Seybold Report ISSN: 1533-9211
K.Sindhya, M.Suganya, S.Santhana Megala
Vol 18, No 6 ( 2023 ) | Licensing: CC 4.0 | Pg no: 298-310 | Published on: 24-06-2023
Abstract
The liver serves as the body's primary organ for detoxifying toxins, which makes it crucial for ensuring survival. The liver may be harmed if it contracts a virus, is the target of the body's immunological response, or is exposed to toxins. The burden of liver illness in the nation is substantial given that India alone accounted for 18.3% of the two million deaths brought on by liver disease worldwide in 2015. Early discovery and the proper therapy at the appropriate time are the only ways to solve the problem. Clinical results are more dependent on data than models. It can be particularly challenging to identify the appropriate target (response variable) and features for classification problems in medical diagnostics. Another common problem in real-world data science applications is missing values in a data set. For handling the missing values two hybrid strategy were proposed, MSMOTE and MFSMOTE employing imputation algorithm along with SMOTE. As evaluation measures, confusion matrix, precision, recall, and f1-score were used. With an accuracy of 87.80% in predicting disease and 11.38% in predicting no disease, Extra Tree - MFSMOTE does well overall.
Keywords:
Imputation Techniques, Missing Values, Preprocessing SMOTE, Liver Disease.