This paper considered a state of art as it employs a brand new technique to estimate the missing values in the dataset helping the classifiers to classify the data with better accuracy. also determining the effect of each attribute on the accuracy that enables researchers to get better or same accuracy with less number of attributes saving processing time, RAM, and memory needed. The aim of this paper is to improve accuracy of expecting Hepatitis mortality using worldwide dataset from Ljubljana University. We present an implementation of two brand new classification techniques. Using confusion matrix and K-fold cross validation technique to calculate classification accuracy. Two experiments have been done and the experimental results show that using the correlation in frequency domain after computing the weight factor for each attribute achieved better accuracy than using the subtraction method in time domain.