Liniowy model prawdopodobieństwa z regularyzacją jako metoda doboru zmiennych
Regularized linear probability model as a filter

Summary/Abstract: The application of regularized linear probability model as a filter which precedes the logistic regression is proposed in this paper. Additionally the cross-validation is applied in the feature selection stage. Such an approach guaranties more efficient elimination of the irrelevant variables than commonly used regularized logistic regression and classification errors of compared methods do not differ significantly. The datasets from UCI Repository were used in empirical study and noisy variables were generated from Bernoulli or normal distributions.

• Details
• Contents