The Use of Neural Network Analysis to Predict Stroke Occurrence

Abstract

This research work employed machine learning for prediction of upcoming strokes in patients. In this study, critical lifestyle related attributes were first gathered through an all-inclusive survey instrument. Moreover, objective medical tests, namely routine blood examinations and ECG, were conducted to collect more relevant health information of the patients. Afterward, gathered information was utilized for training proposed multilayer perceptron neural networks (MLPNN). Neural network model was employed because it is observed that it can effectively capture the interaction between different lifestyle and medical attributes. Consequently, it outperforms several conventional machine learning strategies devised with the recorded information. Comparative results are found very promising and it is anticipated that an analogous networking model can be easily absorbed in portable devices for the early prediction of strokes associated with critical lifestyle and health risks. Rapid advancements in medical knowledge are providing multiple new opportunities for the diagnosis and treatment of several serious health conditions including stroke. Nonetheless, fluent deals among lifestyle trends and medical features necessitate extremely customized medications. In this paper, data collected from an extensive survey conducted on adult stroke patients from a multi-ethnic context have been analysed using different launched and topological properties in the form of observed disputes concerning previous studies on similar datasets have been further investigated. A predictive neural networking framework has been thereby developed aiming to resolve the disagreement and forecasting upcoming strokes associated with critical lifestyle and medical attributes. Implementing the devised approach, nearly 63% of the strokes are anticipated 3 months beforehand.