Abstract:
The abrupt explosion of the Ebola virus in 2018 in DRC was one of the world's most
widespread and deadliest epidemics with the highest number of casualties being reported in
the regions of NORTH KIVU and ITURI provinces. Ebola, a fatal hemorrhagic fever
syndrome, is caused by the Ebola virus (EBOV).
The World Health Organization proclaimed the disease as a world healthcare crisis. In most of
the cases, the patients are known to have died before the antibodies could respond. This
indicates the need to improve upon the diagnosis and prediction techniques available for this
disease. This work aims to analyze and improve upon the accuracy of the prediction systems
for the Ebola disease using several inputs. The input relies on the symptoms shown by the
patient during the early stages of the disease. The data mining techniques employed to carry
out this research include Decision Trees, KNN, Support Vector Machine, Random Forest,
Gradient Boosting classifier. The experimental results show the accuracy obtained by each
classification technique and the best predictive model for both diagnosis and prognosis was
Support Vector Classification that were applied to the dataset with 0,88 accuracy.
We will include these models into an Ebola prediction web app with an API in Flask(Python),
which will aid medical practitioners and people in the early diagnosis of illness.
