A decision support proposal for imbalanced clinical data: Acute appendicitis cases

The difficult diagnosis of acute appendicits of patients appealing to the hospital with abdominal pain often leads to unnecessary acute appendicits operations. Accordingly, the aim of this study is to be able to provide the correct diagnosis whether the existing case indeed neccesiates operation or not through machine learning algorithms based on classification. To that purpose, SMOTE, Random Oversampling and Random Undersampling methods were proposed to reduce the negative effects of common imbalanced data set problem on classification and it was benefitted from the risk factors in relation to Alvarado Score to predict the diagnosis of acute appendicits. Therefore, the result of histopatologic report which is accepted as a gold standart for the diagnosis acute appendicits was also compared with Alvarado Score. Additionally, different classification models were generated by using classification algorithms like Support Vektor Machine, Random Forest, Artificial Neural Network, Naive Bayes, k-Nearest Neighbor and Logistic Regression. Consequently, a decision support system was developed that could contribute to the decision making mechanism by generating interface for the Support Vektor Machine algorithm in which the best performance was obtained and a different perspective was provided with the Alvarado Score.