Mountain Top University Library

This study is aimed at the application of decision trees algorithm for the
classification of the survival of patients living with hepatitis C virus. The study
achieved this by eliciting knowledge on the variables monitored during the treatment
of Hepatitis C patients and collecting relevant data, formulating the classification
model, simulating the model and validating the model.
Structured interview was used to elicit information on the variables monitored
among Hepatitis patients receiving treatment. C4.5 decision trees algorithm was used
to formulate the classification model for the survival of Hepatitis C disease based on
the information collected. The classification model was simulated using Waikato
Environment for Knowledge Analysis (WEKA) software. The classification model was
validated based on accuracy, sensitivity, false alarm rate and precision.
The results of the study showed that 19 variables were associated with the
classification of the survival of patients receiving treatment for Hepatitis C. The results
also identified a number of variables which had missing values from the data collected.
The results of the simulation showed that the higher the proportion of datasets use for
training then the higher the performance of the model on the testing dataset. Therefore,
using 90% of the dataset for training and 10% of the dataset for testing showed the
highest performance. The results also showed that twelve rules were inferred from the
decision trees which was formulated from the simulation with the best performance.
The results also showed that among the 19 variables identified, 11 variables were the
most relevant, namely: ascites, age, spiders, bilirubin, liver firmness, sgot, palpable
spleen, anorexia, and liver size.
The study concluded that the classification model developed can be adopted for
use by physicians in locations where there are no experts to provide as decision support
vi
thus improving survival which may mitigate untimely deaths. The study also concluded
that the identification of a smaller yet relevant number of variables will reduce the
challenge of focusing on irrelevant variables monitored thus reducing the response time
of data collection and survival classification.