Preprocessing Using Correlation Based Features Selection on Naive Bayes Classification

Abstract

The birth of a healthy baby is generally around 38-42 weeks of pregnancy. However, there are many babies born at an inadequate age of birth and the age of birth that is past its time. This study aims to predict the age of the birth of a patient. The method used is a classification with the Naïve Bayes algorithm with input variable (X), the factors experienced by pregnant women in the form of 8 variables X and Y variable in the form of Birth Age. Problems that arise are too many attributes that affect the results of accuracy. To overcome this, preprocessing is used with the Correlation Based Features Selection (CBFS) method. CBFS chose the X variables which had the highest correlation with the Y variable (Birth Age) but had the least correlation between the X variables. From the CBFS that had been done, produced 4 X variables, namely: blood pressure, number of babies, congenital diseases before pregnancy, and problems during pregnancy. The results of the test showed an increase in Precision, recall, and accuracy in the Naïve Bayes classification when implemented CBFS. The highest value of accuracy after preprocessing is 67% with an increase of 2 percent compared to before preprocessing.