TY - JOUR
TI - Heart failure risk prediction using azure data lake architecture with automated machine learning and machine learning approaches
AU - Alghamdi Ahmed Mohammed
AU - Al Shehri Waleed
AU - Almalki Jameel
AU - Jannah Najlaa
AU - Bahaddad Adel A
AU - Bokhary Abdullah M
JN - Thermal Science
PY - 2024
VL - 28
IS - 6
SP - 5059
EP - 5069
PT - Article
AB - Cardiovascular disease is a chronic disease that is a leading cause of death  due to heart failure and blood stroke. The WHO records 17.9 million deaths  yearly due to heart-related diseases. Heart failure occurs worldwide,  especially having a significant impact in low and middle-income countries.  Early diagnosis of heart disease is needed because a patient can face  serious complexities if it is detected in the later stages of disease  progression. In addition, if heart disease is identified early, it is likely  to be cured. On the other hand, symptom identification of heart failure is  necessary for an accurate and optimum solution. The model reported in this  paper suggests a solution for the early diagnosis of heart disease. First,  data analysis is performed, and pre-processing approaches are applied to  prepare the dataset for model training. Raw data has noise and missing  values, which are treated correctly before being passed to the model.  Second, two types of algorithms are trained for the proposed solution.  Traditional machine learning algorithms are used in the form of support  vector machine, k-nearest neighbors, logistic regression, random forest,  artificial neural networks, decision tree, xgboost, and catboost to train  and test the model. In parallel, automated machine learning (AutoML) with an  Azure machine learning cloud instance is used for model training and  testing. Azure data lake cloud storage is utilized for model training and  running the AutoML process. Finally, the performance of the models was  evaluated using a University of California Irvine (UCI) machine learning  open-source dataset for heart failure diagnosis. The AutoML outperformed  when compared with traditional algorithms. The highest accuracy value  obtained for the best machine learning algorithm was xgboost, with an  accuracy of 82.22%, whereas the accuracy value obtained using AutoML was  88%. The proposed model can be used for clinical purposes due to its  performance and the approach applied.