Implementation of ADABOOST Algorithm on C50 for Improving the Performance of Liver Disease Classification

Abstract

The liver is a vital organ that is important for humans because it plays a role in regulating hormone cycles, neutralizing toxins, and controlling the composition of blood. Liver disease is a common ailment worldwide. Often, this disease occurs without specific symptoms (asymptomatic). Therefore, liver disease is known as a "silent killer," and it is necessary to quickly and accurately diagnose and treat liver diseases. Data mining technology can be useful for rapidly detecting liver diseases from laboratory diagnosis results. One classification algorithm that can be used is the C50 algorithm. This algorithm is an improvement over the previous C45 algorithm, with several advantages such as efficient memory usage and more concise tree results. However, the C50 algorithm may experience overfitting on complex medical data, requiring the boosting process using AdaBoost. The AdaBoost algorithm can make the C50 algorithm more susceptible to overfitting. Another advantage of the AdaBoost algorithm is its ability to handle imbalanced datasets in terms of target labels. This research used 583 data from UCI machine learning with two target labels: liver and normal. The research results show that the C50 algorithm is capable of identification with an accuracy rate of 74.58%. Furthermore, the C50 algorithm's accuracy can be maximized by AdaBoost to reach 86.44%