Value prediction techniques aim to predict the value of one or more dependent variables based on the values of one or more predictor variables. They are typically trained using data items for which both predictor and dependent variables are known and then used for data items for which only the predictor variables are known.
The term regression is sometimes used to refer to value prediction functionality. However, the term is avoided here. Regression actually refers to a specific group of algorithms, most classically least squares regression. Furthermore, the group of regression algorithms contains elements like logistic regression that actually have a classification rather than a value prediction function.
Value prediction is captured both as a functional building block and as a use case as it has both functional and business aspects.
- alias
- used by
- ALG_Bayesian linear regression ALG_Decision tree ALG_Least Squares Regression ALG_Local regression ALG_Long short-term memory network ALG_Multivariate adaptive regression splines ALG_Random forest SPT_Bagging SPT_Boosting SPT_Elastic net SPT_Evolutionary selection SPT_LASSO SPT_Locally weighted learning SPT_Ridge regression SPT_Stacking