Paper Type
Completed Research Paper
Abstract
This research focuses on presenting an empirical method to gather necessary data and then developing several models to predict the chance of diabetic retinopathy (proliferative and non-proliferative) by observing HbA1c, duration of disease and albumin excretion rate of diabetic patients. We gathered required knowledge from other studies that have investigated the relation of different risk factors and complications in diabetes. In order to create 1-1 models, curve fitting was performed by using two different software applications: Tiberius (Brierley 2011) and SPSS (IBM 2010), which work based on ANN and least square regression, respectively. To start producing the model, seven different patterns, i.e. linear, logarithmic, quadratic, cubic, power, s and exponential, have been chosen as the best regression options. Using R-squared, it can be clearly seen that the best selected regression models fit the data in all the dataset tables better than ANN, as well as the other six regression patterns.
Recommended Citation
Sangi, Mohsen; Win, Khin Than; Sharvani, Farid; and Fulcher, John, "Model Development for Prediction of Diabetic Retinopathy" (2013). AMCIS 2013 Proceedings. 14.
https://aisel.aisnet.org/amcis2013/HealthInformation/GeneralPresentations/14
Model Development for Prediction of Diabetic Retinopathy
This research focuses on presenting an empirical method to gather necessary data and then developing several models to predict the chance of diabetic retinopathy (proliferative and non-proliferative) by observing HbA1c, duration of disease and albumin excretion rate of diabetic patients. We gathered required knowledge from other studies that have investigated the relation of different risk factors and complications in diabetes. In order to create 1-1 models, curve fitting was performed by using two different software applications: Tiberius (Brierley 2011) and SPSS (IBM 2010), which work based on ANN and least square regression, respectively. To start producing the model, seven different patterns, i.e. linear, logarithmic, quadratic, cubic, power, s and exponential, have been chosen as the best regression options. Using R-squared, it can be clearly seen that the best selected regression models fit the data in all the dataset tables better than ANN, as well as the other six regression patterns.