Release Time:2019-03-11  Hits:

Indexed by: Conference Paper

Date of Publication: 2009-08-25

Included Journals: Scopus、EI

Page Number: 794-797

Abstract: For data sets of customer churn are imbalanced, it's hard for classical BP neural network to make good classification. This paper proposes an improved BP algorithm, namely CostBP. CostBP modifies the error function into costbased error function, based on which error propagation is conducted. The experimental results showed that CostBP can minimize total misclassification cost and was effective in prediction of minority class. ? 2009 IEEE.