- Rows: These are the basic groups in which the data is organized. For example, if we have a dataset with customer information, each row could represent a specific customer. Or, if we have a dataset that records sales, each row could represent a particular transaction. In short, a row is each of the entries in a database.
- Columns: These sections make up a row where we find their characteristics or information. Continuing with the customer example, each column would house information about the customer, such as their name, age, and purchase history… Similarly, in the sales example, each column would indicate the characteristics of that transaction, such as the time and day. What has been done, and what did it cost? In short, the columns are the attributes of each of the entries.
- Values are the data that we find within each row and column, which can be in different formats.
Dataset Types
Types of data sets according to their format
- Numerical: It only contains data in numbers, and you can do quantitative and statistical analysis. That is why it is used primarily in science, statistics and finance.
- Text: In this case, the information is in word and character format and is mainly used to train natural language models and to develop machine translation tools. Within this data set, we can find studies, news, reviews, publications on social networks, articles, blogs, and forums. They are the ones most within reach of the standard user, and many can be found in public online repositories.
- Video and image: As their name indicates, they contain data in video and image format, serving mainly to train computer systems in charge of interpreting and analyzing pictures or videos as well as identifying patterns within them; in short, what is known as computer vision models.
Types of data sets according to their structure
Tables (structured data set)
They are the most common and have the advantage that they are intuitive and easy to understand so that users can use them without high technical knowledge. Relational databases and spreadsheets are examples of structured data sets.
On the other hand, they allow efficient and fast analysis. They are also used in various sectors, such as marketing and finance.
Unstructured Dataset
The data is disorganized, making it more challenging to process and analyze. A perfect example of an unstructured data set would be emails within the email.
Like structured data sets, within this type, we can also encompass different datasets depending on their format.
Where can I find Datasets?
First, you should know that anyone can create a data set by storing data and information digitally. However, some users decide to publish them (autonomously or because it is part of their job) so that the public can access them.
In that sense, we can find public (free) or private data sets.
Any user can access public data sets, and they can be found on specific platforms such as Google Data Search or FiveThirtyEight. The first is the largest online dataset search engine regarding company information. The second houses extensive data on politics, sports and global surveys. Both are reliable; you can use them for free when working on your projects.
For their part, private data sets are usually purchased by private companies or organizations. Because the data is not public, special care must be taken with its privacy when storing and processing it, as it is usually the target of hackers—cyber attacks.
Within private data sets, we also find susceptible government data that is not in the public domain; therefore, not everyone can access it.
