What is a bucketing?

Converting a single feature into multiple binary features called buckets or bins, typically based on a value range. The chopped feature is typically a continuous feature. For example, instead of representing temperature as a single continuous floating-point feature, you could chop ranges of temperatures into discrete buckets, such as: - <= 10 degrees Celsius would be the "cold" bucket. - 11 - 24 degrees Celsius would be the "temperate" bucket. - >= 25 degrees Celsius would be the "warm" bucket. The model will treat every value in the same bucket identically. For example, the values`13` and`22` are both in the temperate bucket, so the model treats the two values identically.

If you represent temperature as a continuous feature, then the model treats temperature as a single feature. If you represent temperature as three buckets, then the model treats each bucket as a separate feature. That is, a model can learn separate relationships of each bucket to the label. For example, a linear regression model can learn separate weights for each bucket. Increasing the number of buckets makes your model more complicated by increasing the number of relationships that your model must learn. For example, the cold, temperate, and warm buckets are essentially three separate features for your model to train on. If you decide to add two more buckets--for example, freezing and hot--your model would now have to train on five separate features. How do you know how many buckets to create, or what the ranges for each bucket should be? The answers typically require a fair amount of experimentation. --- See Numerical data: Binning in Machine Learning Crash Course for more information.

Practitioners refer to bucketing when building, training, or evaluating machine learning systems. It appears in research papers, product documentation, and technical discussions about AI capabilities and limitations.