Diagnostic Predictive Approaches for Liver Disease Detection using Stacked Ensemble Model with Data Augmentation

The global medical fraternity is challenged with evolving a perfect prediction model that diagnoses a liver ailment at the right time and calls for an immediate medical intervention for its critical need. Potential threat to life of liver disease chronically contracted stresses for necessity for its root cause and time-bound medical remedy. By making use of a dataset of Indian liver patients from the pool of data maintained at the national level, this work is entailed with the introduction of an architecture of innovative nature that desegregates Stacked Ensemble Model, feature engineering for foretelling liver ailments. The core contribution to this work is the harness of feature engineering utilizing SHapely Additive exPlanations (SHAP), encompassing the state-of-the-art techniques that are not delved into by data-driven machine learning approaches that are in vogue now as a vaticinator of liver ailment. The labyrinthine design of Stacked Ensemble Model facilitates timely prediction of liver disease, exploiting the maximum of learning techniques that ramp up the detection activity and augment accurate diagnosing. The methodology exploits the potencies of varied base learners like Logistic Regression, Multi-layer Perceptron, Support Vector Machine, Decision Tree, K-nearest neighbor and Extra trees classifier. These diverse views are homogenized as input to the meta-learner, the Random Forest, to build a sturdy and trustworthy predictive model. The commissioning of Stacked Ensemble Model with feature engineering produced an accuracy of a cent percent short of 3. This methodology is aimed at fine tuning the prediction of liver ailments, ensuring smooth, effective and timely intervention, and assuring of an effective management.