Last modified: 2018-07-07
Abstract
Osteoarthritis (OA) is a long-term chronic disease characterized by the deterioration of cartilage in joints which results in bones rubbing together and creating stiffness, pain, and impaired movement. One way to prevent this disease is to do early detection using machine learning. Machine learning uses mathematical algorithms implemented as computer programs to identify patterns in large datasets and to iteratively improve in performing this identification with additional data. One type of machine learning is supervised learning for classification. In this study will be used Adaptive Boosting (AdaBoost) and Support Vector Machines (SVM) together as classifiers.
Â
SVM aims to resolve the classification problem with forming a hyperplane that maximizes the margin (the closest distance between training data and hyperplane) by dividing the two classes of data and solve the mathematical model as follows
EQUATION (1) (PLEASE SEE THE ABSTRACT FILE)
AdaBoost aims to maintain a weight distribution EQUATION (2) (PLEASE SEE THE ABSTRACT FILE) of base classifier (SVM is base classifier in this paper) iteratively. The ensemble method is called AdaBoost SVM. AdaBoost SVM can be used to improve performance and get higher accuracy.
Â
The purpose of this study is to see whether AdaBoost SVM can produce good accuracy with SVM as a comparison. The data used for training is Osteoarthritis patient data which is checked by MRI T2Map. The data is divided into three classes which are not severe, severe, and very severe. Tests were conducted using 10% until 90% training data. SVM and AdaBoost SVM classification results as follows:
Â
Table 1. Results of OA severity classification using SVM and AdaBoost SVM. (PLEASE SEE THE ABSTRACT FILE)
Â
From the result above, we can see that the highest accuracy value of SVM is 75% in 90% training data, while the highest accuracy value of AdaBoost SVM is 85.714% in 80% training data. It can also be seen that AdaBoost SVM improved the accuracy of SVM. So we conclude that AdaBoost can improve the performance of SVM by maintains the weight distribution thus increasing the accuracy. AdaBoost SVM has better accuracy than SVM in the classification of Osteoarthritis disease severity.