| dc.description.abstract | Machine learning techniques offer powerful
tools for early prediction and diagnosis of Coronary Artery
Disease (CAD), a major cause of global mortality, by
analyzing complex medical data to recognize patterns that
conventional methods might miss, potentially leading to
timely interventions and reduced death rates
.
The objective
of this study is to model an Artificial Neural Network for
CAD diagnosis, achieving an accuracy of roughly 90%. The
Synthetic Minority Over-sampling Technique was employed
to address the problem of data imbalance. Leveraging the
expertise of a cardiologist, we tuned the feature search to
include key clinical characteristics only, such as basic
demographics, clinical history, and diagnostic test results.
This approach made the interpretations more feasible, and
at the same time, improved the predictive efficiency of the
model. Optimized tuning of the hyper parameters was done
while designing the Artificial Neural Network and the use of
dropout layers as an anti- over-fitting technique and batch
normalization technique for stabilizing the training phase.
To support the created model cross validation at a lower
level named stratified k-fold cross validation was done. The
model was accurate and reliable compared to conventional
machine learning methods. Moreover, it was beneficial to
combine clinical knowledge in qualitative feature reduction
with the presented technical approach because it led to a
more clinically pertinent and accurate model. | en_US |
