Application of data mining techniques to developing a classification model for glaucoma type identification

Abstract

Data mining, also known as Knowledge Discovery in Databases (KDD), is a process that entails extracting valuable, interpretable, and useful information from raw data. Glaucoma, characterized by an elevation in intraocular pressure (IOP), leads to glaucomatous optic neuropathy and subsequent loss of retinal ganglion cells and their axons, ultimately resulting in blindness. Those tasked with treating glaucoma patients may face challenges in accurately identifying the type of glaucoma and prescribing appropriate treatment, often due to subjective decision-making, limited knowledge, and reliance on instrument visualization. These challenges contribute to resource wastage and time-consuming processes. The primary goal of this research is not to completely eliminate the problem but to alleviate biased decisions made by ophthalmologists. This is accomplished by developing an easily accessible method for identifying glaucoma types through the creation of an improved classification model. In this study, data mining techniques are employed to unveil new knowledge based on the collected dataset. Among various data mining classification algorithms, this paper utilizes naïve Bayes, GRIP, J48, and PART algorithms, along with two test options involving complete and selected features. According to the empirical analysis conducted, the PART algorithm, with a 10-fold cross-validation test option using selected features, yielded the highest accuracy result, reaching 71.4%.