Applicable model of liver disease detection based on the improved CART-AdaBoost algorithm

Abstract

Traditional diagnosis technology on earlier detection of some deadly liver diseases has many disadvantages. These shortcomings are due mainly to inadequate accuracy, which usually leads to failing to give liver patients timely treatment. In order to solve this problem, this paper used Classification and Regression Tree (CART) as a weak classifier of the AdaBoost framework to propose a Classification and Regression Tree-Adaptive Boosting (CART-AdaBoost) model. Moreover, the authors trained and verified the model basing on the Indian Liver Patient Dataset (ILPD) of UCI. The results showed that the model's accuracy was 83.06%, and its precision was 84.31%. Besides, F1-score could reach 80.75%, and the recall metric was 77.48%. All the former three indicators were higher than those produced by single models or combination models (weak classifier + AdaBoost) listed in this paper. Besides, it is worth noting that the prediction accuracy and precision of the CART-AdaBoost model were improved by a maximum value of 18.60% and 23.84%, respectively. Therefore, the suggested model is of great benefit in enhancing the early detection effect of liver diseases.