Semi-supervised learning: Assisted cardiovascular disease forecasting using self-learning approaches

• Main Authors: Tusher, Ekramul Haque, Mohd Arfian, Ismail, Khan, Ferose, Anis Farihan, Mat Raffei, Md Akbar, Jalal Uddin
• Format: Article
• Language: English; English
• Published: Penerbit Akademia Baru 2024
• Subjects: QA75 Electronic computers. Computer science
• Online Access: http://umpir.ump.edu.my/id/eprint/39151/1/Semi-supervised%20learning_%20Assisted%20cardiovascular%20disease%20forecasting.pdf; http://umpir.ump.edu.my/id/eprint/39151/7/Semi-Supervised%20Learning-%20Assisted%20Cardiovascular%20Disease.pdf; http://umpir.ump.edu.my/id/eprint/39151/; https://doi.org/10.37934/araset.56.1.136150; https://doi.org/10.37934/araset.56.1.136150

Cardiovascular diseases (CVDs) are characteristics that affect both the heart and the blood vessels. This disease is the main factor contributing to the greatest number of deaths globally. In the present global context, it is very difficult to detect cardiovascular diseases by early-stage symptoms. If this isn't diagnosed early, it could lead to death. In order to improve the accuracy of the CVD prediction system, a wide variety of supervised and unsupervised learning approaches from the fields of machine learning were used. Only labeled data is used in supervised learning systems to create a classification model but acquiring sufficient amounts of labeled data takes time and typically requires the cooperation of field experts. However, unlabeled samples are readily available in a variety of real-world situations. More effectively than any other machine learning approach, semi-supervised learning (SSL) addresses this problem by integrating quantities of labeled and unlabeled data to improve the classification model. In this work, we propose semi-supervised learning approaches based on self-learning with Support Vector Machine (SVM), Naïve Bayes (NB) and Random Forest (RB). According to the comparison's findings, SVM has a high classification accuracy rate of 94.68%, a recall rate of 94.41%, a sensitivity rate of 94.49%, a F1 score rate of 92.99%, a precision rate of 91.59% a low, a balanced accuracy rate 94%, a G-mean rate of 94.45 and low Error-rate 5.32%. The model may be used to forecast cardiovascular disorders in the medical profession.