Abstract:
In India, liver disease is one of the major causes of deaths in all age groups and moreover its resistance to early detection is more disturbing fact. It is therefore important to have a decision support system for early prediction of liver disease so that proper medical assistance can be given to patients. A number of researches have been done to address this area. However, the current literature fails to offer such a system, which can make such similar kind of early prediction. Hence, this research propose8s a Two-stage CBR (TSCBR ) model to enhance the accuracy of prediction of liver disease. Case Based Reasoning (CBR) is added to 5 different classification models including neural network (NN), decision tree based Random Forest (RF), logistic regression (LR), Support Vector Machine (SVM) and naive Bayes (NB) to enhance the improvement on the correct diagnosis of disease. The main focus of this research is to reduce the number of cases that are wrongly classified as not having liver disease. Every new case passes through two stages of proposed model to see whether the new case belongs to liver disease or not. To analyze the accuracy of the proposed research two stage CBR model is compared with already existing models to diagnose liver disease without CBR. 10-fold cross-validation approach is used to reduce biasness and to analyze the performance of proposed system. It was observed that naive Bayes with CBR model shows the highest accuracy of 0.83 among all the existing compared classification technique. The proposed technique decreases the number of false negative cases. Thus, resulting in enhanced number of early detections of liver cases. It has been observed that the accuracy of TSCBR is better than the existing system.
