Optimizing Sensitivity and Specificity of Ensemble Classifiers for Diabetic Patients
Predicting the right class for a certain disease in the medical-related field is very critical. The effects of misclassification of the class could be very risky because it may lead to the mistreatment of the patient. The most important classification performance measurements in medical fields are s...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English English |
| Published: |
Asian Research Publishing Network
2015
|
| Subjects: | |
| Online Access: | http://eprints.unisza.edu.my/6664/1/FH02-FIK-15-03851.pdf http://eprints.unisza.edu.my/6664/2/FH02-FIK-16-04762.jpg http://eprints.unisza.edu.my/6664/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Predicting the right class for a certain disease in the medical-related field is very critical. The effects of misclassification of the class could be very risky because it may lead to the mistreatment of the patient. The most important classification performance measurements in medical fields are sensitivity, specificity and accuracy. This research aims to focus on the relationship between these three measurements. Misjudgements in classifying a person to a particular disease will prevent him/her from getting the correct treatment. Thus, the accuracy in classifying such medical data should be at the highest. Nevertheless, the most significant measurement is to have the highest sensitivity, because this will show that the classifier correctly classifies the patient who had a positive symptom of a particular disease. By using a single classifier, it is impossible to get the highest sensitivity. Thus, this paper proposed an ensemble method that aimed to increase the sensitivity as well as to improve the accuracy of the classification. The proposed method optimises the three performance measures by giving weights that composed of the proposed objective function. The results showed that the ensemble method is significant to achieve the highest accuracy of 76% with 84% sensitivity and 63% specificity for diabetic dataset from UCI medical data repositories. |
|---|