Early Forecasting of Chronic Liver Disease from Liver Function Test Imbalance Datasets

Abstract

The primary goal of this study is to develop a model for predicting chronic liver disease in its early stages from datasets containing Liver Function Test (LFT) imbalance results, which will aid practitioners in accurately diagnosing liver disease. Detecting disease in its early stages can be difficult, as practitioners often struggle to predict the disease due to its ambiguous symptoms. A total of two data sets were used in this analysis,the second dataset (Primary) was obtained from the Karnataka region of India, and the first dataset (secondary) was taken from the UCI repository. To balance the datasets, we used the Random Forest and K-Nearest Neighbour’s (KNN) algorithms, as well as the Synthetic Minority Oversampling Technique (SMOTE). On both the imbalanced and balanced datasets, as well as the various parameters, we compared the effects of the two algorithms. Random forest outperforms KNN in terms of accuracy, specificity, precision, and false positive rate (FPR) on balanced datasets, while KNN outperforms Random forest in terms of accuracy, specificity, sensitivity, FPR, and FNR parameters.On the majority of parameters, the proposed system is expected to increase the balance dataset's performance. The suggested system is as follows: the balance dataset provides a stronger result for the majority of the parameters. The proposed approach aids physicians in correctly diagnosing liver disease at an early stage.