Abstract:
Predictive modeling plays a vital role in stroke prediction, enabling timely intervention
in healthcare. It is inevitable that any form of prediction especially in the prediction of
stroke would require some form of predictive modeling that empowers timely
healthcare intervention. This work can be used for proposing evaluation criteria of the
stroke prediction model concerning the dataset size and feature influence. Two datasets
were used: the first one with 5000 data points, the program had a 90% accuracy while
the second with 1500 data points had 50% accuracy. Such transitions merged increased
the datasets to a combined percentage of 60% to the other’s benefit, proving the
importance of diverse datasets. Other measures pointing to the exploration of the feature
importance analysis include Average Glucose Level and Body Mass Index (BMI) that
are important in the accurate prediction of strokes. Therefore, these results present task,
data set attributes, and the importance of features as crucial factors to be taken into
account to build stable models. The nature of change in the difference in model
accuracy between two datasets shows that large samples produce higher model
accuracy. Future studies are likely to reveal improved models for different datasets and
different subpopulations in order to ultimately enhance process accuracy and
consequently stroke predictability as well as its effects on its patients. Understanding
these factors can help enrich such determinants among the health-care providers which
in turn would lead to improved ways of intervention and management of the risks
associated with stroke. As such, this study emphasis the possibilities of predictive
modeling to enhance the efficiency and effectiveness of the stroke prevention and care
by providing accurate and reliable predictions.
