Discretization of continuous features

Discretization of continuous features

In statistics and machine learning, discretization refers to the process of converting or partitioning continuous attributes, features or variables to discretized or nominal attributes/features/variables/intervals. This can be useful when creating probability mass functions – formally, in density estimation. It is a form of binning, as in making a histogram.

Typically data is discretized into partitions of K equal lengths/width (equal intervals) or K% of the total data (equal frequencies).[1]

Some mechanisms for discretizing continuous data include:

  • Fayyad & Irani's MDL method[2] - Uses Information Gain to recursively define the best bins.
  • And many more[3]

Many Machine Learning algorithms are known to produce better models by discretizing continuous attributes[4]

See also


  1. ^ "Entropy and MDL Discretization of Continuous Variables for Bayesian Belief Networks". http://sci2s.ugr.es/keel/pdf/specific/articulo/IJIS00.pdf. Retrieved 2008-07-10. 
  2. ^ "Multi-Interval Discretization of Continuous-Valued Attributes for Classification Learning". hdl:2014/35171. 
  3. ^ "Supervised and Unsupervised Discretization of Continuous Features". http://www.ifir.edu.ar/~redes/curso/disc.ps. Retrieved 2008-07-10. 
  4. ^ "S. Kotsiantis, D. Kanellopoulos, Discretization Techniques: A recent survey, GESTS International Transactions on Computer Science and Engineering, Vol.32 (1), 2006, pp. 47-58.". http://www.math.upatras.gr/~esdlab/en/members/kotsiantis/discretization%20survey%20kotsiantis.pdf.