A HIV/AIDS viral load prediction system using artificial neural networks
Date
2017
Authors
Tunduny, Titus Kipkosgei
Journal Title
Journal ISSN
Volume Title
Publisher
Strathmore University
Abstract
Human Immunodeficiency Virus (HIV) has been affecting people since it was first discovered in 1986. This is as a result of the HIV virus being present in the patient bloodstream for the remainder of their normal life, as there is no cure that exists as of now. HIV, if left unmanaged would end up developing into Acquired Immune Deficiency Syndrome (AIDS), a syndrome that weakens a patient’s immune system and leaves them susceptible to other opportunistic infections. Antiretroviral therapy (ART) has been successfully used in managing the progression of the HIV virus in the human body. However, poor adherence attributable to ignorance, adverse drug effects, and age have derailed the attainment of viral load suppression amongst the HIV positive people. The progression of the virus is tracked by counting Cluster of Differentiation 4 positive cells, and the amount of virus in the blood (viral load) every 6 months. This research introduces the use of multi-layer artificial neural networks with backpropagation to predict the HIV/AIDS viral load levels over a given period of time (in weeks). The Data-driven Modelling methodology was used in the development of the model. This methodology was ideal since the model relied solely on pre-existing data, and supports artificial neural networks. The model developed performed at an accuracy level of 93.76% and a mean square error of 0.0323. The results showed that the neural network can be used as a suitable algorithm for HIV/AIDS viral load level prediction. The learning rate used in the study was 0.005 and the momentum was 0.9. The iterations for the training, testing and validation varied.
Description
Thesis submitted in partial fulfillment of the requirements for the Degree of Master of Science in Information Technology (MSIT) at Strathmore University
Keywords
Neural Networks, Viral Load, CD4+, Data-driven Modelling