Abstract:
Glaucoma is a type of eye disease characterized by progressive optic neuropathy and visual field
defect and categorized by damage to the optic nerve with corresponding visual field loss. It affects
through increasing an intraocular pressure (IOP), which is responsible for the glaucomatous optic
neuropathy involving the death of retinal ganglion cells and their axons which resulting in
blindness. A person, who is responsible to treat glaucoma patients might make an error in
glaucoma type identifying and treatment ordering due to subjective decision, knowledge
limitation, and visualization through instruments. This results in resource wastage as well as time consuming. The main aim of this research is to reduce (not removing the problem at all) the biased
decision of the ophthalmologist through making an easy, quick and accessible way for glaucoma
type identification and order the treatment for each type by developing a better classification
model. In this study, data mining techniques are used to discover new knowledge based on the
dataset collected from Borumeda hospital. From numerous data mining classification algorithms,
this research applied Naïve Baye, Jrip, J48, and PART algorithms that are used with two basic test
options based on complete and selected features. Based on the experimental analysis performed,
based on the dataset collected from Borumeda hospital with an instance number of 3535 with
fourteen attributes under four class labels (namely CG,PACG,POAG and SG ), the PART
algorithm with a test option of 10 fold cross-validation using the selected feature scored the highest
accuracy result which is 74.7%. Finally, Then the output of the data mining result achieved through
a classification model based on the PART algorithm incorporated with the knowledge base
represented through a rule-based system and a personalized prototype system for type identifying
and treatment recommendation of glaucoma ailment is develop. Last of all, the system that was
developed through this methodology achieved a remarkable performance by obtaining a user
acceptance score of 82.2%. Which, indicate it achieves a good result for classifying and order
recommended treatment for each classified types of glaucoma ailment. Based on this finding, in
futurity an ensemble method with a hybrid model will recommend to score a good performance
outcome. Nonetheless, further researches should be done to upsurge the merits of combining Data
mining persuaded knowledge with knowledge base system.
