In AI, we can classify data according to how it is structured, labeled, type or purpose
Data classification by structure
Structured Data
- It’s organized in rows and columns (like spreadsheets/databases)
- It’s easy to process with traditional Machine Learning algorithms
- Some formats are CSV files, SQL databases, Excel sheets
- Examples: Customer records, financial transactions, sensor readings.
Semi-structured data
- It has some type of organization but not rigid table format
- It may needs to be parsed, but contains valuable structured elements.
- Some formats are Key-value pairs, nested structures
- Examples JSON files, XML, web logs, emails and metadata.
Unstructured Data
- Data which doesn’t have any predefined organization or format, just raw data.
- It requires specialized processing, like NLP or Computer Vision
- Some examples are text documents, images, videos or audio files.
- Up to 80% of the enterprise data is unstructured.
Data classification by presence of labels
Depending on the type of Learning we are applying data can have labels or not, for instance:
Labeled Data
- It’s used on Supervised Learning to identify the “correct answer” or target outputs that you want your model to learn to predict.
- Examples: A photo of an animal, has a label “dog” or “cat”, so the model can predict the correct label for new photos.
Unlabeled data
- It’s used in Unsupervised Learning when the data has no label to tell the model the “correct answer”
- Example: you group similar images and the goal is the model can find patterns
Data Type or Formats
Depending on the AI application, data can be presented on different types and formats, or a combination of them
Numerical Data
- Data can be Continuous (Like temperature or height)
- Data can be Discrete (number of customers, age in years)
Categorical Data
- It can be Nominal: without natural order (color, nationality, etc.)
- It also can be Ordinal, like sizes, ratings, age.
Text Data
- Examples: reviews, documents, social media posts, emails
- It requires tokenisation, cleaning, vectorization for AI models
Tabular Data
- Different type of data (numerical, categorical, text, etc..) can be presented in form of a table.
Time-series Data
- Data points collected over time intervals
- Some examples are: weather data, website traffic, room capacity over time
- In this type of data, the order matters
Media Data
- Includes:
- Images: Pixel values, requires Computer Vision techniques
- Audio: Sound waves, requires signal processing
- Video: Combines images and audio over time