作者: Fatih Kayaalp , Muhammet Sinan Basarslan , Kemal Polat
DOI: 10.1109/IDAP.2018.8620935
关键词:
摘要: Attribute selection has a significant effect on the performance of machine learning studies by selecting attributes having result, reducing number attributes, and calculation cost. In this study, new attribute method which is combination R-correlation coefficient-based (RCBAS) ρ-correlation (ρCBAS) called Two-Stage Correlation-Based Selection (TSCBAS) proposed to select attributes. The been applied customer churn prediction telecommunications dataset for evaluation. used in study includes real call records details years 2013 2014 obtained from major company Turkey. Apart method, four different methods named Rcorrelation selection, ReliefF, Gain Ratio have creating five datasets. After that, classifier algorithms including Random Forest, C4.5 Decision Tree, Naive Bayes AdaBoost.M1 applied. results compared according metrics comprising Accuracy (ACC), Sensitivity (TPR), Specificity (SPC), F-measure (F), AUC (area under ROC curve), run-time. comparisons show that algorithm outperforms state art prediction.