Pallavi B. Kanawade, Dhiraj V. Kawade, Anuradha V. Yenkikar |
Dept. of Comp. Scie. and Engg.,Vishwakarma Inst. of Information Tech., Pune, Maharashtra.
Abstract
Heart failure is a basic restorative condition with a noteworthy effect on open wellbeing, driving to considerable horribleness and mortality rates around the world. Early expectation and discovery of heart failure can help in opportune interventions and administration, possibly moving forward with quiet results. To this end, we utilize machine learning classification calculations to anticipate heart disappointment based on different clinical highlights. Particularly, we explore three prominent algorithms: calculated relapse, irregular timberland, and back vector machine. These computations have been prepared using a dataset with key information including age, blood pressure, cholesterol, and other crucial restorative guidelines. Through comprehensive experimentation and assessment, we survey the execution of every calculation regarding exactness, review, and F1 points. Moreover, we analyse the highlighted significance given by each calculation to pick up experiences into the variables contributing most altogether to heart failure expectation. Our findings illustrate the adequacy of machine learning procedures in foreseeing heart failure and give important bits of knowledge for clinicians in recognizing people at hazard, encouraging early intercession procedures, and making strides in understanding care. Apart from assessing the machine learning algorithms’ effectiveness, we also evaluate their interpretability by looking at the main clinical characteristics that have the biggest impact on heart failure prediction. We find that factors including age, blood pressure, and cholesterol are the main predictors in all models by examining feature importance scores. During patient assessments, clinicians can benefit from this analysis’s insightful explanation of the underlying risk factors for heart failure, which can help them prioritize crucial indicators. Additionally, the random forest model performs better in terms of accuracy and recall than logistic regression and support vector machines due to its capacity to manage non-linear relationships and complicated feature interactions.
Keywords: Healthcare Analytics, Heart Failure Prediction, Machine Learning, and Support Vector Machine, Logistic Regression
View PDF