AIB-7 Use of Deep Learning Algorithms in Improving Stroke Prediction Accuracy using Multi-layer Perceptron (MLP) Neural Networks
Abstract
The use of Artificial Intelligence (AI) and Machine Learning (ML) algorithms have been a popular tool as a predictive instrument in the modern world of medicinal science. Many machine learning algorithms, such as classification and logistic Regression, have been frequently used in achieving efficient and accurate predictions using several general parameters such as a patient’s age, weight, body mass index, smoking status, and gender. While those machine learning algorithms were successful in producing effective accuracy scores, there still remain several missing layers in the use of ML predictive algorithms in today’s current medical space. In this study, we propose the use of deep learning algorithms, a subset of machine learning, which uses artificial neural networks to mimic the behavior of the human learning process. We implemented the use of Multi-layer Perceptron (MLP) neural networks on our real-world stroke patient dataset of over 7,000 patients that included several unique parameters such as total cholesterol, triglycerides, blood glucose level, heart rate, and blood pressure rates. The use of deep learning algorithms, such as MLP neural networks, showed to be a more efficient tool when compared to its machine learning counterparts, as the MLP algorithm was able to use a larger dataset with more complex variables and parameters.
Keywords
artificial intelligence, neural networks, machine learning algorithms, deep learning algorithms, multi-layer perceptron, stroke predictions
AIB-7 Use of Deep Learning Algorithms in Improving Stroke Prediction Accuracy using Multi-layer Perceptron (MLP) Neural Networks
University Readiness Center Greatroom
The use of Artificial Intelligence (AI) and Machine Learning (ML) algorithms have been a popular tool as a predictive instrument in the modern world of medicinal science. Many machine learning algorithms, such as classification and logistic Regression, have been frequently used in achieving efficient and accurate predictions using several general parameters such as a patient’s age, weight, body mass index, smoking status, and gender. While those machine learning algorithms were successful in producing effective accuracy scores, there still remain several missing layers in the use of ML predictive algorithms in today’s current medical space. In this study, we propose the use of deep learning algorithms, a subset of machine learning, which uses artificial neural networks to mimic the behavior of the human learning process. We implemented the use of Multi-layer Perceptron (MLP) neural networks on our real-world stroke patient dataset of over 7,000 patients that included several unique parameters such as total cholesterol, triglycerides, blood glucose level, heart rate, and blood pressure rates. The use of deep learning algorithms, such as MLP neural networks, showed to be a more efficient tool when compared to its machine learning counterparts, as the MLP algorithm was able to use a larger dataset with more complex variables and parameters.