WALLAGA UNIVERSITY

SCHOOL OF GRADUATE STUDIES

INSTITUTE OF TECHNOLOGY
DEPARTMENT OF COMPUTER SCIENCE
PROGRAM MASTERS WEEKEND

PREDICTING AND FORECASTING ANTI-RETROVIRAL THERAPY

TREATMENT FAILURE AMONG HIV PATIENT USING ENSEMBLE
LEARNING APPROACHES : CASE OF NEKEME SPECIALIZED
HOSPITAL

A THIS PROPOSAL IS SUBMITTED IN PARTIAL FULFILLMENT OF

THE REQUIREMENT FOR THE DEGREE OF MASTER OF SCIENCE IN
COMPUTERSCIENCE

SUBMITED: BY Gadise Mangasha

ADVISORS : KEMAL.M(ASSISTANT PROFESSOR)

FEBRUARY , 2025
NEKEMTE, ETHIOPIA
 WALLAGAUNIVERSITY

SCHOOL OFGRADUATE STUDIES

P.O.BOX:395, NEKEMTE, ETHIOPIA.

Developing deep learning based predictive model for Network

intrusion detection system using CNN approaches.

Submitted By: Gadise Mangasha

1______________________ _____________ _____
Name of student Approved by Signature Date
2___________ ______________ ____________
Name of Major Advisor Signature Date
3___________ ___________ ____________
Name of Co-Advisor Signature Date

4.______________ ___________ ___________

Name of Chairman, DGC Signature Date

5 ______________ ___________

Name of Dean, SGS Signature Date

DECLARATION

I declare that this research titled Predicting and forecasting Anti-retroviral therapy

treatment failure among HIV patient using Ensemble learning approaches: In

Case Of Ambo Referral Hospital” is my original work, and has not been submitted as a
requirement for the award of any degree in the university or elsewhere.
PG Candidate: Gadise Mangasha Signature-----------------Date----------------
 ACKNOWLEDGEMENT
First and foremost I want to thanks the Almighty of God for helping me to complete my proposal
work Next, I would like to sincerely thanks my advisor Dr.______________ (PHD) and my Co-
advisor Dr.__________________ (PHD),for his guidance, support, insightful feedback, and
helpful suggestions beginning with the early phases of conceptual inception and continual
guidance that encourage me and overcome the challenges I encountered to complete my proposal
work. I could not finish this Research work without his sincere and competent advice. Moreover,
I give a special appreciation to all Computer Science department and the entire ART department
personnel at Nekeemte specialized Hospital. Finally, I would like to thanks, all my parent and
friend who encourage me in doing this proposal work.
 ABSTRACT
No a day’s , HIV/AIDS disease become an international public health problem. In Ethiopia, HIV
viruses have been widespread in different areas of the country. Many researcher were conducted
. a research on treatments of the virus using data mining and knowledge based system. However,
there still an issues of taking anti-retroviral therapy treatment among a patient who start
medication. Beginning ART treatment was significantly, decreasing HIV infection and death of
people by decreasing the patients viral load in their cells to measurable level and minimizing the
risk of HIV infection. Although antiretroviral therapy is now widely used , virological failure is
still a major concern for patients with HIV (WHO & 2021., 2022) .This research study utilizes a
retrospective cross-sectional data of patients who are at least three months on ART treatment
when data were collected from Nekemte specialized Hospital. This research study follows an
experimental research design methodology as it used an experimental research. The experimental
parts of this research proposed a prediction of ART failure. The ensemble learning approaches
was used comparatively to accurately predict the ART failure. We will proposed seven
classification model including, random forest, decision tree, naive Bayesian network, KNN,
neural network, support vector machine, gradient boasting and XB-boasting. This research work
will be conducted using a retrospective cross-sectional data of patient who were started ART
treatment from Neekemte specialized Hospital. The ART data set of a patient record will splited
as 80% the total record for training and 20% for testing respectively for cross validation
purpose. The research work is implemented using ensemble learning approach. We have
conducted a predictive model using ensemble learning approach to predict the success and
failure of ART treatment based on Adherence information of the patients at Nekemte specialized
Hospital. The Research funding is predicting and forecasting the ART treatment as success or
failure at Nekemte specialized Hospital based on Adherence of the patient taking medicine .The
experimental results of this research study will evaluated in terms of accuracy , precision , recall
and F1-score . These evaluation method would be different in the time taken to build the
predictive model. Decision tree algorithm would be considered as a better time efficiency. So ,
the decision tree model will be selected as the best algorithms to develop prediction model for
ART treatment success and failure.

Keywords: ensemble learning, random forest, classification algorisms, ART, machine learning,
Adherence.
Introduction

HIV/AIDS is the human immune deficiency virus and acquired immune deficiency syndrome
(Ghasabi FHA, 2023). It is still remains a global public health issue. As the Global HIV/AIDS
data states, about 38 million peoples are living with HIV, 28.7 million people were retrieving
antiretroviral therapy, 1.5 million persons are newly infected with HIV, and 650,000 people were
died (EPHI. & 2022). HIV/AIDS infection has become serious problem through the world ,
nearly 67% of HIV-positive people were live in Africa. nearly, 20.6 million people are living
with HIV in Eastern and Southern Africa, as well 16.2 million peoples are receiving
antiretroviral therapy.

About 78% of adults and children are using ART therapy and an estimated 617,921 persons are
living with HIV/AIDS. Ethiopia is amongst the worst nations in Sub-Saharan Africa by the HIV
epidemic (EPHI., 2022). The Amhara region has the biggest number of HIV patients, at over
191,480, while the Oromia, SNNP, Tigray, and beneshangul gumuz regions account for more
than 60% of new infections. In Ethiopia, the adult HIV prevalence is 0.9%. Resistance and an
elevated risk of viral infection may result from an inability to attain and sustain viral suppression
(EPHI., 2022). Consequently, persons with HIV are more likely to experience the progression of
their disease and pass away if early detection of virological failure is delayed. Therefore, an
effective antiretroviral therapy lowers HIV-positive patients' incidence and death.

The Antiretroviral therapy or ART is the method of using the HIV/AIDS treatment medicine.
UNAIDS nation expected that by the end of 2020, 90% of peoples are living with HIV/AIDS
and saw their HIV status, 90% of those who saw their HIV-positive status were on treatment, and
90% of those on treatment had suppressed viral load ART treatment method was significantly
decreasesed (approach., 2016). HIV infection and death was reduced the patients’ HIV viral
status in their normal states to undetectable levels using ART treatment. For most people living
with HIV who never start ART before, first-line ART treatment regimens is important drugs
that are used for safe, effective, and convenient recommendation. However, virological failure
occurs over 24% of patients within a year after beginning treatment. Since the objective is not
achieved, UNAIDS nation improved it to 95-95-95 targets to end the HIV/AIDS pandemic by
2030 For the effectiveness of ART, patients have to take medication every day and regularly see
a doctor for the entirety of their life (UNAIDS, 2022).
Antiretroviral therapy is the use of antiretroviral drugs to safe HIV infection. Patients takes a
combination of anti-HIV medicines called an HIV regimen every day. ART medicine cannot
cure HIV, but it can help people to live long with HIV. ART treatment also reduces the risk of
HIV transmission.

Programs for HIV/AIDS prevention and control must comprehend the factors that contribute to
virological failure in individuals receiving antiretroviral medication (ART). Clinical prediction
models that employ patient-level evidence can also be used in medical decision-making to
reduce these negative effects (UNAIDS, 2022).

Various Machine learning (ML) methods have been used to identify and discover patterns in
complex datasets and predict future outcomes of HIV treatment with good predictive
performance (Young SD & 74, 2017). By using ensemble learning methods we predict the CD4
count and their predictors between patients suitability for ART clinics, the termination of ART
among patients, ART-induced dangerous , viral status , patient Adherence on ART treatment and
also to predict estimate survival of HIV/AIDS patients (Maskew M & 12(1):12715., 2022).
Traditional statistical modeling, on the other hand, was developed for data with small to medium
sample sizes and a few dozen input variables. Traditional statistical reasoning may be more
difficult to handle due to the complexity of the data (Bzdok D & 15(4):233–4., 2018).
Consequently, more researchers study the health sciences using machine learning techniques.

The purpose of this proposal work is developing anti-retroviral therapy treatment failure
predictive model using an ensemble learning approach. We will conduct a predictive model
using seven machine learning algorithm to predict HIV patients treatment failure.

Motivation of the study

It is important to use the current technology called machine learning method to inform
people who delays in starting ART especially, those who have major repercussions, particularly
for individuals who are at high risk of death due to significant immunosuppression or
tuberculosis. HIV/AIDS prevention and control should be used the latest trends such as, AI,
ML and ensemble learning to understand the factors that affects virological failure of HIV
patients taking antiretroviral treatment (ART) (Supriya M & 5(1):58–75., 2020). Since HIV virus
is an epidemic disease and shortening human life, we should implement a predictive model
using the current and most recent advancement such as ML, ensemble learning and DL to
improves the utilization of patient retroviral therapy treatment.

Previously, various researchers and scholars conducted a research on ART medication.

However, the researcher does not include all attributes and features needed for ART
treatment failure in their research while implementing their research on the cases of the
patient who start ART medication. The HIV epidemics virus affects economic growth of
our country and also death of people from time to time. Understanding the reasons that lead
to virological failure in HIV patients on antiretroviral therapy (ART) is a vital for HIV/AIDS
prevention. Also, machine learning is used to detect HIV patients who are at high risk of failing
to adhere and being a virological failure (Maskew M & 12(1):12715., 2022).

Therefore, we have motivated to do this proposal using the recent technological

advancement including, machine learning ad ensemble learning techniques to develop a
predict model for anti-retroviral therapy treatment of the patient.

Statement of the problem

Based on the Global HIV/AIDS Figures, about 38 million people are living with HIV/AIDS,
28.7 million people were retrieving antiretroviral therapy, and about 1.5 million individuals are
newly infected person with HIV, and 650,000 people died due to the epidemic disease around the
world in 2021 (UNAIDS & from:, 2022). United AIDS expected that by the end of 2030, about
90% of individuals were living with HIV knows as they live with HIV and their status also,
moreover, 90% of those who knew their HIV-positive level were on treatment, and 90% of those
on treatment had suppressed viral load (UNAIDS J. U., 2022). In addition , Ethiopia is one of the
most harmed countries in Sub-Saharan Africa by the HIV epidemic, by an estimation of 617,921
people living with HIV and about , 78% of adults and children receiving ART (UNAIDS &
from:, 2022).

The Virological failure is a vital problem faced by HIV patients after starting treatment. WHO
rules recommend that virological failure should not above 10% after the beginning ART
treatment (Organization & 2016., 2016). HIV patients who knows virologic failure were at
increased risk of clinical progression to AIDS and death compared with patients who had a
complete virological response (Organization & from:, 2022). So, with late early prediction of
virological failure, various patients with HIV/ AIDS are further to have sickness progression and
death (Bernabé KJ & 9(5):ofac121., 2022). consequently, a real antiretroviral therapy treate
Previously, there have been many researchers and scholars conducted a research on ART failure
prediction. However, the researchers only conducted a research using KBS , rule based and case
based method and this generates low accuracy rate on ART failure prediction. The research study
conducted by masafint Dagnachew using Hybrid knowledge based clinical decision support
system using Rule based and Case based reasoning has a lots of limitation including, the
researcher employed rule based and case based reasoning to predict the ART clinical information
using prolog tools and the total accuracy result of the model was also too low (D. Mesafint and
D. H. Manjaiah, 2019). Numerous investigations have revealed that virological failure in HIV
patients that can be influenced by socio demographic predictors (age, gender, marital status,
religion), clinical predictors (low BMI, TB co-infection), and treatment-related predictors
(excessive dose frequency, poor medication adherence). On the other hand in previous studies,
conventional statistical modeling was developed for data with small to medium sample sizes and
a few dozen input variables.

Traditional statistical reasoning could be more difficult to handle due to the intricacy of the data
management and mortality in HIV-positive patients. However, traditional statistical modeling
was created for data with a few dozen input variables and sample sizes that would be considered
small to medium size. Beside these facts, various researchers were used data from different
Hospital have only used statistical methods to recognize factors of viral failure (Mulisa D &
19(1):1–13., 2019). Although prior studies have been employed a research to investigate the
factors of virological failure in various regions of Ethiopia, most of them were applied classical
statistical models to identify a vital predictors of virological failure (Assemie MA & 4(1):1–10.,
2019).

Therefore, advanced predictive and forecasting modeling is needed to improve the targeting of
interventions through differentiated care models, to be more cost-effective, to improve patient
outcomes, and to initiate treatment based on achieving maximum patient survival and clinical
benefit. Hence, a machine learning techniques was used more efficiently . Machine learning
approaches can classify and discover patterns in complex datasets and predict future of HIV
treatment outcomes with excellent predictive performance (Supriya M & 5(1):58–75., 2020).
Machine learning predictive algorithms have the potential to improve the quality of care and
predict the needs of HIV patients by analyzing huge amounts of data, and enhancing prediction
capabilities
Research question
1. What are the determinant factors affecting ART treatment failure
2. What ensemble learning models we conduct to predict and forecast virological failure?
3. Can ensemble learning approach effectively predict and forecast antiretroviral therapy
treatment failure ?
4. How can we evaluate the performance of ensemble learning model for effective ART failure
treatment prediction and forecast?
Objective
The objective of this research study is stated as follow.
General objective
The general objective of this research is to predict and forecast Antiretroviral therapy treatment
failure among HIV patient using ensemble machine learning techniques : case of Nekemte
specialized Hospital.

Specific Objectives

To identify determinant factors affecting HIV patients virological failure in Nekemte

specialized Hospital.
To conduct preprocessing task including, removing inconsistence data, noise and
redundancy.
To build predictive model using ensemble learning models for virological failure prediction.
To evaluate the performance of an ensemble learning predictive model for its relevancy and
accuracy.

Scope of the study

The scope of this research work is implementing Anti-retroviral therapy treatment failure using
ensemble learning model only in Nekemte specilized Hospital. This research work only uses
institution based cross sectional dataset from Ambo Nekemte specialized Hospital, and Wollega
university referral Hospital . The research study scope is limited to predicting an Adherence of
HIV patients taking anti-retroviral therapy at specialized Hospital. The study considered domain
experts who faced challenges in managing Adherences of people with HIV AIDS. This study
focuses on side effects of antiretroviral drugs and considered modeling and representing the
ensemble machine learning techniques, rather than data mining and rule based techniques. This
is due to high prevalence of adverse drug reaction among HIV infected person on antiretroviral
therapy.

The task involved in conducting this proposal work is including literature review, data set
collection, data, preprocessing , model training, representation , implementation and evaluation.
The ensemble learning model was conducted to analysis the data set including, preprocessing,
attribute selection and classification . In this research, we focused on ensemble learning approach
that integrates different individual models and classifiers and finally, combines the results of
each models to produce model result with high and better accuracy than individual model result.

1.7.Significance of the Study

This study has vital significance to the body of knowledge, policy maker and practice of ART
treatment. In Ethiopia, a very few studies have been done to predict ART failure using
knowledge based method. This research work uses an ensemble learning method to increase our
understanding of ART failure diagnosis. Additionally, the investigation was encourage more
researchers into Artificial intelligence based applications in related and other Ethiopian health
care sectors.

The ability of machine learning to improve targeting of interventions through differentiated

models of care, increase cost-effectiveness, and improve patient outcomes will be understood by
decision- and policy-makers relevant to HIV/AIDS care. This understanding will help them
design and implement appropriate ML policies and strategies on HIV/AIDS patient data.

The results of the proposed study can be employed as a treatment-decision support tool in
clinical practice, which will aid in the identification of ART treatment failures and improve the
accuracy of ART diagnoses. Consequently, this will strengthen the evaluation of HIV patients
who are susceptible to ART failure, with the aim of quickly identifying and efficiently executing
ART therapies. The system might provide timely, complete, consistent and relevant information
concerning support, treatment and follow-up of HIV/AIDS patient treatement. This study builds
an ensemble learning classification algorithms to make clinical decision system for HIV patient
ART treatment failure.
 CHAPTER TWO
REVIEW OF LITRETURE AND RELATED WORK