Optimization of the Artificial Neural Network Algorithm with Genetic Algorithm in Stroke Prediction

Abstract

This study aims to optimize Artificial Neural Network with Genetic Algorithm in predicting stroke. This research is motivated by health problems in the community that are less considered that cause a disease such as stroke. Factors of lifestyle, poor diet and other factors that can be the cause of stroke. Therefore, where later the data that has been obtained will be processed to see what factors determine the cause of stroke. The data used, namely kaggle and mendeley, will be processed using RapidMiner, with a development method (CRISP-DM) and a testing method using a Confusion Matrix. The results of this study, stroke disease classification model accuracy kaggle Artificial Neural Network dataset with Genetic Algorithm accuracy 95.13% and AUC 0.667 and mendeley dataset accuracy 98.20% and AUC 0.712. For model evaluation with Artificial Neural Network algorithm with Artificial Neural Network algorithm with kaggle dataset genetic algorithm using X-fold validation average accuracy of 95.14% and AUC 0.686.7 and mendeley dataset resulted in accuracy of 98.20% and AUC 0.712.5. So as to produce from an algorithm a new attribute from the results of the classification model that has been carried out, namely heart disease, ever married, work type and residence type