Data Mining · CSC213

Data Reduction: Sampling, Aggregation, Dimensionality Reduction and Compression

Data Mining Study guides

Data reduction produces a smaller representation of the data that gives the same or almost the same analytical results.

Dimensionality reduction is the process of reducing the number of random variables or attributes under consideration. Dimensionality means the number of attributes of the data. Too many dimensions make data sparse and distances less meaningful (the curse of dimensionality). Techniques include feature selection, feature extraction such as PCA, and wavelet transforms.

Numerosity reduction replaces the original data volume by alternative, smaller forms of data representation: sampling (random sampling picks a subset, with or without replacement), histograms, clustering, and parametric models. Aggregation combines two or more attributes or objects into one (for example daily sales into monthly sales), which gives a higher-level view and less variability.

Data compression applies transformations to get a reduced representation. If the original data can be reconstructed from the compressed data without any loss of information, the compression is lossless; if only an approximation can be reconstructed, it is lossy. Attribute construction creates new attributes from existing ones, which can improve accuracy and understanding of structure in high-dimensional data.

Key points

  • Dimensionality reduction: fewer attributes (feature selection, PCA).
  • Sampling: choose a representative subset.
  • Aggregation: combine into higher-level summaries.
  • Lossless compression: original fully recoverable; lossy: only approximate.
  • Attribute construction builds new attributes from old ones.

Worked example

Store sales per transaction for 5 years is huge. Aggregate to daily totals per store, or sample 10 percent of transactions at random, and trend analysis still works with a fraction of the storage.

Common questions

What is lossless compression?

Compression where the original data can be reconstructed exactly from the compressed data.

What is the curse of dimensionality?

As the number of attributes grows, data becomes sparse and distance-based methods lose meaning.

Practice questions on this topic

Try these, then open the app to check your answers with explanations.

  1. Where data are transformed or consolidated into forms appropriate for mining by performing summary or aggregation operations
    • A. Data transformation
    • B. Data mining
    • C. Data knowledge
    • D. Data integration
  2. If the original data can be reconstructed from the compressed data without any information loss, the data reduction is called lossless
    • A. False
    • B. True
  3. Attribute construction can help improve accuracy and understanding of structure in high dimensional
    • A. False
    • B. True
  4. The process of reducing the number of random variables or attributes under consideration.
    • A. Numerosity reduction
    • B. Dimensionality reduction
    • C. Random variables reduction
    • D. Filtering of random variables
  5. Aggregation in data mining involves:
    • A. Combining multiple datasets into one
    • B. Reducing the number of attributes in a dataset
    • C. Removing outliers from the dataset
    • D. Transforming continuous variables into discrete categories
  6. Random sampling is a data transformation technique used for:
    • A. Balancing imbalanced datasets
    • B. Removing noisy data points from the dataset
    • C. Reducing the size of a large dataset
    • D. None of the above