Selected Publications

Bariatric surgery (BAR) has become a popular treatment for type 2 diabetes mellitus (T2DM) which is among the most critical obesity-related comorbidities. Patients who have bariatric surgery, are exposed to complications after surgery. Furthermore, the mid- to long-term complications after bariatric surgery can be deadly and increase the complexity of managing safety of these operations and healthcare costs. Current studies on BAR complications have mainly used risk scoring for identifying patients who are more likely to have complications after surgery. Though, these studies do not take into considera-tion the imbalanced nature of the data where the size of the class of interest (patients who have complications after surgery) is relatively small. We propose the use of imbalanced classification techniques to tackle the imbalanced bariatric surgery data: synthetic minority oversampling technique (SMOTE), random undersampling, and en-semble learning classification methods including Random Forest, Bagging, and AdaBoost. Moreover, we improve classification performance through using Chi-Squared, Information Gain, and Correlation-based feature selection (CFS) techniques. We study the Premier Healthcare Database with focus on the most-frequent complications includ-ing Diabetes, Angina, Heart Failure, and Stroke. Our results show that the ensemble learning-based classification techniques using any feature selection method mentioned above are the best approach for handling the imbalanced nature of the bariatric surgical outcome data. In our evaluation, we find a slight preference toward using SMOTE method compared to the random undersampling method. These results demonstrate the potential of machine-learning tools as clinical decision support in identifying risks/outcomes associated with bariatric surgery and their effectiveness in reducing the surgery complications as well as improving patient care.
, 2017.

The computational complexity of solving nonlinear support vector machine (SVM) is prohibitive on large-scale data. In particular, this issue becomes very sensitive when the data represents additional difficulties such as highly imbalanced class sizes. Typically, nonlinear kernels produce significantly higher classification quality to linear kernels but introduce extra kernel and model parameters which computationally expensive fitting. This increases the quality but also reduces the performance dramatically. We introduce a generalized fast multilevel framework for regular and weighted SVM and discuss several versions of its algorithmic components that lead to a good trade-off between quality and time. Our framework is implemented using PETSc which allows an easy integration with scientific computing tasks. The experimental results demonstrate significant speedup compared to the state-of-the-art nonlinear SVM libraries.
In ArXiv, 2017.

The support vector machine is a flexible optimization-based technique widely used for classification problems. In practice, its training part becomes computationally expensive on large-scale data sets because of such reasons as the complexity and number of iterations in parameter fitting methods, underlying optimization solvers, and nonlinearity of kernels. We introduce a fast multilevel framework for solving support vector machine models that is inspired by the algebraic multigrid. Significant improvement in the running has been achieved without any loss in the quality. The proposed technique is highly beneficial on imbalanced sets. We demonstrate computational results on publicly available and industrial data sets
In ESSAN 2017, 2017.


  • McAdams Hall Office 224. McMillan Rd, Clemson, SC 29631
  • Email for appointment