Off-campus WSU users: To download campus access dissertations, please use the following link to log into our proxy server with your WSU access ID and password, then click the "Off-campus Download" button below.

Non-WSU users: Please talk to your librarian about requesting this thesis through interlibrary loan.

Access Type

WSU Access

Date of Award

January 2019

Degree Type


Degree Name



Industrial and Manufacturing Engineering

First Advisor

Celestine C. Aguwa

Second Advisor

Leslie F. Monplaisir


With the existence of cutting-edge technology and research and development present in the world today, there is elevated expectations that defective product could be minimized, if not eliminated. This has dramatically led to the inevitability of product recall in this era. The motivation of this research stems from an increase in software failure recalls throughout the years, with in-vitro diagnostics and imaging as the two main categories of software medical device recall, which results in faults going undetected before getting in contact with the end-user, increase in the number of injuries and death and increase in financial losses accrued by the recalling firm.

In this study, a framework comprising of text mining, latent semantic analysis and classification algorithms that predict the failure type experienced by the above-mentioned devices via the application of machine learning algorithms is developed by using the Food and Drug Administration Weekly Enforcement Report dataset. Four popular machine learning algorithms, the percentage-split method, and seven classifiers performance evaluation metrics such as classification accuracy, specificity, sensitivity, Matthews’ correlation coefficient, and execution time is used. The framework can easily identify and classify devices with control flow faults from those with integration fault. Furthermore, receiver optimistic curves and area under the curves for each classifier is computed.

The performance of the proposed system has been validated on full features, with an 80% on the training set, and the remaining 20% on the test set. The framework presented in this paper would act as a machine-learning-based decision support system that will assist the medical device manufacturers to detect medical devices with faults efficiently.

Off-campus Download