In the present study, Wavelet analysis of P wave for the prediction of Atrial Fibrillation after CABG is evaluated. Continuous Wavelet Transform is applied to ECG and Mean/Max parameters are calculated within the P window for different frequency bands. Thus, 24 parameters are available, which, along with the 4 window lengths (corresponding to P wave length of X, Y, Z, V signals), make a pool of available parameters to be used for classification. Linear regression is used for the classification of the two groups and bootstrapping is applied in order to enhance statistical robustness. The features to be used in the regression model are selected from the pool of available parameters by use of an iterative procedure. The outcome of the feature selection procedure shows that X and Z-axis features as well as vector-magnitude features are the most important ones for the prediction of Atrial Fibrillation after CABG.